Home

correlation between the flicker and the microwave noise of gallium

1. doublem Vr doublem Vg doublem doublem R doublem P2 doublem P1 doublem FStop doublem FStart BOOLm CheckAverage int m Radio AveragingMode BOOLm CheckLogAverage CString m TimeToGo AFX DATA ClassWizard generated virtual function overrides AFX VIRTUAL CFlickerNoiseMeasurementDlg protected virtual void DoDataExchange CDataExchange pDX DDX DDV support AFX VIRTUAL Implementation protected HICON m hIcon Generated message map functions AFX MSG CFlickerNoiseMeasurementD1g virtual BOOL OnInitDialog fx msg LONG OnStartCountdown WPARAM Seconds LPARAM lParam msg LONG OnDisplayStatus WPARAM Message LPARAM Status fx msg LONG OnSaveData WPARAM wParam LPARAM lParam msg LONG OnMeasurementStoped WPARAM wParam LPARAM lParam msg void OnSysCommand UINT nID LPARAM lParam fx msg void OnPaint fx msg HCURSOR OnQueryDragIcon msg void TOBIAS WERNER msg void OnKillfocusEditR msg void OnSetfocusEditFstart fx msg void OnSetfocusEditFstop msg void OnSetfocusEditPl msg void OnSetfocusEditP2 x msg void OnSet focusEditR 51 51 51 x msg void OnSet EditVd x m
2. PostMessage pApp m hWnd WM MEASUREMENT STOPED 0 0 return 0 Start pApp gt m_AveragingRanges CurrentLimit INDEX Frequency pApp gt m_NumberOfPassedFrequencyRangest break default start and stop are above the frequency limit break do nothing PostMessage pApp m hWnd WM SAVE DATA 0 0 Call save data return 0 Thread finished successfully bool CFlickerNoiseMeasurementDlg DoMeasurement double StartFrequency double StopFrequency int Averaging CFlickerNoiseMeasurementDlg int FrequencySpanIndex TimeForSettling TimeForMeasurement double FrequencySpan FrequencySpanIndex pApp gt GetFrequencySpan StartFrequency StopFrequency sprintf buffer FrequencySpanIndex 0 100 000Hz if pApp gt SendDataSR770 buffer return FALSE sprintf buffer STRFS 2 StartFrequency 0 100 000Hz if pApp gt SendDataSR770 buffer return FALSE FrequencySpan 100000 pow 2 19 FrequencySpanIndex TimeForSettling int 400 FrequencySpan if TimeForSettling 0 PostMessage pApp gt m_hWnd WM DISPLAY STATUS WPARAM Frequency range set Waiting for FFT settling to finish LPARAM STATUS MEASURE i PostMessage pApp m hWnd WM START COUNTDOWN WPARAM TimeForSettling 0
3. CLPTSwitchRemoteApp BEGIN MESSAGE CLPTSwitchRemoteApp CWinApp AFX MSG MAP CLPTSwitchRemoteApp NOTE the ClassWizard will add and remove mapping macros here DO NOT EDIT what you see in these blocks of generated code AFX MSG ON COMMAND ID HELP CWinApp OnHelp END MESSAGE MAP CLPTSwitchRemoteApp construction CLPTSwitchRemoteApp CLPTSwitchRemoteApp TODO add construction code here Place all significant initialization in InitInstance The one and only CLPTSwitchRemoteApp object CLPTSwitchRemoteApp theApp CLPTSwitchRemoteApp initialization BOOL CLPTSwitchRemoteApp InitInstance Standard initialization If you are not using these features and wish to reduce the siz of your final executable you should remove from the following the specific initialization routines you do not need ifdef AFXDLL Enable3dControls Call this when using MFC in a shared DLL else Enable3dControlsStatic Call this when linking to MFC statically fendif CLPTSwitchRemoteDlg dlg m pMainWnd amp dlg int nResponse dlg DoModal if nResponse IDOK TODO Place code here to handle when the d
4. 109 E 112 Fe MATEAB PROGRAM 114 VERIBICATIONOE o mex 115 BoD IVAW AB 6 116 APPENDIX MICROWAVE SWITCH CONTROL eee ee eee ee 121 Cr MICROWAVE Bero ae eu EAE d a 122 nuce 123 GZ A SCHON 123 TT NR TEE rcr DEN 23 OZS AT A CW CU DOPL 126 CAN SOFTWARE A amen teed 126 ERROR MI EI DILE 127 3 311 Header LPT Switch Remote eds 127 G 3 1 2 Header file LPTSwitchRemoteDlg h 2 127 G 3 1 3 Implementation file 5 4 000 128 G 3 1 4 Implementation file 5 129 APPENDIX 132 E 132 H2 INDEX qe 136 EOVA TION INDE e uta dL Eid 137 139 IBN er 139 0 2 TION 141 L5 MEASUREMENT 2 a Ut HE 141 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE
5. 71 252 LOVI SI HO 71 Character SOS Gea 22 uU E 73 TOS 74 DOT Typicat materia DEODEPES a 74 D32 LOVE 74 DC harde ADD tue 75 DSW arera asi eset a 76 DECA NEEE T 77 Dos T Typical MATCH DV OP 77 Ld LOVE STU CTT esed ete 727 78 D 4 gain Parameters 79 DAD WGN CP IN GD TEES 80 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 3 141 APPENDIX E LOW FREQUENCY NOISE MEASUREMENT 81 Il SMOOTHING OF THE GRAPH en Uxor 81 2 DESIGN OF THE PROGRAM 83 Re 85 I d PROGRAM C ODE a t 86 E 4 1 Header file FlickerNoiseMeasurement h eese 86 4 2 Header file 86 4 3 Implementation file 88 4 4 Implementation file 89 APPENDIX FIXTURE CHARACTERIZATION 1 41 1 8 24 0 2 4 44 107 IIS YS TIC APPROACH
6. E 58 58 59 UR 59 15 E 60 EE 60 USE 61 61 61 61 ee 61 ee 61 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 138 141 AATE EE EEEE T O EE ATTE 62 UD ecu A RE 62 AEE EAA A 63 OD 63 e o 63 ur 64 65 1 3 VENN 65 65 X 67 67 109 jv O AR 110 00 M 110 110 111 111 111 LT 111 P 111 111 112 112 M I UM 112 op PEENE A
7. tet tete 69 Iug Da Waver map een den deii 70 structure of Cra NELICEGEUS teda odit 71 Fig D 5 Typical I V curve for 04 eee eene nene 72 Fis 2 0 Map of the waver GaN AFIT 04 73 Fig 2 7 Layer structure of GANT EIS cie t e eg eite iei 74 Fig D 8 Typical I V curve for GaNHFETOSG esses eene eene nnne 75 pug opo Map of the waver GaNIHEBTUS ctt 76 DIO0 bayer of GANHFETI O 77 Fig D 11 Typical I V curve for GANHPET 10 nee 78 Fig D 12 Noise parameters and gains of GaNHFETIO 7 79 Fie waver GaNHEBEBT TIU teipsum ons 0 Fig E 1 Comparison between the original blue and the converted data red 82 Fig E 2 Comparison between combined sections of measurements taken at linear left blue and logarithmic right red equidistant 51 82 Fig E 3 The user interface of the automated low frequency noise measurement jo 42151427 E M 83 Fig 4 Expressive icons are used to reflect the status of the program Their meaning from left to right GPIB card search GPIB card found GBIP card failure SR770 search SR770 found SR770 failure SR770 warning SR770 save data SR770 measureme
8. void CFlickerNoiseMeasurementDlg OnDestroy CDialog OnDestroy SendDataSR770 AOFM1 this enable auto offset SendDataSR770 ALRM1 this enable sound KillTimer ID TIMER1 ThreadExitFlag TRUE Exit measurement void CFlickerNois asurementDlg SetDlgControlState bool DisableFlag GetDlgItem IDC_EDIT VG gt EnableWindow DisableFlag GetDlgItem IDC_EDIT VD gt EnableWindow DisableFlag GetDlgItem IDC_EDIT_ VR gt EnableWindow DisableFlag GetDlgItem IDC EDIT 1 gt EnableWindow DisableFlag GetDlgItem IDC EDIT R EnableWindow DisableFlag GetDlgItem IDC EDIT P2 EnableWindow DisableFlag GetDlgItem IDC EDIT FSTART EnableWindow DisableFlag GetDlgItem IDC EDIT FSTOP EnableWindow DisableFlag GetDlgItem IDC RADIO AVG AUTO gt EnableWindow DisableFlag GetDlgItem IDC RADIO AVG FILE gt EnableWindow DisableFlag if m Radio AveragingMode GetDlgItem IDC BUTTON EDIT AVG gt EnableWindow DisableFlag GetDlgItem IDC CHECK LOG AVG gt EnableWindow DisableFlag GetDlgItem IDC COMBO LOG SAMPLE gt EnableWindow DisableFlag GetDlgItem IDC BUTTON EDIT INIT gt EnableWindow DisableFlag GetDlgItem IDC BUTTON EDIT SETUP gt EnableWindow DisableFl
9. AFX DATA INIT CAboutD1g AFX DATA INIT void CAboutDlg DoDataExchange CData CDialog DoDataExchange pDX 4 AFX DATA CAboutDlg AFX DATA MAP BEGIN MESSAGE CAboutDlg CDialog AFX MSG MAP CAboutD1g No message handlers MSG MAP END MESSAGE MAP Exchange pDX CFlickerNoiseMeasurementDlg dialog CFlickerNoiseMeasurementDlg CFlickerNoiseMeasurementDlg CWnd pParent NULL TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 90 141 CDialog CFlickerNoiseMeasurementDlg IDD pParent AFX_DATA_INIT CFlickerNoiseMeasurementD1g m Statu m Vr m Vg m Vd 0 0 2 1 FStop 0 0 FStart 0 0 CheckAverage FALSE m Radio AveragingMode 1 m CheckLogAverage FALSE m TimeToGo T AFX DATA INIT Note that LoadIcon does not require a subsequent DestroyIcon in Win32 m hlcon AfxGetApp gt LoadIcon IDR_MATNFRAME lt gt M E 35 88 8 void CFlickerNoiseMeasurementDlg DoDataExchange CDataExchange pDX CDialo
10. Qv S 5 aO 2 A Y y Y y 4 GU 4 y Le m 4 10 LO LO gJ 20 LOL 0 C 20 B F F E F T y F E F F F UJ J 5 pem Bar e que E j J i Lam Ltn rm T i E 10 g LO J LC 0 LUJ JJ 0 LOL OJ D Lam T E F Y y X S ea y y y X E ES Jg 10 C 0 LU J J for LO 10 LO F y me dE Fo D Jg 20 20 cop LO 20 LO 0 Firm F sh m dr F E F UJ CJ LU 0 LU U LQ LUI OU G E E E X X E de g a y GU y a g Yg N gt of a8 Q 2 lt lt gt HE ob og OH on o J ob ob ob bob oh ob ob ob ob ob ob ob ob od ob ob ob oh pb oh ob od bb ob pb ob o ore of
11. a 8 Termination im Power Calibration Sensor Power Meter 2 gt Attenuator 10dB Sensor Attenuator 2048 SQ Microwave Switch Microwave Switch ij 9 Noise a 7 d B Noise mo ult Noise Em ii Bias Source Tuner pom H Load Tuner Bias 11 lt mx Extender a Bo z z E H Adapter RF Probes JejeuieJeg S J9jouigJeq sS Bias Control Parameter Analyzer r Adapter 1 Network Analyzer a Fig 6 1 The new microwave setup to characterize DC characteristics S parameters power noise of transistors in the microwave frequency range from 2 GHz to 18 GHz Microwave switches at the input and output allow to switch between different setup configurations without having to calibrate the system again TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 122 141 Fig 6 1 shows the changed setup with the microwave switches at the input and output of the system If the entire setup is initially characterized it is now possible to switch between S parameter power and noise measurement without having to disassemble or calibrate the system again DC measurements can be performed at any time independent of the selected measurement path Measurements can now be carried out much quicker It was desirable to control the switches by software and to fully integrate it into the existing software
12. DisplayStatus Initializing the instrument STATUS SR770 sprintf Command FNM InstrumentInit txt Handle fopen Command rb if Handle NULL no fil xiste Create file with standard init commands Handle fopen Command w if Handle NULL SetDlgControlState ENABLED DisplayStatus Error creating file FNM SystemSetup txt Write protection Disc space STATUS ERROR return FALSE CreateInitInstrumentStandardFile Handle fclose Handle Handle fopen Command rb while feof Handle TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 95 141 fgets Command 100 Handle get Command if ValidCommand Command if SendDataSR770 Command this SetDlgControlState ENABLED DisplayStatus ERROR Initializing the SR770 STATUS ERROR return FALSE if strncmp Command ARNG1 6 pause 5000 after Autoranging DisplayStatus Instrument initialized STATUS SR770 pause 1000 return TRUE bool CFlickerNoiseMeasurementDlg SaveMeasurementDataToFile int Answer char MessageString 200 FILE Handlel Handle2 int CurrentRangeINDEX CurrentValueINDEX CurrentFrequencyRange INDEX CurrentSeriesValueINDEX FirstRefValue MeanCounter double LastFrequency RefFrequencyStep log ThisFrequency log NextFrequency log
13. ifdef DEBUG define new DEBUG NEW undef THIS FILE static char THIS FILE endif 1 Hj E LH CFlickerNoiseMeasurementApp BEGIN MESSAGE CFlickerNoiseMeasurementApp CWinApp AFX MSG MAP CFlickerNoiseMeasurementApp NOTE the ClassWizard will add and remove mapping macros here DO NOT EDIT what you see in these blocks of generated code AFX MSG ON COMMAND ID HELP CWinApp OnHelp END MESSAGE CFlickerNoiseMeasurementApp construction CFlickerNoiseMeasurementApp CFlickerNoiseMeasurementApp TODO add construction code here Place all significant initialization in InitInstance The one and only CFlickerNoiseMeasurementApp object CFlickerNoiseMeasurementApp theApp CFlickerNoiseMeasurementApp initialization BOOL CFlickerNoiseMeasurementApp InitInstance AfxEnableControlContainer Standard initialization If you are not using these features and wish to reduce the siz of your final executable you should remove
14. 0 01 20 Ghz GPIB Oscillator m Hewlett Packard oo HP85052D E 0 001 2 047 Ghz GPIB cid Noise Meter gt 22 Maury Mi MS 59 MT2075C 25 3 0 01 18 Ghz 0 45 12 5 Ghz 0 8 18 Ghz 0 8 18 Ghz 0 45 12 5 Ghz 1 8 18 Ghz Maury Microwave R MT7618E ih a GNE RF Probes _ rl Noise cw Bias Source Tuner 1 Load Tuner Bias Extender 5 698 Ys r aury Microwave Ss SS Microwave MI Microwave S eS MT982A01 a MT982A01 Ege MT868C OH GGB Industries OB HS 40A GSG 150P GS H Tuner Control Maury Microwave MT986B02 GPIB Bias Control m J pi Parameter Analyzer Hewlett Packard Fig 4 8 The microwave setup configured for noise measurements At the input of the system a noise source MT7618E Fig 4 9a from Maury Microwave 18 connected which can be switched between two noise temperatures by a noise figure meter MT2075C Fig 4 9b of Maury Microwave The direct input tuning range of the noise figure meter is limited to frequencies below 2 047 GHz The microwave setup however is designed to cover a frequency range up to 18 GHz which is above the maximum direct input frequency of the noise figure meter An external down convers
15. dd lt 4 N tatus un 8 un Nh W m UJ gt ontrol Yes E tatus ontrol ontrol Y es 2 2 2 2 2 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 125 141 Only the three least significant bits of the parallel port signal are utilized to control the position of the switch Table G 3 shows which one of the three possible signal paths is connected to the common contact depending on the data byte and the enable port G state Enable port G Data byte Connected signal path 0 pxxx Common with port 1 O 010 with port 2 ______ x100 Common with port 3 0 x000 No port connected to common 12 Kxxx port connected to common Table G 3 Control settings and the relating activated signal path of the microwave switch To protect the bus transceiver from oscillations in the supply voltage a 100 nF capacitor 1s applied The output ports of the transceiver are three state ports 1 e the port can be high 5 V low GND or can be in a high ohmic state To guarantee a proper voltage level in the latter case according to the TTL definition pull down resistors are connected to the outputs of the transceiver The control signals and the 12 V coil supply voltage are interfaced by a female 9 pin sub D connect
16. latte gr 50 2 0 T ARCS UTS GONE TA UI 50 FR 53 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 2 141 APPENDIX A SYMBOLS AND 5 54 APPENDIX B DERIVATIONS FOR THE FLICKER NOISE eee 57 DERIVATION OF THE EQUATION RELATING S WITH 5 0 4 000000 000 57 ON 666 57 59 B 2 ESTIMATION OF FOR THE PRESENT MEASUREMENT SYSTEM 60 ESTIMATION OF THE DIFFERENTIAL RESISTANCE 62 APPENDIX C CORRELATION 63 C ITHECORRELNTIONACOEEEIGIBENT 56 63 LEAST SOUARE REGRESSIONS 5 Lie da 64 65 C THE COEFFICIENT OF DETERMINATION 66 C 5 ERROR DUE TO SMALL NUMBER OF 65 2 67 APPENDIX D VAVERINFORMATION e eee eee eee eene eene eee eese esos sess esses s 68 BO 68 DTT picai material DEODEFIS a ttn 68 DD Ts ye U 68 DLT pica IC ra SCS acd eod te EE eut 69 LIS SV GV CT quads yu teat ticis eae Sade aaah es 70 D2 GANAF E TO eite 71 DOT proper
17. 15 Vg MEN Vs 1 00 V 1 5 1 50 V s 0 Dv 0 00 M MM 2 00 0 0 00 2 00 4 00 6 00 8 00 10 00 12 00 Drain Source Voltage Vp V Created by macro HP OutputCharacterizticz written by Tobias Werner 2000 Fig D 11 Typical I V curve for GaNHFETIO TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 79 141 0 4 4 Noise and gain parameters The transistor F07 on waver GaNHFET109 had the lowest measured noise level and the noise properties of this transistor will be therefore mentioned Fig D 12 shows the noise parameters and different gains as a function of the selected bias condition The minimum noise figure the relative noise resistance the associated gain at minimum noise figure and the maximum available gain Gay are specified for ax each working point of the analyzed transistor The optimum source reflection coefficient 1s also indicated as a red dot in a small smith chart Noise parameters and gains at 2 GHz for different bias conditions HFET10_F07 25 20 VG 0 00 V T 15 gt F min 1 64 dB 1 21 dB 1 00 dB z 0 50 V 8 ANF min 3 04 dB Gam 11 85 dB dB dB Fmin 0 81 dB Eas Tz 34008 2 85 dB G 7 03 dB 9 97 dB eo 2 0 3 05 dB 0 12 52 dB n 9 46 3 40 Fmin 1 80 dB F min 0 99 dB 0 87 dB
18. 21 4 8 The microwave setup configured for noise measurements 22 4 9 a Noise source MT7618E and b Noise Gain Analyzer MT2075C 22 4 10 Example of source and load positions for a noise measurement The light blue crosses in the smith chart indicate the range which can be covered by the mechanical tuners at 2 GHz The red crosses in the source chart are the positions where the noise figure is measured and will be used to determine the noise parameters The red cross in the load chart point to the position of the load impedance ees 23 4 11 Typical noise measurement screen in the ATS of Maury Microwave The optimum source reflection coefficient is indicated by the square in the smith chart Circles of constant noise figure for different source impedances are plotted in red color into the smith chart suus 24 4 12 The representation of the noise setup in the ATS software of Maury ta 24 4 13 Picture of the setup to measure the low frequency noise properties of Se an 29 4 14 User interface of the program Automated IV Curve Measurement 29 4 15 a The noise floor of the low frequency noise measurement system The spikes at harmonics of the power system frequency 60 Hz cannot be suppressed by the shielding enclosure The graph also shows disturbances at 15 k
19. Fig F 8 The slope of the line segment can be easily calculated in the left example but will lead to a wrong result for the electrical length in the right example F 3 Matlab Program The presented procedure to calculate the S parameters of a fixture from the error coefficients of two VNA one port calibrations was implemented in Matlab The program accepts two standard calibration files of the HP8510C and calculates the S parameters with corrected phase information for 512 and 21 The S parameters are then written to a file in a touchstone compatible format and can be used to setup the microwave systems TOBIAS WERNER F 4 Verification of the results To control that the program is working correctly the S parameters of a fixture were calculated by applying the presented algorithm and the result was compared to that of a commercial fixture characterization program Maury Microwave MT956 Fixture Software 3500 The conditions were not exactly the same because the rigid low loss cable was bended between the two measurements and therefore changed its electrical characteristics slightly Nevertheless the agreement between these two measurements is very good and Wei CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 115 141 shows that the implemented algorithm is working excellent siop REHE VM 52 22 55 Vis gts Ma
20. Surface passivation Process MOCVD low pressure D 2 2 Layer structure SOURCE GATE DRAIN WIDE gN 600 2DEG helero interface NARROW GaN 1 5 um SiC m Fig D 4 Layer structure of 04 PAGE 71 141 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 72 141 D 2 3 Typical DC characteristics A typical output characteristic and typical gate leakage current is shown in Fig D 5 for a transistor 707 on waver GaNHFETO04A and vs Vp GaNHFETO4 J07 25 40 Gate Source Voltage 35 20 Gate Source Voltage 30 Vg 2 00 Vg 0 00 V 20 Vg 1 50 V Vg 0 50 V Drain Source Current Ip mA Gate Current 15 WA 10 Vg 7 1 00 V Vg 7 1 00 V 5 c 0 50 V Vg 1 50 V 0 0 00 2 00 V 0 0 00 5 00 10 00 15 00 20 00 25 00 30 00 D rain Sou rce Vo Itage Vp V Created by macro HP OutputCharacterizticz written by Tobias Werner 2000 Fig D 5 Typical I V curve for 04 TOBIAS WERNER D 2 4 Waver map Transistors marked red were working devices The DC characteristics S parameters and the flicker and microwave noise were measured for these devices Devices marked gray were not working All other device were either working but had differing properties or were not CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NO
21. dB Ta 4 lo a 14 27 dB 15 36 dB min G 0 94 dB pe eee G 14 03 dB ra 4 89 ies Lake hs r 4 45 li a gt 1 02 dB 0 71 dB 0 59 dB F min 0 62 dB en po 10 53 dB 12 89 dB ma 14 08 dB 2525 6 7 90 dB G 10 85 dB 0 12 93 dB 15 32 98_____ f Q 1 0 dB 14 13 dB 0 2 430 4 501 Gr 6 06 8r 7 08 10 12 n G Drain Source Voltage Vp V Fig D 12 Noise parameters and gains of GaNHFETIO 07 PAGE 80 141 2 2 2 5 23 598 Bo See gt N 22 2 SFH FH FH FH FH EH HA FHH d E E HH HJ HJ E HY EJ HJ HJ FH HH 2 5 HH HH eles UU PH O OH HH HH 5 2 SIPH FH FH FH FH FH UEH FH FH ee UH H FH 2 9
22. sprintf sSeconds d Seconds CorrectTimeFormat sSeconds sprintf sMinutes d Minutes CorrectTimeFormat sMinutes m_TimeToGo Format Ss s sMinutes sSeconds SetTimer ID TIMERI 1000 NULL else m TimeToGo KillTimer ID TIMER1 UpdateData FALSE PAGE 105 141 TOBIAS WERNER void CFlickerNoiseMeasurementDlg ConvertToMinutesAndSeconds int Seconds int Minutes Minutes int ceil Seconds 60 Seconds Seconds Minutes 60 void CFlickerNoiseMeasurementDlg CorrectTimeFormat char Element char DummyString 2 if strlen Element 1 sprintf DummyString 0 s Element else sprintf DummyString s Element strcpy Element DummyString LONG CFlickerNoiseMeasurementDlg OnStartCountdown WPARAM Seconds LPARAM 1Param CountDownSeconds int Seconds return 0 int CFlickerNoiseMeasurementDlg CalculateMeasurementTime CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS int EntireTime OffsetTime SettlingTime MeasurementTime TransferTime OverheadTime AveragingRange INDEX double Start Stop FrequencySpan OffsetTime 16 s TransferTime 10 s OverheadTime 7 s EntireTime OverheadTime OffsetTime Start m FStart Stop m FStop for AveragingRangeINDEX 0 AveragingRangeINDEX m Numbe
23. 2 0 2 Device dB dB dB dB 802 1 42 1 22 1 20 1 20 co1 1 08 1 05 0 76 0 77 006 2 86 2 47 2 18 2 48 DO 1 51 1 43 1 55 1 46 E10 2 20 222 2 46 2 45 12 2 84 2 40 2 13 1 84 E16 2 79 247 2 35 2 29 FO 0 73 0 54 0 62 0 58 F10 2 29 2 46 3 41 3 34 F16 1 56 1 30 1 65 1 65 F17 1 69 1 71 1 49 501 1 25 1 28 T27 1 04 602 1 49 1238 1 58 1 54 505 2 33 2 27 2 95 2 85 G10 2 50 2 44 3 12 2 99 G11 1 29 103 0 94 0 86 G13 1 02 0 96 0 79 0 80 G14 1 42 1 38 1 46 1 44 G17 1 94 1 76 1 55 1 48 G20 2 41 2 20 1 84 1 71 H01 1 30 1 14 1 18 1 10 H11 1 06 0 84 0 79 H12 1 87 1 53 1 83 1 60 05 1 88 1 85 2 49 2 40 2 03 1 89 2 33 2 28 1 70 1 50 1 20 1 10 min 132 94 175 38 126 14 7 29E 04 0 54 0 62 0 58 average 125 97 167 51 118 41 4 45E 03 1 79 1 65 1 74 1 67 correlation matrix Table 5 1 Correlation results for bias conditions of lowest noise 0 Table 5 1 shows the individual low frequency and microwave noise results for each device on the waver GaNHFETIO The header row gives information about the measurement conditions parameters and units The lower main row shows the minimum maximum and average values of the corresponding columns The correlation coefficient was calculated for each column of the low frequency noise and the microwave noise and is shown in a matrix at the bottom of the table For the waver GaNHFETIO the correlation coefficient demonstrates only a negligible dependence on the m
24. EARG Invalid argument gt if iberr ESAC strcat str ESAC Not Sys Ctrlr gt if iberr EABO strcat str EABO Op aborted gt if iberr ENEB strcat str ENEB No GPIB board gt if iberr EOIP strcat str EOIP Async I O in prg if iberr ECAP strcat str ECAP No capability gt if iberr EFSO strcat str EFSO File sys error gt if iberr EBUS strcat str EBUS Command error gt if iberr ESTB strcat str ESTB Status byte lost gt if iberr ESRQ strcat str SRQ stuck on if iberr ETAB strcat str ETAB Table Overflow gt Add ibcntl information to the Message String strcat str nibcntl ultoa ibcntl tempbuf 16 streat str tempbuf AfxMessageBox str MB ICONSTOP 0 Call the ibonl function to disable the hardware and software 1 1 0 0 bool CFlickerNoiseMeasurementDlg FindGBIPCard DisplayStatus Searching GPIB card STATUS CARD FIND pause 1000 SendIFC 0 at board address 0 if ibsta amp ERR SetDlgControlState ENABLED sprintf buffer No GPIB card was found on this computer DisplayStatus buffer STATUS CARD ERROR return FALSE DisplayStatus GPIB card found STATUS CARD pause 1000 return TRUE bool CFlickerNoiseMeasurementDlg InitInstrument FILE Handle char Command 100
25. PollForSettlingCompletion pApp FrequencySpanIndex Calculating the measurement tim TimeForMeasurement int 400 FrequencySpan Averaging PostMessage pApp gt m_hWnd WM START COUNTDOWN WPARAM TimeForMeasurement 0 sprintf buffer Measuring from 0f Hz to 0f Hz Average Sd StartFrequency StopFrequency Averaging PostMessage pApp m hWnd WM DISPLAY STATUS WPARAM buffer LPARAM STATUS EASURE if pApp gt SendDataSR770 AVGOO pApp return FALSE Averaging OFF if pApp SendDataSR770 STRT pApp return FALSE Start measurement if m FirstAverage TRUE if SendDataSR770 ARNG1 pApp return FALSE Auto ranging pause 5000 if SendDataSR770 ARNGO pApp return FALSE Auto ranging off if pApp gt m_CheckAverage TRUE pApp gt m_FirstAverage FALSE if pApp gt SendDataSR770 STOP pApp return FALSE Stop measurement if pApp gt SendDataSR770 AVGO1 return FALSE Averaging ON sprintf buffer d Averaging Set averaging factor if pApp gt SendDataSR770 buffer return FALSE if pApp gt SendDataSR770 OVLPO pApp return FALSE Overlapping 0 if pApp gt SendDataSR770 AUTSO pApp return FALSE A
26. rpc and can be united to the resistor R Fig B 3c which is given by _ Upc t PETRER B 14 Rx and R form a voltage divider and the resistance of can be calculated by _ Vbat Ry Ty s Ry B 15 Inserting B 13 and B 14 into B 15 and solving for we obtain the equation bat R tVbat ed 221 22 2 2 2 2 2 2 VR P EV bat R TDC 16 2 2 Toc pou tuas oh ba ape This equation consists only of known or measured quantities and can thus be solved R can then be determined from B 13 R P R B 17 With this result the network in Fig B 3a can be transformed to that of Fig 1 with p T SF B 18 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 62 141 B 3 Estimation of the differential resistance r Spontaneous fluctuations in current and voltage is referred to as noise Thus it is no static but a time dependent process To estimate the resistance of the device under test a linear approximation in its working point is made Fig B 4 shows a sector of an IV Curve blue curve with the red dots representing the discrete measured points of the drain current versus drain voltage Vp Differential resistance P Vpi lp i Fig B 4 Calculation of the diff
27. text sprintf Proccessing error correction file WITHOUT FIXTURE s file namel disp text for CurrentErrorTerm 1 NumberOfErrorTermsl Buffer ReadErrorTerms CalFilel NumberOfFrequencies1l if size Buffer 1 NumberOfFrequenciesl disp ERROR The number of frequencies differs from the number of error terms disp E X IT PROGRAM return else ErrorTermArrayl CurrentErrorTerm INDEX1 Buffer end end text sprintf Proccessing error correction file WITH FIXTURI disp text s file name2 for CurrentErrorTerm INDEX2 1 NumberOfErrorTerms2 Buffer ReadErrorTerms CalFile2 NumberOfFrequencies2 if size Buffer 1 NumberOfFrequencies2 disp ERROR The number of frequencies differs from the number of error terms disp E X IT PROGRAM return else ErrorTermArray2 CurrentErrorTerm INDEX2 Buffer end end 55 5 5 5 5 5 5 5 Calculate the S parameters for the fixture S S Fixture S MA CalculateFixtureSParameter ErrorTermArrayl ErrorTermArray2 FrequencyArrayl Electrical Length 55 5 5 5 5 5 5 5 Store the S parameters for the fixture 5 if StoreFixtureFile Fixture S MA file namel file name2 return Error end text sprintf nPROGRAMM FINISHED SUCCESSFULLY disp text fclose CalFilel fclose CalFile2 return oo END OF MAIN PR
28. 156 00 350 400 450 500 5 050 10 150 200 250 300 950 1000 1050 1100 5S4100Hz dBArmsHz Sq 100Hz vost 83 1 dBArmsh Hz S4 100Hz BV Vp Hz Fs 2GHz SV vp 48 Fd 2GHz 8V 10 88 FF 28GHz dB One result Method B GaNHFETIO worst case yielded to a correlation coefficient of 0 56 0 15 but considering all reliable other cases always a small negative correlation coefficient between 0 2 0 2 Method and 0 3 0 2 Method B or close to zero Method C was obtained see table The presented conclusion is supported by the constellation of widely spread points in the scatter diagrams which showed no clear tendency concerning reliance between the flicker and the microwave noise This result indicates that two independent noise sources are responsible for the flicker and the microwave noise noise model can thus be established which describes the microwave and the flicker noise as a superposition of uncorrelated noise sources TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 54 141 Appendix A SYMBOLS AND ABBREVIATIONS Symbol QOH Unit 1 m m rad gt gt gt gt lt lt lt lt lt lt lt Description Hooge parameter phase constant of the line wave length phase of S parameter 512 standard deviation of data set X standard deviation of data set Y opti
29. Company Inc 1965 Larry D Schroeder Understanding Regression Analysis An Introductory Guide Sage Publications 1986 I 3 Measurement techniques 32 33 34 35 36 37 38 Hewlett Packard Fundamentals of RF and microwave Noise Figure Measurements Hewlett Packard Application Note 57 1 1983 Maury Microwave Automated Tuner System User Manual MT993 2 Maury Microwave 2000 Focus Microwaves Basics on Load Pull and Noise Measurement Focus Microwaves Application Note 8 1994 Hewlett Packard A new technique for measuring components using the HP 8510C Network Analyzer Product Note 8510 13 Hewlett Packard Company 1999 Hewlett Packard Specifying calibration standards for the HP 8510 network analyzer Product Note 8510 5A Hewlett Packard Company 1997 Hewlett Packard In Fixture Measurements Using Vector Network Analyzers Application Note 1287 9 Hewlett Packard Company 1999 Glenn Elmore De Embedding measurements using the HP 8510 Microwave Network Analyzer RF amp Microwave Measurement Symposium and Exhibition Hewlett Packard Company 1985
30. TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 34 141 where is the differential resistance and R the resulting bias network resistance detailed derivation of 5 2 can be found in Appendix Derivation of the Equation relating SI with SV The measured transistors on the wavers had the same gate length 0 25 um and gate width 100 um but different source drain spacings To examine if this influences significantly the result of the correlation between the flicker and the microwave noise the Hooge parameter which takes into account this drain source spacing was calculated To examine if the correlation coefficient is varying considerably dependent on the applied parameter the minimum noise figure was not only compared to Sy S and the Hooge parameter but also to the normalized spectral current noise density S a in the following only referred to as S 7 2 which 15 often referred to in papers It normalizes on the squared drain current due to the fact that the flicker noise in the linear region of a transistor is proportional to n Furthermore the comparison between S 5 1 2 and allows to separate the influence of the normalization to o and the contribution of the different source drain spacings 5 2 Microwave noise To characterize the microwave noise the minimum noise figure was estimated at 2 GHz at room temperature wi
31. 126 00 4 128 00 gt 130 00 4 is 2 32 00 lero 134 00 r S 0 50 1 00 1 50 2 00 2 50 a 1 00 1 50 Fen 2GHz V 7 1V V5210V dB V5 7 1V Vo 10V dB Correlation coefficient 0 04 0 20 Correlation coefficient 0 02 0 20 110 00 4 1 60E 02 4 51 aa S 140E 02 5 eo 114 00 1 20 02 J 116 00 7 gt 1 00 02 118 00 4 F10 gt 8 00 03 120 00 4 5 6 00E 03 4 5 520 122 00 Ye 2 xcd gt 126 00 2 00E 03 4 e cur 128 00 r 0 00E 00 3 0 50 1 00 1 50 2 00 2 50 0 50 1 00 1 50 2 00 2 50 Fminl d GHZ Vg 1V Vp 10 dB Fig 5 9 Scatter diagrams for a comparison between F min at 2 GHz representing the microwave noise and Sy 8 8 VP and ay at 100 Hz representing the flicker noise GaNHFETIO TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 39 141 Fig 5 9 shows four representative scatter diagrams for the bias condition with the lowest averaged minimum noise figure 1 65 dB On the ordinate Sy Sr S J and at at 2 GHz on the abscissa One dot in the diagram represents a certain transistor on the same waver with its minimum noise figure and corresponding low frequency noise parameter The correlation coefficient and the regression line are drawn in red color to each plot Also indicated with the correlation coefficient is the err
32. 4 4 a Typical result of a S parameter measurement obtained with the aid of a b vector network analyzer Hewlett Packard 8150 A tuner controller MT966B02 from Maury Microwave manages two mechanical tuners MT982A01 Fig 4 5 from the same company at the source and load position Tuners are used to change the impedance seen by the DUT at its input and output and are set to 50 Ohms during the S parameter measurement Fig 4 5 Picture of two mechanical tuners MT982A01 with a tuner controller MT966B02 of Maury Microwave Two very high temperature probes VHT40A GSG 150P Fig 4 6a of GGB Industries with ground signal ground configuration and 150 um contact spacing are used to contact transistors on waver gt Industries 4196 Corporate Square Naples FL 34104 USA Phone 1 941 643 4400 Fax 1 941 643 4403 www ggb com TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 21 141 a b Fig 4 6 a Probe tip VHT40A GSG 150P and b probe tips contacting a through on a calibration substrate To get accurate results for the S parameters the VNA has to be calibrated before the measurement This is required to shift the reference plane to the probe tips where the measurement is carried out Precisely known calibration standards open short load through are measured and compared to the expected response The VNA then calculates error coefficients which com
33. File Edi Only initial autorange GPIB Setup Tine 10 Addr 10 Find Comand files Initialization Setup system file Edit Setup system Data file Start sample Far log conversion 20 Average No GPIB card was found on this computer Fig 4 19 The user interface of the program to measure the low frequency noise The program is named Automated Low Frequency Noise Measurement ALFNM and the user interface can be seen in Fig 4 19 The description and manual of the program can be found in Appendix E Low Frequency Noise Measurement Program To display the measurements a macro was written in Microsoft Excel which directly opens and displays the latest measurement result TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 31 141 5 MEASUREMENT RESULTS 5 1 Flicker noise To characterize the low frequency noise the spectral voltage noise density Sy was measured at the drain of a biased transistor at room temperature S vs frequency f VG 0 V VD 0 5 V GaNHFETO4 4J06 100 4 105 Slope 1 f 5 1100 Hz 155 160 4 T 1 10 100 1000 10000 100000 Frequency f Hz Spectral voltage noise density Sv dBVrms Hz Fig 5 1 Typical result of a low frequency noise measurement with the slope proportional to I f Fig 5 1 shows a typical noise measurement result of Sy versus frequency from 1 Hz to 1
34. GaNHFETIOQ 37 Table 5 2 Correlation results for bias conditions of lowest noise GaNHFETO6 40 Table 5 3 Correlation results for bias conditions of lowest noise GaNHFETO0d4 43 Table 5 4 Correlation results for an identical bias conditions in saturation TU Occit IE 46 Table 5 5 Correlation results for an identical bias conditions in saturation Gron ol TIO vied 48 Table 5 6 Correlation results for working points covering the entire bias range of one transist r onm 51 Table G 1 RF properties of the used microwave switches 122 Table G 2 Parallel port LPT assignment and signal names eee 124 Table G 3 Control settings and the relating activated signal path of the microwave SAUL GI aiiis DA LM ID IE 125 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 137 141 21 7 7 OTR 8 D ccu Mu MM Ree Oe 9 DS uc uuu I E LL 9 ee 14 to Xn 14 a EE EAEN E 15 CE ASEA NE 27 EL RERUM REPE 33 33 IE 58 d m 58 LI M 58 vc 58 URN 58 O 58
35. pDX AFX DATA CLPTSwitchRemoteDlg DDX Radio pDX IDC RADIO OFFLINE m SetupConfig AFX DATA MAP BEGIN MESSAGE CLPTSwitchRemoteDlg CDialog AFX_MSG MAP CLPTSwitchRemoteDlg ON WM SYSCOMMAND ON WM PAINT ON WM QUERYDRAGICON ON BN CLICKED IDB EXIT OnExit ON BN CLICKED IDC RADIO OFFLINE OnChangeConfig ON BN CLICKED IDC RADIO SPARAMETER OnChangeConfig ON BN CLICKED IDC RADIO POWER OnChangeConfig ON BN CLICKED IDC RADIO NOISE OnChangeConfig TOBIAS WERNER AFX MSG MAP END MESSAGE CLPTSwitchRemoteDlg message handlers BOOL CLPTSwitchRemoteDlg OnInitDialog CDialog OnInitDialog Add About menu item to system menu IDM ABOUTBOX must be in the system command range ASSERT IDM_ABOUTBOX amp ID ABOUTBOX ASSERT IDM ABOUTBOX OxF000 CMenu pSysMenu GetSystemMenu FALSE if pSysMenu NULL CString strAboutMenu strAboutMenu LoadString 105 _ABOUTBOX if strAboutMenu IsEmpty pSysMenu gt AppendMenu MF_SEPARATOR pSysMenu gt AppendMenu MF_STRING IDM ABOUTBOX strAboutMenu Set the icon for this dialog The framework does this automatical
36. return FALSE buffer ibcnt 1 0 PAGE 97 141 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 98 141 counter tt Status atoi buffer while counter 1 Status 128 amp amp ThreadExitFlag FALSE amp amp counter 100000 return TRUE bool CFlickerNoiseMeasurementDlg PollForCommandCompletion CFlickerNoiseMeasurementDlg char serPol do ibrsp pApp gt m_InstrumentHandle amp serPol while serPol amp 2 0 amp amp ThreadExitFlag FALSE Frage ob das ueberhaupt gebraucht wird da der Networkanalyzer sowieso das Flag hochzieht nachdem er ein Stopkommando erhalten hat if ThreadExitFlag TRUE PostMessage pApp gt m_hWnd WM MEASUREMENT STOPED 0 0 ThreadExitFlag FALSE return FALSE return TRUE bool CFlickerNoiseMeasurementDlg GetMeasurementData CFlickerNoiseMeasurementDlg char dataBuffer 30 char Message 200 int CurrentValueINDEX sprintf Message Getting measurement data from SR770 PostMessage pApp m hWnd WM DISPLAY STATUS WPARAM Message LPARAM STATUS TRANS for CurrentValueINDEX 0 CurrentValueINDEX lt 400 CurrentValueINDEX sprintf dataBuffer SPEC 0 d CurrentValueINDEX if pApp gt SendDataSR770 dataBuffer pApp return FALSE if pApp gt ReceiveDataSR7
37. 0 2 4 6 8 10 12 0 0 0 5 1 0 1 5 2 0 2 5 3 0 3 5 Drain Source Voltage Vp V Drain Source Voltage Vj V Fig 3 3 Typical current voltage characteristic of a GaN HFETs in comparison to b GaAs PHEMTS The gate current is often referred to as the gate leakage current because ideally it should be zero for field effect transistors The HFET devices show a low gate leakage current in the range of uA what is especially important for low flicker noise 12 Fig 3 4 shows typical S parameters of a GaN HFET from 2 GHz to 10 GHz for a bias condition of 10 and Ve 0 5 V The S parameters S and S are typically presented in a smith chart lower half of Fig 3 4 and 8 and 5 in a polar chart upper half of Fig 3 4 Nearly all measured devices demonstrated a S smaller than unity at a reference system impedance of 500 T zx gt Gate Current lc uA TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 13 141 S parameters GaNHFET10_F07 Vds 10V Vg 0 50V 2 10step1GHz 2 GHz 22 10 GHz Fig 3 4 S parameters of a GaN HFET versus frequency at a certain bias condition Typically the S21 is smaller than unity at a reference impedance of 50 Ohms Over all analyzed samples the minimum noise figure F which describes the noise level of the transistor in the microwave region varied between
38. 0 54 to 10 dB at 2 GHz at an optimum bias condition in the saturation region of the transistor The frequency dependence of the minimum noise figure and the associated gain at n room temperature from 2 GHz to 26 GHz is shown in Fig 3 5 For increasing frequencies F nin quickly increases and decreases To compare this result to the state of the art technology the lowest noise level of commercial GaAs transistors is also plotted to the chart The comparison indicates that the noise performance of GaN HFETs still has to be improved to compete with GaAs and vs frequency f V 5 V V 8 V IGaNHFETO1 6 3 Minimum Noise Figure Finis dB Associated Gain G dB Frequency f GHz Fig 3 5 Frequency dependence of the minimum noise figure and the associated gain of GaN HFETs at room temperature Minimum Noise Figure F rin dB o N e TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 14 141 The bias dependence of two noise parameters can be seen in Fig 3 6 In the x y plane the known output characteristic like in Fig 3 3 is plotted On the z axis the minimum noise figure Fig 3 6a and the normalized equivalent noise resistance Fig 3 65 at 2 GHz are drawn for different bias conditions b 2000 Tobias Wemer 946 2000 Tobias Werner 1 64 121 4 0 81 665 10 05 6 3 8 8 Drai
39. 133 4 17 NSERTEL 134 4 i i i i gt 2 25 a 35 4 45 5 5 5 6 6 5 2 25 33 4 45 5 B Fmin 2GHz V 5 1V V522V dB Fmin 2GHz V 7 1V V5 72V dB TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 43 141 5 4 5 Results for GANHFET04 The minimum noise figure for waver GaNHFET04 was measured at 2 GHz for two different bias points in the saturation region and was compared as before to Sy Sr S J and at 100 Hz for one bias point in the linear region 0 V Vp 0 5 V 1 f noise microwave noise VG 0 5 VG 0 5 VD BV VD 8V Fan GHZ Fak GHA Device dB dB E04 3 66 3 60 E05 3 00 3 04 E06 3 01 3 05 203 3 95 4 08 206 3 88 4 02 504 3 26 222 GO3 3212 3 10 GO8 474 HOS 3 34 HOS 2 92 2 86 HO 5 07 5 07 HOS 5 62 5 63 105 2 93 2 97 06 5 58 5 62 405 3 39 3 42 406 3 69 3 68 407 2 43 3 21 139 50 186 30 131 57 1 39E 03 2 92 2 85 average 129 44 173 99 119 29 2 40E 02 3 79 3 80 Correlation matrix Table 5 3 Correlation results for bias conditions of lowest noise 04 Table 5 3 shows the individual low frequency and microwave noise results for each device on the waver GaNHFETO04A The correlation coefficient for this waver differs only slightly for the two microwave noise bias conditions Independent of the selected flicker noise parameter the correlation coefficient
40. 141 TABLE OF CONTENTS PIN TRODUGC TION er 4 M 6 NOIB EOR ESNOISB ue 6 EIEEE 7 2 2 MICROWAVE NOISE 8 2 2 1 Theory of microwave noise measurement 9 JSANALYZED TRANSISTORS 10 3 F TYPICAL DEVICE PERFORMANCE 12 TNIEASUREMENESETUP deua o er te e a 17 AT MICROWAVE 17 dd s parameter Config tei an tete beers 19 4 1 2 NOUS CON OURO OM n 22 422 LOWPBRPOUENCY NOISE SETUP 25 42 1 Measurement wth the OR IU is iacta 2 SMEASUREMENT RESULTS Vu Ro e 31 S L EDICRBER NOISE 31 MICROWAVE NOISE 34 5 3 CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE 35 5 4 METHOD A BIAS CONDITION FOR LOWEST 36 3 TOP GANAF BTL i eet etta A eb xi tea reto ta Bae 37 FOV GAN HF EL US 40 35 4 3 JOV GOIN TE LOE Rt 43 5 9 METHOD B SAME BIAS CONDITION reb epe i tetas tees 45 2 91 Results for GaNHPETIO bids condition I ein gee 46 5 5 2 Results for GaNHFETIO bias condition 2 worst case 48 S 0 METHODO ENTIRE BIAS RANGE
41. 19 202122 2324 2526 27 28 2930 3132 Microwave noise 1 f noise IV charcteristics S Parameters 4 HH H j NOT WORKING Fig 0 6 Map of the waver GaNHFETO04 PAGE 73 141 m TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS D 3 GaNHFETO0S D 3 1 Typical material properties Mobility 1400 cm2 Vs Carrier sheet density 1 1x10 cm 2 Aluminum mole fraction 15 Surface passivation Process MOCVD low pressure D 3 2 Layer structure SOURCE GATE DRAIN WIDE Alo gN 600 2DEG hetero interface NARROW GaN 0 5 um SiC A Fig D 7 Layer structure of GaNHFETO6 PAGE 74 141 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 75 141 D 3 5 Typical DC characteristics A typical output characteristic and typical gate leakage current is shown in Fig 0 8 for a transistor C12 on waver GaNHFETOS and vs GaNHFETO8 C12 60 Gate Source Voltage 50 Gate Source Voltage 0 00 Y I e Gate Current Ic A ho e Drain Source Current Ip mA 0 00 2 00 4 00 6 00 6 00 10 00 Drain Source Vo Itage Vp V Created by macro HP OutputCharacteristics written by Tobias Werner 2000 Fig D 6 Typical I V curve for GaNHFETO6 2 5 N ut ea E PAGE 76 141 CORRELATION BETWEEN THE FLICKER AND THE MICROWAV
42. 2fHz stop 2fHz raw txt m Vd m Vg m FStart m FStop generate the FilName Handle fopen FNM LastSettings txt w if Handle NULL DisplayStatus Error opening file FNM LastSettings txt for writing Continueing STATUS WARNING pause 1000 else UpdateData DIALOG TO DATA fprintf Handle s n VERSION STRING fprintf Handle s n m FileName fprintf Handle 4 6fNn m fprintf Handle 4 6fNn m Vd fprintf Handle 4 6fNn m Vr fprintf Handle 4 6fNn m P1 fprintf Handle 4 6f n m_R fprintf Handle 4 6fNn m P2 fprintf Handle 4 6fNn m FStart fprintf Handle 4 6f n m FStop fprintf Handle Sd n m_PrimaryInstrumentAddress fprintf Handle Sd n m_TimeOut fprintf Handle Sd n m_CheckAverage fprintf Handle Sd n m_ Radio AveragingMode fprintf Handle Sd n m_ComboRefStartValue fprintf Handle d m_CheckLogAverage Last no carriage return fclose Handle void CFlickerNoiseMeasurementDlg CalculateCurrentID double dCurrentId if m_R 0 dCurrentId m Vr m Vd m m Id Format ID 4 2f mA dCurrentId else DisplayStatus Resistor 0 is not allowed STATUS WARNING void CFlickerNoiseMeasurementDlg OnKillfocusEditR UpdateData TRUE CalculateCurrentID DisplayStatus void CFlickerNoiseMe
43. 9 9 S gt E o PH FH HH FH H EH FH FH TEH FH FAH 5 IQQ HH HWH PNA HH H H H lt HJ og z HH BA FH gt BH OEH FH FAH E JHA HJ HH JUHA FHH FH FH FH HA FH ng 2 5 FH HH HH HH HU HA 2 HH FH HH HJ FH HH EJ c HH HA HH HH 2 2 2 HH H H H H H H FH FH FH HH HJ ol HA FH HH EY UEH EN HH HH gt 5 5 Ae HA FAH HH EH FAH A H FH 5 50g TPA HHH A 15 E eum LiL T HH H HH gt FH FH FH 2 x E 589 HI HIPH PH 2 2 5 oo HH H HH IHH UN EE Es gt c lt O LL I T 222 86 5 lt EO ag E S84 4 zy E 32 analyzed D 4 5 Waver map Fig D 13 Map of the waver GaNHFETIO TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 81 141 Appendix E LOW FREQUENCY NOISE MEASUREMENT PROGRAM Before the program was created the low frequency noise measurement was very time consuming and complicated Fortunately the SR770 is equipped with a GPIB interface and can be therefore controlled by a host computer equipped with a GPIB card The program provides a Windows compatible user
44. Array frequency range 1 frequency range 2 frequency range 1 frequency rang 3 1 frequency range 2 y Frequency Array return else y 0 return end return FUNCTION ReadErrorTerms FileHandler NumberOfFrequencies INPUT FileHeandler OUTPUT Vektor Frequency list read from indicated file AUTOR Tobias Werner Universitaet Stuttgart DATE February 29 2000 function y ReadErrorTerms CalFile NumberOfFrequencies exit counter 1 while isempty findstr fgetl CalFile BEGIN exit counter exit counter 1 if exit counter 20 disp ERROR in file format disp EX IT PROGRA Read the error terms for every frequency text sprintf Reading error terms for d frequencies n NumberOfFrequencies disp text ErrorTerm counter 0 Current ErrorTerm 1 CalFile while isempty findstr Current ErrorTerm END ErrorTerm counter ErrorTerm counter 1 zahl str2num Current ErrorTerm zahll zahl 1 i zahl 2 ErrorTerm Array ErrorTerm_counter 1 11 Current ErrorTerm fgetl CalFile if ErrorTerm counter NumberOfFrequencies disp ERROR in file format disp EX IT PROGRA 0 return end end y ErrorTerm Array return TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 119 141 F
45. COMBO LOG SAMPLE OnSelchangeComboLogSample IDC COMBO LOG SAMPLE OnSetfocusComboLogSample DC EDIT VR OnKillfocusEditR DC EDIT VD OnKillfocusEditR IDC EDIT FSTOP OnKillfocusEdit IDC EDIT P1 OnKillfocusEdit US IDC EDIT P2 OnKillfocusEdit EN LFOCUS IDC EDIT VG OnKillfocusEdit CBN KILLFOCUS IDC COMBO GPIB ADDR OnKillfocusEdit CBN KILLFOCUS IDC COMBO GPIB TOUT OnKillfocusEdit CBN KILLFOCUS IDC COMBO LOG SAMPLE OnKillfocusEdit WM TIMER AFX MSG MAP END MESSAGE w Z ao UJ UJ 2 a 1 2 2 w A Q n a 2 w A Q n z z AE H E Q Q U 2 00 DANHHNAA UNDANHDHHHHHH z A A U 2 z A A z A A CFlickerNoiseMeasurementDlg message handlers BOOL CFlickerNoiseMeasurementD1lg OnInitDialog CDialog OnInitDialog Add About menu item to system menu IDM ABOUTBOX must be in the system command range ASSERT ABOUTBOX amp OxFFFO ID ABOUTBOX ASSERT IDM ABOUTBOX lt OxF000 CMenu pSysMenu GetS
46. Error opening file s for writing nPlease make sure that file is not opened by another application m FileNameRaw Answer MessageBox MessageString m MessageBoxCaption MB RETRYCANCEL MB ICONWARNING if Answer 2 Cancel selected return FALSE while Handle2 NULL m ComboLogSample GetLBText m ComboRefStartValue buffer FirstRefValue atoi buffer 1 LastFrequency 1 0Hz can also be for the first value Output all the settings and frequency ranges to the file fprintf Handlel VG t 6 6f tVD t 6 6f tVR t 6 6f n m Vg m Vd m Vr fprintf Handlel P1 Nt 6 6fNtR t 6 6f tP2 t 6 6f n m m R m P2 fprintf Handlel ConvStart tsample dNtAVG t d n FirstRefValue 1 m CheckLogAverage fprintf Handlel FStart Nt 6 6fNtFStop t 6 6f n m FStart m FStop fprintf Handlel RANGE tFREQ Hz tAveraging tRang t d n m NumberOfAveragingRanges for CurrentRangeINDEX 0 CurrentRangeINDEX m NumberOfAveragingRanges CurrentRangeINDEX fprintf Handlel RangeSd t 6 6f t d n CurrentRangeINDEX 1 m AveragingRanges CurrentRangeINDEX Frequency m AveragingRanges CurrentRangeINDEX Averaging Output the measurement values for CurrentFrequencyRangeINDEX 0 CurrentFrequencyRangeINDEX m NumberOfPassedFrequencyRanges CurrentFrequencyRangeINDEX MeanCounter 0 MeanAccuY 0 for Cur
47. F Fixture Characterization which calculates the S parameters of probe tips connected to a fixture from two sets of HP8510C error coefficients The latest version of the system uses microwave switches to select different setup configurations and is described in Appendix G Microwave Switch Control With the system power S parameters and noise in a frequency range from 2 GHz to 12 4 GHz and DC characteristics can be measured For the topic of this work all measurement types except power were relevant and the different setups and configurations of the instruments will be presented The usual measurement sequence to examine the microwave noise of a transistor is the following DC measurement Fig 4 2a at the beginning to get a general idea about the maximum bias ranges of the device under test DUT S parameter measurements Fig 4 4a at the frequencies of interest for different bias conditions within the limits for the gate and drain voltage determined by the DC measurement Noise measurements Fig 4 11 for the same conditions concerning frequency and bias range as used for the S parameter measurements The DC characteristics can be measured with the program Automated IV Curve Measurement Fig 4 14 which controls the HP4145A parameter analyzer via the GPIB at any time independent of the setup configuration a Ip Vc and vs Vp GaNHFET08_C12 60 12 Gate Source Voltage 5 00 V 0 4 i T 10 Ga
48. FStop log FStart 109 m NumberOfAveragingRanges int ceil FSpan 109 m AveragingRanges CAveragingRangeType calloc unsigned m NumberOfAveragingRanges sizeof CAveragingRangeType if m_NumberOfAveragingRanges 0 m NumberOfAveragingRanges 1 FStep log FSpan log m NumberOfAveragingRanges FStart log 10910 m FStart FStop log 10910 m FStop FStart m FStart for CurrentFrequencyINDEX 0 CurrentFrequencyINDEX m NumberOfAveragingRanges 1 CurrentFrequencyIND EXT PAGE 101 141 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 102 141 FStop pow 10 FStart_log CurrentFrequencyINDEX 1 FStep_log m AveragingRanges CurrentFrequencyINDEX Frequency FStop m AveragingRanges CurrentFrequencyINDEX Averaging CalculateAveraging FStart FStop CurrentFrequencyINDEX 41 FStep log FStart FStop m AveragingRanges Current m AveragingRanges Current FrequencyIND FrequencyIND EX Frequency EX Averaging m FStop CalculateAveraging FStart m FStop CurrentFrequencyINDEX 1 FStep 109 return TRU E void CFlickerNoiseMeasurementDlg SetStandardVariableValues m Vg 1 V m Vd 3 V m Vr 4 V m Pl 10 kOhm m 10 kOhm m P2 10 kOhm m FStart 100 Hz m FStop 3000 Hz Standard GPIB address 10 m
49. Field Effect Transistor LF Low Frequency MESFET Metal Epitaxial Semiconductor Field Effect Transistor MHEMT Metamorphic High Electron Mobility Transistors MISFET Metal Insulator Semiconductor Field Effect Transistor MMIC Monolithic Microwave Integrated Circuit MODFET Modulation Doped Semiconductor Field Effect Transistor MOSFET Metal Oxide Semiconductor Field Effect Transistor MW Microwave PHEMT Pseudomorphic High Electron Mobility Transistor PHFET Pseudomorphic Heterojunction Field Effect Transistor PSD Power Spectral Density SDHFET Selectively Doped Heterostructure Field Effect Transistor SiC Silicon Carbide TEGFET Two Dimensional Electron Gas Field Effect Transistor VNA Vector Network Analyzer TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 57 141 Appendix DERIVATIONS FOR THE FLICKER NOISE B 1 Derivation of the Equation relating S with Sy The spectral voltage noise density Sy measured at the drain of a biased transistor depends on the following drain bias network which is changing if the voltage 1s adjusted Therefore the spectral current noise density S is calculated from Sy to have a common base to compare measurement results Further in papers this parameter 15 mainly used to characterize the flicker noise level and the results of this investigation can be easier compared to results referenced in publications 1 1 Linear region If a field effect transistor is ope
50. Fig C 2 The least squares line minimizes the squared distances between the line and the points The line is given 29 by 5 with the slope m calculated by COV y y P C 6 and the intercept with the y axis 7 Observation range If a dependence of two measurement sets should be analyzed it is also important to know the possible range of the parameters To illustrate this idea two graphs are shown in Fig The left picture shows an example where measurement results cover the entire range of possible values In this case a correlation coefficient close to 1 1s obtained If the same graph is taken but only a sector 1s considered Fig C 3b the correlation coefficient drops to a much smaller value TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 66 141 a b Correlation coefficient 0 99 Correlation coefficient 0 11 8 50 1 80 ML 3 00 i 1 1 70 gt 2 50 gt 5 p a a gt gt 1 50 1 60 T L S S 1 00 E ns 150 0 50 1 00 1 50 2 00 2 50 3 00 3 50 1 50 1 60 1 70 1 80 2 2 1 5 10 dB 2 2 1 5 10 dB Fig C 3 Different correlation results dependent on the chosen range a Entire range b small sector In practice the parameter range is sometimes unknown or the measurement results occur only in a limit
51. Frequency Noise Measurement takes this into account and provides an option to reduces the overhead of information in the higher frequency band If selected only relevant samples in a logarithmic reference frequency interval are taken red graph Fig 1 General Purpose Instrument Bus TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 82 141 log f Fig E 1 Comparison between the original blue and the converted data red The low frequency noise was measured from 1 Hz to 100 kHz As described before the entire graph consists of several single measurements in order to accurately describe as well the lower part of the spectrum log f log f Fig E 2 Comparison between combined sections of measurements taken at linear left blue and logarithmic right red equidistant steps If measurement sections have to be combined the graph will show steps due to a continual increase and a sudden decrease of information when the next measurement section begins This case is illustrated the left graph of Fig E 2 for three frequency sections The right picture shows the same graph if the presented algorithm which eliminates redundant measurement points is used The graph now looks smooth like a single measurement TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 83 141 E 2 Design of the program The program is written i
52. High Electron Mobility Transistor PHEMT The drain source current and the gate current are plotted versus the drain source voltage Vp for different gate source voltages The GaN HFET can stand much higher drain source voltages due to the elevated breakdown voltage of GaN devices 15 17 what can be noticed by the increased drain voltage range in Fig 3 3a While the slope in saturation of the output characteristic Ip vs in saturation is positive for GaAs devices Fig 3 35 it shows typically a negative tendency for GaN HFETs Fig 3 35 b and vs GaNHFETO08 C12 and vs Vp UMS 2612 T2 0 Ye E Vg 1 00 V 12 45 4 Gate Source Voltage Gate Source Voltage 40 Gate Source Voltage Tu M Sd 1 T 10 0 80 V i Gate Source Voltage B 0 10 V RV 4 00 V Ve 0 00 V 1 0 70 V Vs 0 60 V Ve 7 020V mA 9 Vs 0 50 V Gate Current lc uA Drain Source Current b i Ve 030V LV 0 40 V KV 2 00 V 77 1 M us 15 Vs 0 30 V 0 40 V Ve 2 00 V Il IN UE 10 0 20 V 0 50V 4 1 00 V Ve 0 10 V Ve 0 60 V Ve 3 00 V kK Ve 0 70V 0 VG 0 00 4 00 V 0 0 V 0 00 Vg 0 80
53. LI P FH H4 c HA FH gt vim Microwave noise IV charcteristics S Parameters NOT WORKING 1 f noise HH HH Fig D 9 of the waver 08 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS D 4 GaNHFETIO The microwave noise on this waver was the lowest of all measured wavers D 4 1 Typical material properties Mobility 1400 cm2 Vs Carrier sheet density 1 1 1013 cm 2 Aluminum mole fraction 15 20 Surface passivation Process MOCVD low pressure D 4 2 Layer structure SOURCE GATE DRAIN WIDE Alp Gag gN 600 A 2DEG doped GaN 6A hetero interface NARROW GaN 0 5 um ins SIC gt m m Fig D 10 Layer structure GaNHFETIO PAGE 77 141 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 78 141 0 4 3 Typical DC characteristics A typical output characteristic and typical gate leakage current is shown in Fig D for a transistor F07 on waver GaNHFET 10 Ip Vc and vs GANHFET10_F07 25 r 45 Gate Source Voltage Vg 2 00 V 4 Gate Source Voltage 20 Vo 0 00 VV 35 Vg 1 50 V lt 3 2 15 5 E 25 Vg 1 00 V Vg 0 50 V 2 2 10 S
54. Line 100 Handle m Vd atof Line fgets Line 100 Handle m Vr atof Line fgets Line 100 Handle fgets Line 100 Handle fgets Line 100 Handle m P2 atof Line fgets Line 100 Handle m FStart atof Line Line 100 Handle m FStop atof Line CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 91 141 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 92 141 fgets Line 100 Handle m PrimaryInstrumentAddress atoi Line fgets Line 100 Handle m TimeOut atoi Line fgets Line 100 Handle m CheckAverage atoi Line fgets Line 100 Handle m Radio AveragingMode atoi Line fgets Line 100 Handle m ComboRefStartValue atoi Line fgets Line 100 Handle m CheckLogAverage atoi Line fclose Handle else MessageBox Old file version of settings file FNM LastSettings txt nStandard settings are used m MessageBoxCaption MB_OK MB_ICONWARNING SetStandardVariableValues void CFlickerNoiseMeasurementDlg SaveLastVariableSettingsToFile FILE Handle Generate data file name sprintf m FileName Vd 2 4fV Vg 2 4fV start 2fHz stop 2fHz txt m Vd m Vg m FStart m FStop generate the FilName Generate raw data file name sprintf m FileNameRaw Vd 2 4fV Vg 2 4fV start
55. MICROWAVE NOISE OF GAN HFETS PAGE 4 141 1 INTRODUCTION Electronic devices and systems exhibit random fluctuations in the voltage or current at their terminals and these fluctuations are usually referred to as noise Noise is present in every electronic system and is unwanted because it limits its performance a b Fig 1 1 Noise distorting a an analog signal b regeneratable digital signal and c a non regeneratable digital signal In analog systems noise distorts a processed signal Fig a and cannot be separated from it because of the non deterministic nature of noise As a result the signal to noise ratio of an amplifier chain decreases since not only the signal but also the noise is amplified and every stage adds its own noise contribution In digital systems the signal Fig b can be regenerated but the importance of noise becomes more and more significant as the trend to lower supply voltages continues Fig c Thus the reduction of noise is important for both analog and digital systems Electronic noise may be conveniently divided into two categories determined by the noise source In the first category the noise source 1s external to the device under consideration In the second category the noise is generated within the circuit element and is called intrinsic noise This work investigates the intrinsic noise behavior of the basic component of electronics semiconductor devices Several aspects lik
56. MeanAccuY MeanFrequency LastMeanFrequency DisplayStatus Storing data STATUS SAVE pause 1000 to display the information try to open the data file for read to look if file existes Handlel fopen m FileName rb if Handlel NULL fclose Handlel sprintf MessageString The file s existes nDo you want to overwrite the existing file m FileName Answer MessageBox MessageString m MessageBoxCaption MB OKCANCEL MB ICONWARNING if Answer 2 Cancel selected return FALSE do Handlel fopen m FileName w if Handlel NULL sprintf MessageString Error opening file s for writing nPlease make sure that file is not opened by another application m FileName Answer MessageBox MessageString m MessageBoxCaption MB RETRYCANCEL MB ICONWARNING if Answer 2 Cancel selected return FALSE while Handlel NULL try to open the raw data file for read to look if file existes Handle2 fopen m FileNameRaw rb if Handle2 NULL fclose Handle2 sprintf MessageString The file s existes nDo you want to overwrite the existing file m FileNameRaw Answer MessageBox MessageString m MessageBoxCaption MB OKCANCEL MB ICONWARNING if Answer 2 Cancel selected return FALSE do Handle2 fopen m_FileNameRaw w if Handle2 NULL sprintf MessageString
57. PrimaryInstrumentAddress 10 m TimeOut 13 Standard 10s m CheckAverage FALSE m Radio AveragingMode 1 m CheckLogAverage 1 m ComboRefStartValue 4 take from file void CFlickerNoiseMeasurementDlg OnSelchangeComboLogSample m ComboRefStartValue m ComboLogSample GetCurSel void CFlickerNoiseMeasurementDlg OnSelchangeComboGpibTimeOut CString SelectedText m_TimeOut m Combo GPIB Timeout GetCurSel if InstrumentInitialized ibtmo m InstrumentHandle m TimeOut if ibsta amp ERR m Combo GPIB Timeout GetLBText m TimeOut SelectedText sprintf buffer Time out successfully changed to s SelectedText DisplayStatus buffer STATUS 5 770 else DisplayStatus 1 ERROR changing time out STATUS ERROR bool CFlickerNoiseMeasurementDlg ValidCommand char Command char Search Search2 Search Command while Search lt 32 amp amp Search 0 Search skip leading spaces Search2 Search if Search 0 return FALSE Quit Line is empty while Search2 0 amp amp Search2 37 Search2 Go to the end of the string while Search2 lt 32 Search2 37 amp amp Search2 gt Search Search2 if Search Search2 return FALSE Search2 Search2 0 if strlen Search lt 2 return FALSE GPIB commands are longer then
58. ReceiveDataSR770 Command 30 TRUE this return Get result for Averaging AveragingFlag atoi Command SendDataSR770 AVGOO this return Averaging OFF SendDataSR770 STRT this return Start measurement SendDataSR770 ARNG1 this return Auto ranging on ause 5000 F SendDataSR770 ARNGO this return Auto ranging off SendDataSR770 STOP this return Start measurement rintf Command AVGO d AveragingFlag SendDataSR770 Command this return Set Averaging back to last setting Fh Fh Fh Fh H H H H mh O DisplayStatus System is now ready for setup STATUS SR770 OK SendDataSR770 ALRM1 this SendDataSR770 MSGSSYSTEM READY this bool CFlickerNoiseMeasurementDlg CheckVariableRanges double Dummy if m_FStart lt 0 SetDlgControlState ENABLED GetDlgItem IDC EDIT FSTART gt SetFocus CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS ERROR PAGE 100 141 void CFlickerNois void CFlickerNois void CFlickerNois void Cl void Cl TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS DisplayStatus Start frequency has to be positive and unequal to unity STATUS WARNING return FALSE if m_FStop lt 0 SetDlgControlState ENABLE
59. SendByte int ReadPortStatus CLPTSwitchRemoteDlg CWnd pParent NULL standard constructor Dialog Data AFX_DATA CLPTSwitchRemoteD1g enum IDD IDD_LPTSWITCHREMOTE DIALOG D int m SetupConfig AFX DATA ClassWizard generated virtual function overrides AFX VIRTUAL CLPTSwitchRemoteD1g protected virtual void DoDataExchange CDataExchange pDX DDX DDV support AFX VIRTUAL Implementation protected HICON m hIcon Generated message map functions AFX MSG CLPTSwitchRemoteDlg virtual BOOL OnInitDialog afx msg void OnSysCommand UINT nID LPARAM lParam afx msg void OnPaint afx msg HCURSOR OnQueryDragIcon afx msg void OnExit afx msg void OnChangeConfig AFX MSG DECLARE ESSAGE MAP AFX INSERT Microsoft Visual will insert additional declarations immediately before the previous line fendif defined AFX_LPTSWITCHREMOTEDLG 0BD27EC6 5C15 11D4 B354 0006295283FA INCLUDED G 3 1 3 Implementation file LPTSwitchRemote cpp LPTSwitchRemote cpp Defines the class behaviors for the application include stdafx h include LPTSwitchRemote h include LPTSwitchRemoteDlg h ifdef _DEBUG define new DEBUG NEW undef THIS FILE static char THIS FILE FIL endif Lu
60. StatusIcon LoadBitmap IDB ICON 8770 MessageBeep MB OK break case STATUS 5 770 OK StatusIcon LoadBitmap IDB ICON SR770 OK MessageBeep MB break case STATUS FIND StatusIcon LoadBitmap IDB ICON FIND MessageBeep MB break case STATUS STOPPED StatusIcon LoadBitmap IDB ICON STOP MessageBeep MB ICONEXCLAMATION break default MessageBox Wrong argument for function DisplayStatus m StatusIcon SetBitmap HBITMAP StatusIcon CDialog OnPaint UpdateWindow RedrawWindow StatusIcon Detach m Status StatusMessage UpdateData DATA TO DIALOG void CFlickerNoiseMeasurementDlg OnSetfocusEditVg m LastMessage Pleas nter the voltage for VG in V DisplayStatus m LastMessage void CFlickerNoiseMeasurementDlg OnSetfocusEditVd m LastMessage Pleas nter the voltage for VD in V DisplayStatus m LastMessage void CFlickerNoiseMeasurementDlg OnSetfocusEditVr m LastMessage Pleas nter the voltage for VR in V DisplayStatus m LastMessage void CFlickerNoiseMeasurementDlg OnSetfocusEditPl m LastMessage Pleas nter the resistor value for the potentiometer Pl in kOhm DisplayStatus m_La
61. THE MICROWAVE NOISE OF GAN HFETS PAGE 112 141 Es EpytEgiEpi Epy EgrEpi Egi 5 pem Es Egr Egi Epi Egi EprE so Epi Egi F 11 Ex Epy Egi Ep Es EpytEgiEpi Egi Equation F 11 represents the S parameters of the unknown fixture by the error coefficients of two one port calibrations As can be seen from this equation there are two possible solutions for 512 and S21 equal in magnitude but 180 degrees apart in phase To get the correct physical result further investigations has to be made and will be discussed in the next chapter F 2 Phase ambiguity For an ideal loss less transmission line the S parameters are given by 0 PU with 4 the electrical length of the line and the phase constant From F 12 follows for the phase of 512 and 551 9 p t F 13 2 C4 with B 2 and the wavelength 4 F in F 13 the phase will be 2 7 6 00 Me F 14 From F 14 can be seen that the phase has a linear frequency dependence The slope m of the line 15 F 15 Fig F 6 shows the course of 512 5 1 if only one solution of F 11 is taken into account The phase varies between 90 and 90 degreed and not 180 and 180 as expected TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 113 141 512 821 Phase uncorrected Phase degree Frequency f GHz Fig F 6 Uncorrected phase of S12 and S21 The dots represent
62. VERA A E A A EAE 112 eR 112 113 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 139 141 Appendix I REFERENCES I 1 Noise LL LI 12 13 Aldert van der Ziel Noise in Solid State Devices and Circuits John Wiley amp Sons 1986 J Verdier O Llopis R Plana and J Graffeuil Analysis of noise up conversion in microwave field effect transistor oscillators JEEE Transactions on Microwave Theory and Techniques vol 44 no 8 p 1478 83 Aug 1996 J B Johnson Thermal agitation of electric charge in conductors Physical Review vol 28 pp 74 103 1928 H Nyquist Thermal agitation of electricity in conductors Physical Review vol 28 pp 110 113 1928 S Rumyantsev M E Levinshtein R Gaska M S Shur J W Yang and M A Khan Low frequency noise in AlGaN GaN heterojunction field effect transistors on SiC and sapphire substrates Journal of Applied Physics vol 87 no 4 p 1849 54 15 Feb 2000 P J Fish Electronic Noise and low noise design McGraw Hill Inc New York 1994 L McWhorter 1 f noise and germanium surface properties Semiconductor Surface Physics Philadelphia PA Univ of Pennsylvania Press p 207 1957 F N Hooge 1 f noise Physica vol 83B p 14 1976 V Sommer P B Albert T Zerbe A Schnell A Mesquida Kusters and K Heime Characterization of heter
63. Voltage 5 A 40 oe 4 OV a MEE 5 15 X E i peal YW 2 5 SASS Vg 7 1 Vi 25 8 LUE Me oe eee UN UN UL e US LOU 8 E 30 9 5 35 40 45 0 50 0 1 2 3 4 5 7 B 8 11 1 12 14 15 6 17 1B 8 21 22 23 Drain Source Voltage Vp V Created by macro HP OutputCharacterizticz written by Tobias Werner 2000 Fig D 2 Typical IV Curve for GaNHFETO1 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS D 1 4 Waver map PAGE 70 141 Transistors marked red were working devices The DC characteristics S parameters and the flicker and microwave noise were measured for these devices All other device were either working but had differing properties or were not analyzed 10 1 12 16 HO gt oo VY Q 25 SS e Qe gt Ne v lt SO S G gt S 2 2
64. WM START COUNTDOWN WPARAM 16 0 if pApp gt SendDataSR770 ALRMO pApp return 0 disable sound if pApp gt SendDataSR770 AOFF pApp return 0 performe auto offset calibration if pApp gt SendDataSR770 pApp return 0 disable auto offset PAGE 103 141 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 104 141 for CurrentLimitINDEX 0 CurrentLimitINDEX lt pApp gt m_NumberOfAveragingRanges amp amp ExitFlag FALSE CurrentLimitINDEX switch pApp gt ControlFrequencyRange Start Stop pApp gt m_AveragingRanges CurrentLimit INDEX Frequency case 0 start and stop are below the frequency limit if pApp gt DoMeasurement Start Stop pApp gt m_AveragingRanges CurrentLimitINDEX Averaging pApp sprintf Message Measurement failed PostMessage pApp m hWnd WM DISPLAY STATUS WPARAM Message LPARAM STATUS ERROR PostMessage pApp m hWnd WM MEASUREMENT STOPED 0 0 return 0 ExitFlag TRUE Exit the while loop pApp m NumberOfPassedFrequencyRangestt break case 1 only start is below the frequency limit if pApp gt DoMeasurement Start pApp m AveragingRanges CurrentLimit INDEX Frequency pApp gt m_AveragingRanges CurrentLimitINDEX Averaging pApp sprintf Message Measurement failed PostMessage pApp gt m_hWnd WM DISPLAY STATUS WPARAM Message LPARAM STATUS ERROR
65. from the following the specific initialization routines you do not need ifdef AFXDLL Enable3dControls Call this when using MFC in a shared DLL else Enable3dControlsStatic Call this when linking to MFC statically fendif CFlickerNoiseMeasurementDlg dlg m pMainWnd amp dlg int nResponse dlg DoModal if nResponse IDOK TODO Place code here to handle when the dialog is dismissed with OK else if nResponse IDCANCEL TODO Place code here to handle when the dialog is dismissed with Cancel TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS Since the dialog has been closed return FALSE so that we exit the application rather than start return FALSE the application s message pump PAGE 89 141 4 4 Implementation file FlickerNotseMeasurementDlg cpp FlickerNoiseMeasurementDlg cpp implementation file include stdafx h include FlickerNoiseMeasurement h include FlickerNoiseMeasurementDlg include lt string h gt include lt math h gt h include lt sys timeb h gt used in pause include lt process h gt beginthr ifdef DEBUG define new DEBUG NEW undef THIS FILE static char THIS FILE _ FILE endif define DISABLED FALSE define ENABLED TRUE 2 j O o define STATU d
66. interface where measurement parameters can be set and it is intended to automate the entire measurement process According to the settings in the interface the measurement is divided into several sub sections The parameters are transferred to the instrument and the measurement is started When the measurement for a single section has ended the measurement results are transmitted back to the computer the parameters are adjusted for the next section and the measurement process starts again When the last measurement section has passed the results are transferred to the host computer and all single measurement results are combined to one data file which 1s then stored on hard disk A macro for Microsoft Excel written in Visual Basic for Applications accesses the data file and displays the entire measurement result The following chapters describe roughly the implementation of the program and the user interface It was not the main topic of the Diploma thesis to create this program and the documentation will therefore only give a brief overview E 1 Smoothing of the graph The FFT spectrum analyzer SR770 is only capable of measuring at equidistant linear frequency steps If the acquired data is then plotted versus a logarithmic frequency scale the density of measurement points increases for higher frequencies without providing additional information about the qualitative course of the graph blue graph Fig The program Automated Low
67. lies in the same range what does not necessarily imply that the normalization to and the source drain spacing have no effect on the location of the points in the scatter diagram Sy 100Hz Vc 0V Vo 0 5 dBVrmsiyHz SI 100Hz V 20V Vo 0 5V dB Hz TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 44 141 Correlation coefficient 0 27 0 23 Correlation coefficient 0 24 0 23 120 00 4 163 00 4 07 122 00 4 m 124 00 4 5 168 00 4 o E 126 00 4 lt di 128 00 9 173 00 130 00 4 5 132 00 4 178 00 4 134 00 4 5 138 00 4 183 00 4 138 00 140 00 pu 8 18800 2 50 3 00 3 50 4 00 4 50 5 00 5 50 6 00 a 2 50 3 00 3 50 400 4 50 5 00 5 50 Fei l 2GHz V57 0 5V V5 28V dB Fmin 2GHz Vc 0 5V V5 26V dB Correlation coefficient 0 24 0 23 Correlation coefficient 0 27 0 23 108 00 4 1 80E 04 4 110 00 4 e 1 60 01 4 112 00 c 44 E 114 00 4 1 40E 01 116 00 gt 120E 01 e 118 00 4 w 100E 01 4 120 00 4 gt 47200 i 8 00E 02 4 124 00 4 2s Y ODE 02 4 126 00 4 e 4 00E 02 4 128 00 4 e 43000 9 200E 02 s 132 00 0 00 00 oo 2 50 3 00 3 50 4 00 4 50 5 00 5 50 6 00 2 50 3 00 3 50 4 00 4 50 5 00 5 50 Fu 2GHz V 7 0 5V V5 28V dB F minl 2GHz Vg 0 5V Vp 6V dB Fig 5 14 Scatter diagrams for a comparison between the minimum no
68. n fprintf Handle WNDOO 2Nt window function Hanning n fprintf Handle MRKWO 2 t smarker spot last Command no carriage return return TRUE NOCH Diese Funktion wieder loeschen UINT TestThread LPVOID pParam char Message 100 int Nummer 0 CFlickerNoiseMeasurementDlg pApp CFlickerNoiseMeasurementDlg pParam while ThreadExitFlag Nummer sprintf Message Die Nummer ist d Nummer PostMessage pApp m hWnd WM DISPLAY STATUS WPARAM Message LPARAM STATUS MEASURE pApp gt pause 1000 if Nummer 3 break PostMessage pApp m hWnd WM MEASUREMENT STOPED 0 0 return 0 void CFlickerNoiseMeasurementDlg OnMeasure CString ButtonCaption UpdateData DIALOG TO DATA m Button MeasureStop GetWindowText ButtonCaption if ButtonCaption amp Measure SetDlgControlState DISABLED if CheckVariableRanges return if m Radio AveragingMode if GetAveragingRangesFromFile return Get the frequency ranges from file else if CalculateAveragingRanges return Get the frequency ranges from file SaveLastVariableSettingsToFile Save last used settings to a parameter file CalculateCurrentID NOCH Wieder irgendwo im Dialog die Variable Id einfuegen if FindGBIPCard return if InstrumentActive m IdentificationString return if InitInstrument return
69. noise power is always k T B where B is the equivalent noise bandwidth the absolute temperature and Kk the Boltzmann constant 4 Shot noise and flicker noise which are also called excess noise as a generic name are present only when current flows through a conductor by an external means Flicker noise is influenced by surface state and bulk imperfection and cannot be described by simple physical explanations as thermal noise Since flicker noise has a 1 f dependence with y close to unity this noise 1s often called f noise The dominant noise types in GaN HFETs are the flicker noise in the low frequency range and the microwave noise for elevated frequencies in the microwave range The flicker noise is often overlaid by the so called generation recombination noise GR noise which 15 caused by fluctuations in the number of free carriers and can be related to the presence of trap centers in the forbidden gap 5 2 1 Flicker noise or 1 f noise When current passes through a resistor or a semiconductor noise is generated in excess of the thermal noise in a resistor or the thermal noise plus the shot noise in a semiconductor This excess noise has a spectral density which increases as the frequency decreases and because it 15 most noticeable at low frequency it is also known as low frequency noise In addition it is known as flicker noise because of the flickering of a the needle on a meter measuring the current with a significant co
70. pApp char Command 100 sprintf Command s command strcat Command n NOCH Ueberpruefen ob richtig 13 ibwrt pApp gt m_InstrumentHandle Command strlen Command if ibsta amp ERR sprintf buffer Send ERROR Unrecognized command gt gt s lt lt Check commands in initialization or system setup file command pApp gt DisplayStatus buffer STATUS ERROR return FALSE Quit function if pApp gt PollForCommandCompletion return FALSE return TRUE bool CFlickerNoiseMeasurementDlg ReceiveDataSR770 char data int count bool exit CFlickerNoiseMeasurementDlg ibrd pApp gt m_InstrumentHandle data count if ibsta amp ERR amp amp exit TRUE sprintf data Error reading from instrument at address d pApp m PrimaryInstrumentAddress Post gt hWnd WM DISPLAY STATUS WPARAM data LPARAM STATUS ERROR return FALSE else data ibcnt 1 0 if pApp gt PollForCommandCompletion return FALSE return TRUE int CFlickerNoiseMeasurementDlg ControlFrequencyRange double Start double Stop double Limit if Start lt return 0 if Start lt return 1 Limit amp amp Stop lt Limit ooth are in the region Limit amp amp Stop gt Limit overlapping if Start gt
71. produces 1 f noise in the drain current It is not quite clear how trapping and detrapping of electrons in the space charge region can give a l f modulation a distribution in time constants might be responsible 1 For HFET devices with carrier transport exclusively in the 2DEG it seems reasonable that the hetero interface contributes to the noise by exchanging electrons with the channel 9 Flicker noise depends on many parameters like temperature frequency bias condition device geometry or material characteristic doping level layer structure and thickness etc The noise level can be expressed by the spectral voltage noise density Sy the spectral current noise density Sy the relative spectral current noise density S J 2 by the dimensionless Hooge parameter 2 1 1 The Hooge parameter Hooge proposed in the late 1960s that f fluctuations in all homogeneous materials could be represented by the empirical formula 10 5 r amp H 5 EC 2 1 where is the total number of charge carriers in the specimen R is the mean resistance 9 40 is the power spectral density of resistance fluctuations and 2 0E 3 is a universal constant Later experiments showed that the Hooge parameter is not constant but can vary between several orders of magnitude of GaN devices lies in a range of about 1 0E 5 to 1 0E 1 5 11 The Hooge parameter can also be written as 12 51 N dm P
72. switched to the three configurations S parameter Power and Noise or can be turned offline in this case no signal path is activated TOBIAS WERNER PAGE 127 141 CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS Microwave Setup Remote Control 2000 Tobias Werner This program i used ta change the configuration of the microwave setup via the LPT port Please select the setup fram the options below Setup Configuration OFFLINE 5 C Power Noise Fig G 6 The user interface to control the microwave switches G 3 1 Program code In the following the source code with all functions can be found G 3 1 1 Header file LPTSwitchRemote h LPTSwitchRemote h main header file for the LPTSWITCHREMOTE application if defined AFX LPTSWITCHREMOTE 0BD27EC4 5C15 11D4 B354 0006295283FA INCLUDED define LPTSWITCHREMOTE 0BD27EC4 5 15 11D4 B354 0006295283FA INCLUDED if MSC VER 1000 pragma once endif MSC VER gt 1000 ifndef _ AFXWIN H _ error include stdafx h before including this file for PCH endif include resource h main symbols 71177 ATTA ATT MM CLPTSwitchRemoteApp See LPTSwitchRemote cpp for the implementation of this class class CLPTSwitchRemoteApp public CWinApp public CLPTSwitchRemoteApp Overri
73. tool of Maury Microwaves For this purpose a system consisting of hardware and software was developed and will be discussed in the following chapters G 1 Microwave switches Microwave switches from DowKey were chosen because of their superior RF characteristics and power handling capabilities Table 1 VSWR dB min dB max Watts CW 812 140 60 00 J 60 1218 150 0 1050 Table 6 1 RF properties of the used microwave switches A three way normally open switch model with the part number DowKey 535 520822A was selected because the coil voltage was specified for 12 V and it was possible to control the switch with TTL compatible logic 5 V The switch was equipped with SMA connectors and indicator contacts to reflect the currently selected path Fig G 2 left The two microwave switches with indicator LEDs connected to the control cable Rigid microwave cables at the inputs and outputs are mounted to connect to the system components right A mounted microwave switch on the load side of the system 1 DowKey Microwave Corporation 1667 Walter Street Ventura California 93003 USA PHONE 805 650 0260 FAX 805 650 1734 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 123 141 G 2 Hardware In general computers provide supply voltages of 12V and 5 V for the main board and additional devices Therefore the microwave switch control was designed to us
74. 0 15 000 m 0 209 M 50 000 m Init Single amp Save Show IG Single Append Stop Save Close Fig 4 14 User interface of the program Automated IV Curve Measurement TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 26 141 The parameter analyzer can be remoted over the GPIB by the program Automated IV Curve Measurement which is based on a driver from Maury Microwave and was extended in functionality The interface of this program can be seen in Fig 4 14 The main instrument to measure the noise level of transistors is the SR770 a FFT spectrum analyzer Fig 4 15b of Stanford Research Systems It measures the power spectral density PSD of a signal at its input at a resolution of 5 nVrms Hz 166dBVrms Hz a 100 110 120 130 measured 140 4 150 160 Spectral voltage noise density Sy dBVrms Hz 170 1 10 100 1000 10000 100000 Frequency f Hz Fig 4 15 a The noise floor of the low frequency noise measurement system The spikes at harmonics of the power system frequency 60 Hz cannot be suppressed by the shielding enclosure The graph also shows disturbances at 15 kHz and harmonics of unknown origin b The FFT spectrum analyzer SR770 of Stanford Research Systems The measurement system s noise floor versus frequency is shown in Fig 4 5a and is limited by the performance of the FFT spectrum analyzer The lowest measured
75. 00 kHz for a transistor biased at 0 V and Vp 0 5 V linear region The noise spectrum in the presented example is ideally proportional to 1 f but the slope showed a 1 f dependence with y varying from 0 9 to 1 4 over all measurements The dominant source of noise in the low frequency range the flicker noise was sometimes overlaid by the so called Generation Recombination Noise for frequencies above 100 Hz Fig 5 2 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 32 141 Bump indicates contribution of _ Generation Slope 1 f Recombination Noise Relative spectral current noise density 145 4 1 1 T 1 10 100 1000 10000 100000 Frequency f Hz Fig 5 2 Contribution of generation recombination noise To refer only to the flicker noise the reference value of the spectral voltage noise density Sy was taken at 100 Hz to express the noise level of the low frequency flicker noise Sy is function of the voltage applied to the gate and drain of the transistor Linear Saturation lg region region 0 Bias condition for lowest nore Drain Source Current p Drain Source Voltage Vg Vot Fig 5 3 Bias condition for lowest flicker noise In the linear region the flicker noise is proportional to the squared drain current o and is much smaller than in the saturation region of the transistor Fig 5 3 TOBIAS WE
76. 1 39 1 42 220 2 40 1 93 1 99 1 43 1 38 1 41 402 3 98 2 79 2 75 2 69 1 89 1 89 405 2 68 2 05 2 08 1 56 1 59 1 62 LO2 2 36 1 98 2 12 1 55 1 54 1 61 min 133 11 170 01 126 15 7 28E D4 6 36 6 30 6 49 6 67 6 65 6 52 average 129 21 155 73 122 21 2 01E 03 3 37 2 56 2 60 2 24 2 13 2 11 Correlation matrix Table 5 2 Correlation results for bias conditions of lowest noise 08 Table 5 2 shows again the individual low frequency and microwave noise results for each device on the waver GaNHFETOS Independent of the low frequency noise parameter the correlation coefficient 15 increasing if the bias condition of the microwave noise is chosen closer to that where the low frequency noise was measured Fig 5 11 e Sy 100Hz Vo 0 5V dBVrms SI 100Hz V s0V Vo 0 5V dB Hz TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 41 141 Correlation caetficient Gate Source Voltage NER An We 100 W Drain Source Current lo rm 5 20 5 200 10 condition for bow frequency nose Measurement 0 0 00 2 00 4 00 6 00 8 00 10 00 12 00 Drain Source Voltage Vp V Fig 5 11 The correlation increases for decreasing source drain voltages This suggests that a better correlation might be found if the microwave and the low frequency noise are examined for an identical bias condition what will be analyzed in 5 5 Me
77. 10 GHz Electronics Letters vol 36 no 5 p 469 71 2 March 2000 O Breitschadel H Grabeldinger B Kuhn F Scholz W Walthes M Berroth I Daumiller K B Schad E Kohn and H Schweizer Short channel AlGaN GaN HEMTs with 70 nm T gate Electronics Letters vol 35 no 23 p 2018 19 11 Nov 1999 A Vescan R Dietrich A Wieszt H Tobler H Leier J M Van Hove P P Chow and A M Wowchak MBE grown AlGaN GaN MODFETs with high breakdown voltage J Cryst Growth Netherlands Journal of Crystal Growth vol 201 202 p 327 31 May 1999 Michael Shur GaAs Devices and Circuits Plenum Press New York 1987 Michael Shur Introduction to Electronic Devices John Wiley amp Sons 1996 G D Vendelin A M Pavio and U L Rohde Microwave circuit design using linear and nonlinear techniques John Wiley amp Sons New York 1990 A Cappy Noise modeling and measurement techniques HEMTs IEEE Transactions on Microwave Theory and Techniques vol 36 no 1 p 1 10 Jan 1988 Jin Wei P C H Chan S K H Fung and P K Ko Shot noise induced excess low frequency noise in floating body partially depleted SOI MOSFET s IEEE Transactions on Electron Devices vol 46 no 6 p 1180 5 June 1999 J A Garrido B E Foutz J A Smart J R Shealy M J Murphy W J Schaff L F Eastman Low frequency noise and mobility fluctuations in AlGaN GaN heterostructure filed effect transistors Applied Physics Letters vol 7
78. 2 0 Search2 Search2 0 Frequency atof Search Search Search2 while Search2 0 Search2 Search2 0 Averaging atoi Search return TRUE void CFlickerNoiseMeasurementDlg OnFind int loop num_listeners pad sad Dev Addr4882 t instruments 32 result 31 char Message 100 if FindGBIPCard return DisplayStatus Searching for SR770 on GPIB bus STATUS FIND pause 1000 for loop 0 loop lt 30 1 instruments loop Addr4882_t loop instruments 31 NOADDR FindLstn 0 amp instruments 1 Addr4882 t result 31 if ibsta amp ERR DisplayStatus ERROR finding SR770 No instruments found on the bus Connected to bus STATUS ERROR return num listeners ibcnt for loop 0 loop lt num listeners 1 pad GetPAD result loop sad GetSAD result loop sprintf Message Trying to query instrument at address d pad DisplayStatus Message STATUS_SR770 Normally replaced quickly by following messages pause 1000 Dev ibdev BOARD INDEX pad sad 13 EOTMODE EOSMODE if ibsta amp ERR no error occured ibclr Dev Clear the devic if ibsta 6 ERR no error clearing the device sprintf Message Trying to get identification string from instrument at address d pad DisplayStatus Message STATUS SR770 Normally replace
79. 2 The representation of the noise setup in the ATS software of Maury Microwave TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 25 141 4 2 Low frequency noise setup The low frequency noise setup is used to characterize the noise properties of transistors in a frequency range of typically 1 Hz to 100 kHz The system is very sensitive to vibrations and electromagnetic disturbances and is therefore placed inside a shielded metal box on an air suspended table In Fig 4 13 a picture of the employed instruments to measure the low frequency noise can be seen spectrum analyzer m Lae digital multimetgt Fig 4 13 Picture of the setup to measure the low frequency noise properties of transistors The output characteristic of the transistors was measured with the parameter analyzer HP4156B of Hewlett Packard The measurement gives on the one hand an idea about the maximum bias ranges of the transistor and is on the other hand necessary to extract the differential resistance of the transistor for a certain bias condition The use of this resistance will be explained later AUTOMATED I CURYVE MEASUREMENT T Werner 2000 Integration Time Bias Setting Sequence C Shot Medium Long First Set Output First Input Bias Start Stop Step Current limit 0 000 2 500 M 0 500 m 1 000 mes Output Bias WDS 05 0 00
80. 2 2 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 8 141 with N the number of carriers and the drain source current and can be used to express the noise level of a transistor for low drain biases As mentioned before equation 2 2 1s only valid for homogenous materials If high drain source voltages are applied to the device the electron distribution in the channel gets inhomogeneous and the use of the Hooge parameter is questionable Furthermore the flicker noise is proportional to in the linear region of the transistor 1 for low drain source voltages Therefore the flicker noise 15 often normalized to 1 2 and also the Hooge parameter normalizes on this parameter In saturation however 1 for elevated drain source voltages the flicker noise is no longer proportional to the squared drain source current and the normalization and with it the use of the Hooge parameter to express the noise level of the transistor gets again doubtful The Hooge parameter takes into account different drain source spacings of compared transistors by the estimation of the number of carriers N in the source drain region N can be calculated by N w n 2 3 with the electron sheet density which is determined by measurement the gate width w and 4 the spacing between source and drain Fig 2 1 drain source Spacing lt gt Source D RAIN Gate gate width
81. 2 letters Command Search return TRUE int CFlickerNoiseMeasurementDlg CalculateAveraging double Start double Stop double Factor int Averaging int NumberOfSeconds 200 max time to wait for 400 samples Averaging int ceil NumberOfSeconds Factor Stop Start 400 return Averaging void CFlickerNoiseMeasurementDlg OnKillfocusEdit UpdateData DIALOG TO DATA keep database consistent DisplayStatus bool CFlickerNoiseMeasurementDlg CreateAveragingRangesStandardFile FILE Handle fprintf Handle This File is used to set the averaging for different frequency ranges n fprintf Handle You can erase or add ranges manually but always sort in increasing n fprintf Handle order of frequencies and use the following format Nn fprintf Handle FREQUENCY AVERAGING n fpr ntf 1 fprintf 1 n fprintf Handle 0 Hz lt FREQUENCY lt 100000 Hz Nn fprintf Handle 2 lt AVERAGING lt 32000 Nn fprintf Handle The entry indicates what AVERAGING should be used Nn fprintf Handle below the specified FREQUENCY Nn fprintf Handle Nn fprintf Handle 2000 Tobias Werner Nn fprintf Handle University of Stuttgart f
82. 250 return 15 if FrequencySpan lt 12500 return 16 if FrequencySpan lt 25000 return 17 if FrequencySpan lt 50000 return 18 else return 19 bool CFlickerNoiseMeasurementDlg PollForMeasurementCompletion CFlickerNoiseMeasurementDlg char serPol int Status delete the OVERLOAD flag ignore any previous overloads ibwrt m InstrumentHandle ERRE7 0 7 if ibsta amp ERR return FALSE start polling do ibrsp pApp gt m_InstrumentHandle amp serPol Check overload ibwrt m_InstrumentHandle ERRS 5 if ibsta amp ERR return FALSE ibrd m_InstrumentHandle buffer 100 if ibsta amp ERR return FALSE buffer ibcnt 1 0 Status atoi buffer if Status 128 while serPol amp 1 0 amp amp ThreadExitFlag FALSE Frage ob man das ueberhaupt braucht if ThreadExitFlag TRUE PostMessage pApp gt m_hWnd WM MEASUREMENT STOPED 0 0 return FALSE return TRUE bool CFlickerNoiseMeasurementDlg PollForSettlingCompletion CFlickerNoiseMeasurementDlg pApp int Span int Status UINT counter 0 do ibwrt m_InstrumentHandle FFTS 5 if ibsta amp ERR return FALSE ibrd m_InstrumentHandle buffer 100 if ibsta 6 ERR
83. 6 no 23 p 3442 44 5 Jun 2000 M E Levinshtein S L Rumyantsev R Gaska J W Yang and M S Shur AlGaN GaN high electron mobility field effect transistors with low 1 f noise Applied Physics Letters vol 73 no 8 p 1089 91 24 Aug 1998 M E Levinshtein F Pascal S Contreras W Knap S L Rumyantsev R Gaska J W Yang and M S Shur Low frequency noise in GaN GaAIN heterojunctions Applied Physics Letters vol 72 no 23 p 3053 5 8 June 1998 M E Levinshtein S L Rumyantsev D C Look R J Molnar M A Khan G Simin V Adivarahan and M S Shur Low frequency noise in n GaN with high electron mobility Journal of Applied Physics vol 86 no 9 p 5075 8 1 Nov 1999 S Rumyantsev M E Levinshtein R Gaska M S Shur A Khan J W Yang G Simin A Ping and T Adesida Low 1 f noise in AlGaN GaN HEMTs on SiC substrates Phys Status Solidi A Germany Physica Status Solidi A vol 176 no 1 p 201 4 16 Nov 1999 Jin Wei P C H Chan S K H Fung and P K Ko Shot noise induced excess low frequency noise JEEE Transactions on Electron Devices vol 46 no 6 p 1180 5 June 1999 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 141 141 I 2 Correlation 29 Stuart L Meyer Data Analysis For Scientists And Engineers John Wiley amp Sons 1975 30 B M Shchigolev Mathematical Analysis Of Observations American Elsevier Publishing 31
84. 70 dataBuffer 30 TRUE pApp return FALSE pApp gt m_MeasurementDataY pApp gt m_DataPointerY atof dataBuffer pApp gt m_DataPointerY M for CurrentValueINDEX 0 CurrentValueINDEX lt 400 CurrentValueINDEX sprintf dataBuffer BVAL 0 d CurrentValueINDEX if pApp gt SendDataSR770 dataBuffer pApp return FALSE if pApp gt ReceiveDataSR770 dataBuffer 30 TRUE pApp return FALSE pApp m MeasurementDataX pApp m DataPointerX atof dataBuffer pApp gt m_DataPointerX return TRUE bool CFlickerNoiseMeasurementDlg GetAveragingRangesFromFile FILE Handle char Line 100 int CurrentLimitINDEX Handle fopen FNM AveragingRanges txt rb if Handle NULL no file existe set options to standard values FNM AveragingRanges txt w Handle if Handle NULL SetDlgControlState ENABLED DisplayStatus Error creating file FNM AveragingRanges txt Write protection Disc space STATUS ERROR return FALSE CreateAveragingRangesStandardFile Handle fclose Handle Handle fopen FNM AveragingRanges txt rb m NumberOfAveragingRanges GetNumberOfLinesInFile Handle 17 fclose Handle Handle fopen FNM AveragingRanges txt rb m AveragingRanges CAveragingRangeTyp
85. 9 Scatter diagrams for a comparison between at 2 GHz representing the microwave noise and Sy 5 S F and at 100 Hz representing the NASE CANF dese at cette desea 38 Fig 5 10 Comparison between the point constellation of 5 1 filled dots solid line and the Hooge parameter in dB non filled dots dashed line 39 Fig 5 11 The correlation increases for decreasing source drain voltages 41 Fig 5 12 Scatter diagrams for a comparison between the minimum noise figure F min at 2 GHz and Sy 8 S F and at 100 Hz GaNHFETOS e 4 Fig 5 13 Dependent on the constellation of the points in the scatter diagram the correlation coefficient is very sensitive to errors Only one additional point can change the correlation coefficient from 0 564 a to 0 223 b 42 Fig 5 14 Scatter diagrams for a comparison between the minimum noise figure F min at 2 GHz and Sy 8 S F and at 100 Hz 44 Fig 5 15 Compared parameters and number of devices on waver GaNHFETIO for 1112 111017 Ni rp RO 45 Fig 5 16 Scatter diagrams for a comparison between the minimum noise figure F min at 2 GHz and Sedndsrat OD sioe n ceu ere db d 47 Fig 5 17 Scatter diagrams for a comparison between at 2 GHz and Sy and 5 logarithmic scale at 100 Hz f
86. D GetDlgItem IDC EDIT FSTOP SetFocus DisplayStatus Stop frequency has to be positive and unequal to unity STATUS WARNING return FALSE if m_FStart gt 100000 SetDlgControlState ENABLED GetDlgItem IDC EDIT FSTART gt SetFocus DisplayStatus Star t frequency has to be below 100000 Hz STATUS WARNING return FALSE if m_FStop gt 100000 SetDlgControlState ENABLED GetDlgItem IDC EDIT FSTOP gt SetFocus DisplayStatus Stop frequency has to be below 100000 Hz STATUS WARNING return FALSE if m_FStart m_FStop SetDlgControlState ENABLED GetDlgItem IDC EDIT FSTART gt SetFocus DisplayStatus Please select different values for start and stop frequency STATUS WARNING return FALSE if m_FStop lt m FStart Dummy m FStart m FStart m FStop m FStop Dummy UpdateData DATA TO DIALOG return TRUE FILE Handle asurementDlg OnButtonEditSetup Handle fopen FNM SystemSetup txt rb if Handle NULL no fil xiste Create file with standard init commands Handle fopen FNM SystemSetup txt w if Handle ULL return CreateSystemSetupStandardFile Handle fclose Handle system Notepad FNM SystemSetup txt NOCH esgibt noch einen
87. E NOISE OF GAN HFETS D 3 4 Waver Transistors marked red were working devices The DC characteristics S parameters and the flicker and microwave noise were measured for these devices Devices marked gray were not working All other device were either working but had differing properties or were not analyzed 6 7 8 910 1112131415 1617 18 19 20 2345 1 FH HJ FJ FJ H H H p HAHH HH HH EH HH a HJ FH HH UR lt HH HH FH H H py H H H H wes HH H H FH HH FH FH FH FH IHH HH HH HH HH HH HH FH HH HH FH HH HAH HH LH FH FH FH HH FH HH H FA Ue FH H HH FH HA FH FH HH 0 HH HH HH IHH HH H lt O Q LLI LL cm HUM Lae M Ls FH HM H HH HH HH H HH H LH HJ H ip HH HH JH H J HH H H H HH H H HH H HH HH HH FH HJ FH FH HJ FH FH HJ FH FH FH FH FH E ron LJ HH HH HH ip HH HH HH HH PH PH HH PH PH FH HA FH HH H HH HH ex HH H HH FH HH HH PH jp HH HH HH HH HH HH PH PH LLL LL i
88. E OF GAN HFETS PAGE 59 141 1 2 Saturation region Expression B 8 can also be written in the form poa 9 res Linearregion Saturation region Tk 10k 12k 354 7156 Differential resistance r Ohms 7k 9 20 363 451 83 B Dk 4k Ok 1 18k 40k 7 359 186 21k 3k Bk 12 37 Fig B 2 Differential resistance the linear and saturation region In the saturation region the differential resistance r is much higher than in the linear region Fig 2 If r gt gt R 4 the second term in B 9 can be neglected and the equation can be written as B 10 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 60 141 a DUT 4 Fig B 3 Derivation of all parameters by the measured voltages and known resistors A schematic drawing of the noise measurement setup is shown in Fig B 3a The resistance of and the potentiometer P are known and the voltages and are determined for each noise measurement The drain source current can be estimated by the expression Vg Vp ipee B 11 and with this result the DC resistance rpc of the FET can be determined _ i B 12 The potentiometer be replaced by two resistors and Fig B 3b which related by TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 61 141 P R R 13
89. EditFstop FOCU DC EDIT P1 OnSetfocusEditPl ETFOCU DC EDIT P2 OnSetfocusEditP2 ETFOCU DIT R OnSetfocusEditR ETFOCU DIT VD OnSetfocusEditVd ETFOCU ETFOCU DIT VG OnSetfocusEditVg ELCHA TS j a ss z ANNNNNANNNA ANO Er z Z E 92999999g9g9g99999g9ggogclH w Z UO Q p Ed Dd DIT VR OnSetfocusEditVr IDC Find OnFind GE IDC COMBO GPIB ADDR OnSelchangeComboGpibAddr DC SETUP SYSTEM OnSetupSystem DC BUTTON EDIT SETUP OnButtonEditSetup DC BUTTON EDIT INIT OnButtonEditInit DC BUTTON EDIT AVG OnButtonEditAvg DC RADIO AVG AUTO OnRadioAvgAuto DC RADIO AVG FILE OnRadioAvgFile C COMBO GPIB ADDR OnSetfocusComboGpibAddr C COMBO GPIB TOUT OnSetfocusComboGpibTout DC COMBO GPIB TOUT OnSelchangeComboGpibTimeOut C EDIT FSTART OnKillfocusEdit 0 Q Q UJ UJ UJ UJ UJ UJ 2 gt Q W bj amp Q UJ B Q E 22 E IDC
90. Hz and harmonics of unknown origin b The FFT spectrum analyzer SR770 of Stanford Research Systems 26 4 16 Schematic view of the low frequency noise 0 0202020000000000 26 4 17 Comparison between the same function proportional to I f plotted versus a linear left and a logarithmic right 27 4 18 The measurement has to be split into several measurement sections because an insufficient number of samples represent the lower part of the SC CUM 28 4 19 The user interface of the program to measure the low frequency noise 30 5 1 Typical result of a low frequency noise measurement with the slope 92 32 SLO ona 8 20 E T RENE eee 31 5 2 Contribution of generation recombination noise sss 32 2 2 BIOS condition TOF lowest Jicker noise 32 JA Dram DIGS We WOOD sedisse 33 5 5 Frequency dependence of the minimum noise figure 34 5 6 Bias conditions for lowest microwave noise 33 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 134 141 Fig 5 7 Different approaches to analyze the correlation between the flicker and the AVE 36 Fig 5 8 Compared parameters and number of devices on wavers for method 36 Fig 5
91. ICROWAVE NOISE OF GAN HFETS PAGE 50 141 5 6 Method C Entire bias range In contrast to the preceding methods where numerous transistors on the same waver were examined method was carried out for 29 working points of a single transistor on waver GaNHFETO as illustrated in Fig 5 19 The minimum noise figure and the flicker noise parameter were estimated for an identical bias condition covering the entire bias range of the analyzed transistor 2 GHz entire bias range sp cu 100 Hz linear region microwave noise flicker noise Fig 5 19 Compared parameters and number of devices on wavers for method C 5 6 1 Results for GANHFETOI The minimum noise figure was measured at 2 GHz 10 GHz and 26 GHz and was min compared to Sy Sz and S 2 at 100 Hz As before S 2 only listed for reference and is no expressive parameter because working points in the saturation region of the transistor are compared to each other TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 51 141 bias voltage 1 f noise microwave noise VG VD Fai2GHz Fimin 1OGH2 4 260 2 i 5 5 5 5 5 8 8 8 8 min 100 36 155 10 95 37 97 83 151 95 66 09 average Correlation coefficient Table 5 6 Correlation results for working points covering the entire bias range of one transistor on GANHFETO1 Table 5 6 shows the indivi
92. ISE OF GAN HFETS analyzed 12 3456 78 910 1112 1314 1516 1718 19 202122 2324 2526 27 28 2930 3132 AEE AR EERE LL PEER LA TE EA EL LL EE d AL LE gt gt gt gt gt gt 3p op op op op o op 3p 3b op F SEE opp ER G gt gt gt E OD OE E op op op 9 H gt SS gt gt X ob op 9p op op op o 3b op ob 9B on 222 Jb ob 9p op of of OB 2 iz LF 9E 9b 9b DLS 9b op EE PEER QUEER PEEP FEEL TE EE EE ES EE ES BREE TEES ELEE LE ES BREE TL EE ES 12 3456 78 910 1112 1314 1516 1718
93. Limit amp amp Stop gt Limit return 2 not in region return 2 void CFlickerNoiseMeasurementDlg pause long delay delay time to pause in milliseconds long stop_ms ms struct timeb tbuf ftime amp tbuf stop ms tbuf time 1000000 stop ms tbuf time stop ms 1000000 stop ms 1000 stop ms tbuf millitm delay do ftime amp tbuf ms tbuf time 1000000 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS ms tbuf time ms 1000000 ms 1000 ms tbuf millitm while ms stop ms end pause int CFlickerNoiseMeasurementDlg GetFrequencySpan double StartFrequency double StopFrequency double FrequencySpan StopFrequency StartFrequency if FrequencySpan lt 0 191 return 0 if FrequencySpan lt 0 382 return 1 if FrequencySpan lt 0 763 return 2 if FrequencySpan lt 1 5 return 3 if FrequencySpan lt 3 1 return 4 if FrequencySpan lt 6 1 return 5 if FrequencySpan lt 12 2 return 6 if FrequencySpan lt 24 4 return 7 if FrequencySpan lt 48 75 return 8 if FrequencySpan lt 97 5 return 9 if FrequencySpan lt 195 return 10 if FrequencySpan lt 390 return 11 if FrequencySpan lt 780 return 12 if FrequencySpan lt 1560 return 13 if FrequencySpan lt 3125 return 14 if FrequencySpan lt 6
94. OGRAM FUNCTIONS 0000000000000000000000000000000000000000000000000000000000000000000000 5 55 5 5 5 5 5 5555 5 55 5 55 5 55 5 5 5 5 5 5 5 5 5 5 5 55 5 5 5 5 5 5 95 5 FUNCTION CheckFor8510CalibFile FileHandler INPUT FileHeandler OUTPUT Number of error terms AUTOR Tobias Werner Universitaet Stuttgart 5 DATE February 29 2000 5 09 0 000000000000000000000000000000000000000000000000000000000000000000000 function y CheckFor8510CalibFile CalFile Look for the word DATA in CalFile exit counter 1 while isempty findstr fgetl CalFile DATA amp exit counter lt 10 exit counter exit counter 1 end Count the occurence of the word DATA in CalFile DATA counter 1 while isempty findstr fgetl CalFile DATA DATA counter DATA counter 1 end switch DATA counter case 3 disp 1 PORT HP8510C CALIBRATION FIL y DATA counter ERROR TERMS FOUND case 12 disp ERROR The algorithm for a 2 Port calibration file is not yet implemented disp EX IT PROGRAM y DATA counter otherwise disp ERROR The file is not a HP8510C calibration file disp E X IT PROGRAM y 0 end return PAGE 117 141 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 118 141 FUNCTION ReadFrequencyList FileHandler INPUT FileHeandler OUTPUT Vektor Frequency list read from ind
95. RENSSELAER POLYTECHNIC INSTITUTE ELECTRICAL COMPUTER AND SYSTEMS ENGINEERING DEPARTMENT PROF MICHAEL SHUR B amp UNIVERSITAT STUTTGART 6050500000009 INSTITUT FUR ELEKTRISCHE UND OPTISCHE NACHRICHTENTECHNIK SINN PROF DR ING MANFRED BERROTH PFAFFENWALDRING 47 110 8TH STREET 70569 STUTTGART TROY NY 12180 DEUTSCHLAND USA Correlation coefficient 0 06 0 19 149 00 4 100 2 4 Hz 2 50 2 75 3 00 3 25 3 50 3 75 26 2 dB CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GALLIUM NITRIDE HETEROSTRUCTURE FIELD EFFECT TRANSISTORS TOBIAS WERNER ELECTRICAL ENGINEERING UNIVERSITY OF STUTTGART MATRIKELNUMMER 1710551 DIPLOMA THESIS DECEMBER 1999 DECEMBER 2000 ACKNOWLEDGMENT This Diploma thesis owes its existence to the help support and inspiration of many people First of all I would like to express my gratitude to Prof Dr Ing Manfred Berroth head of the Institut fur Elektrische und Optische Nachrichtentechnik INT at the University of Stuttgart for taking the responsibility for these studies and making this stay at the Rensselaer Polytechnic Institute RPI in New York possible I also would like to express my sincere appreciation and thankfulness to Prof Michael Shur of the Semiconductor Device Research Group SDG at RPI for his support during this year of my thesis work The discussions and cooperation with all of my colleagues have contrib
96. RNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 33 141 Quantity to represent the flicker noise The spectral voltage noise density Sj measured at the drain of the transistor Meas depends on the following drain bias network which can be represented like depicted in Fig 5 4 Fig 5 4 Drain bias network The potentiometer compare Fig 4 16 to adjust the drain source voltage be described by two resistors and The differential resistance of the transistor depends on the drain voltage and is expressed by the resistor r The resistor R is used to determine the drain current by measuring the two voltages Vp and Vp When adjusting Sy to the desired drain source voltage by changing the resistance ratio between and the bias network will change its resulting resistance R seen by the DUT which can be calculated by res R 5 1 The resistors R and R can not be measured in situ and were determined by measured voltages The derivation of the formula relating the voltages with the resistors can be found in Appendix B 2 Estimation of Rres for the present measurement system The altered drain bias network also affects of course the measured spectral voltage noise density To have a common base to compare measurement results independent of the bias network the spectral current noise density 5 was calculated by 2 5 2 S s 2 res
97. ROWAVE NOISE OF GAN HFETS PAGE 121 141 Appendix G MICROWAVE SWITCH CONTROL The microwave setup is intended to characterize transistors on waver in the microwave frequency range from 2 GHz to 18 GHz concerning S parameters power and noise The system 15 also capable to measure DC characteristics I V curves etc For each measurement type a different system configuration is necessary This means that different instruments have to be combined and as a result the sensitive connections have to be unscrewed and reconnected very often The contact quality declines this affects the repeatability of the contacts and it 1s then required to perform a time consuming system calibration as often as the configuration changes Furthermore it is not possible to determine several properties of the same transistor without releasing the contacts between the microwave probes and the pads of the transistor In case the transistor remains connected vibrations can damage the device while the system configuration is altered To make measurement less time consuming and to get rid of the mentioned disadvantages the system was extended by two microwave switches cie Noise Meter Oscillator Circulator Amp Termination cm 227 3
98. ThreadExitFlag FALSE GetDlgItem IDB Measure gt EnableWindow ENABLED m_hWnd GetSafeHwnd Save window handle in member variable CountDownSeconds CalculateMeasurementTime AfxBeginThread MeasureThread LPVOID this THREAD PRIORITY NORMAL else Button caption Stop DisplayStatus Terminating the measurement process Please wait STATUS STOPPED GetDlgItem IDB Measure gt EnableWindow DISABLED CountDownSeconds 0 SendDataSR770 STCO this Stop measurement ThreadExitFlag TRUE Exit measurement NOCH Reihenfolge veraendert nochmal ueberdenken UINT MeasureThread LPVOID pParam CFlickerNoiseMeasurementDlg pApp CFlickerNoiseMeasurementDlg pParam double Start Stop bool ExitFlag FALSE int CurrentLimitINDEX char Message 200 pApp gt m_NumberOfPassedFrequencyRanges 0 pApp m MeasurementDataX double calloc unsigned pApp gt m_NumberOfAveragingRanges 400 sizeof double pApp m MeasurementDataY double calloc unsigned pApp m NumberOfAveragingRanges 400 sizeof double Start pApp m FStart Stop pApp m FStop pApp m DataPointerX 0 pApp m DataPointerY 0 pApp m FirstAverage TRUE sprintf Message Performing auto offset calibration PostMessage pApp m hWnd WM DISPLAY STATUS WPARAM Message LPARAM STATUS 5 770 PostMessage pApp m hWnd
99. UNCTION CalculateFixtureSParameter ErrorTermArryl ErrorTermArry2 FrequencyArray INPUT ErrorTermArryl ErrorTermArry2 FrequencyArray Electrical Length OUTPUT Fixture S MA Fixture file with frequency list in Magnitude and Angle AUTOR Tobias Werner Universitaet Stuttgart February 29 0000000000 function CalculateFixtureSParameter el e2 FrequencyArray Electrical Length Calculate the S parameters Re Im for the fixture text sprintf Calculating the S parameters for the fixture n disp text Fixture S 1 Fixture S RI 2 Fixture S RI 3 Fixture S RI 4 e2 1 1 3 e2 1 1 sqrt el 3 e Fixture S RI 2 821 812 1 2 2 3 1 2 e2 1 2 2 e2 2 1 3 e2 2 el 1 1 2 el 2 1 1 e1 2 822 1 1 2 e2 1 1 3 1 1 e1 2 5511 3 e1 2 e2 1 1 3 1 1 e1 2 2 812 2 3 convert the parameters from Re Im to Ma Ag 0 Frequency list Fixture S MA 1 FrequencyArray 1000000000 Frequency in GHz Using TEMPORARY the second column of Fixture S MA to store the uncorrected phase information Fixture 5 MA 2 angle Fixture S RI 2 180 pi ONLY TEMPORARY Fixture S MA 5 CorrectPhase Fixture S MA Electrical Length 511 Magni
100. URSOR CLPTSwitchRemoteDlg OnQueryDragIcon return HCURSOR m hIcon int CLPTSwitchRemoteDlg ReadPortStatus possible 1 2 4 low active int ControlByte int DataByte int OptionSetting ControlByte _inp CONTROL if ControlByte SWITCH DISABLE OptionSetting 0 else DataByte inp LPT1 switch DataByte case 255 OptionSetting 0 break CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 130 141 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 131 141 case 254 OptionSetting 1 break case 253 OptionSetting break case 251 OptionSetting break default _outp LPT1 SWITCH OFFLINE OptionSetting inp LPT1 if OptionSetting SWITCH OFFLINE 3 MessageControlString Format An error occured Wrong data d was read from LPT1 OptionSetting MessageBox MessageControlString return OptionSetting void CLPTSwitchRemoteDlg OnChangeConfig int SendByte 0 UpdateData true switch m_SetupConfig case 1 S parameter setup selected SendByte SWITCH S PARAMETER break case 2 Power setup selected SendByte SWITCH POWER break case 3 Noise setup selected SendByte SWITCH NOISE break default Switches turned off SendByte SWITCH OFFLINE break
101. _outp LPT1 SendByte sends byte to port lptl _outp CONTROL SWITCH_ENABLE enable nPrSel gt send 1111100 gt low active Pin3 low check if changes were reflected if ControlOutput SendByte ERROR _outp CONTROL SWITCH DISABLE disable nPrSel gt send 1110100 gt low active Pin3 high void CLPTSwitchRemoteDlg OnExit if exit program gt set port disabled set LPT data ports to disabled outp LPT1 SWITCH OFFLINE set all high LOWACTIVE 0 set LPT control ports to disabled _outp CONTROL SWITCH DISABLE disable nPrSel default 111110100 dec228 if switches were not yet turned offline manually display message if m_SetupConfig 0 MessageBox Switches turned OFFLINE CDialog OnOK int CLPTSwitchRemoteDlg ControlOutput int SendByte int ReadByte ReadByte inp LPT1 if ReadByte SendByte MessageBox ERROR setting switch return ERROR else switch ReadByte case SWITCH S PARAMETER S Parameter essageBox Switched to S PARAMETER configuration break case SWITCH POWER Power essageBox Switched to POWER configuration break case SWITCH NOISE Noise essageBox Switched to NOISE configuration break case SWITCH OFFLINE Switches turned off essageBox Switches turned OFFLINE brea
102. aNHFETIO one of the wavers method M was already applied to As for method the minimum noise figure at 2 GHz was compared to Sy Sy 51 1 and at 100 Hz but this time for an identical bias condition in the saturation region of the transistor same bias point 2GHz 100Hz 21 devices Fig 5 15 Compared parameters and number of devices on waver GaNHFETIO for method B linear region saturfation GaNHFET10 microwave noise flicker noise Dependent on the chosen bias conditions large variations in the correlation result were obtained Therefore two cases where only the gate source voltage Vg was altered will be compared to each other TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 46 141 5 5 1 Results for GaNHFETIO bias condition 1 The minimum noise figure for waver GaNHFETIO was measured at 2 GHz for a working point in the saturation region 1 V 10 V and was compared to Sy Sr 5 1 and a y at 100 Hz for the same bias condition 1 f noise microwave noise VGz 1 0V 10 2 GHz Device dB C01 1 05 006 2 47 007 1 43 E12 2 40 E16 2 47 FO 0 54 F10 2 46 F16 1 30 F17 1 71 G01 1 28 G02 133 505 22 G10 2 44 G17 1 76 G20 2 20 01 1 14 H11 0 84 H12 1 63 105 1 85 1 89 17 1 50 101 41 155 25 101 69 0 0051 0 54 average 92 44 144 80 95 32 0 0536 ra correlation matrix Table 5 4 C
103. about the same point constellation Comparing the flicker noise scales in Fig 5 9 and Fig 5 16 it can be noticed that the flicker noise level is much higher in saturation as in the linear region as mentioned at the beginning of this chapter Fig 5 5 Whereas the correlation coefficient of the same waver for method N was close to zero 0 01 0 20 when correlating and Sy the regression line for this investigation method shows a negative tendency the correlation coefficient is 0 24 0 21 The comparison between S resulted in both cases in a negative correlation coefficient of 0 19 0 19 for method of 0 32 0 21 for method BI The points are as before randomly distributed and the correlation coefficient is still small enough to classify the result as uncorrelated TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 48 141 5 5 2 Results for GaNHFETIO bias condition 2 worst case The same examination was carried out for three additional bias conditions The obtained results differ from the one before and the worst case will be presented in this section The noise level of transistors depend on a vast number of influences and the differing results show how problematical it is to find the right conditions for investigating the reliance between the flicker and the microwave noise In this example all parameters were chosen as before except the gate sour
104. ag GetDlgItem IDC CHECK AVG EnableWindow DisableFlag GetDlgItem IDC COMBO GPIB ADDR gt EnableWindow DisableFlag GetDlgItem IDC COMBO GPIB TOUT gt EnableWindow DisableFlag GetDlgItem IDC SETUP SYSTEM gt EnableWindow DisableFlag GetDlgItem IDC_Find gt EnableWindow DisableFlag if DisableFlag m Button MeasureStop SetWindowText amp Measure GetDlgItem IDB Measure gt EnableWindow DisableFlag else m Button MeasureStop SetWindowText amp Stop GetDlgItem IDB Measure gt EnableWindow DisableFlag void CFlickerNoiseMeasurementDlg OnTimer UINT nIDEvent char sSeconds 2 sMinutes 2 CTime curTime CTime GetCurrentTime int Minutes Seconds m SecondsCounter Seconds m SecondsCounter if m SecondsCounter 0 KillTimer ID 1 m TimeToGo else ConvertToMinutesAndSeconds amp Seconds amp Minutes sprintf sSeconds d Seconds CorrectTimeFormat sSeconds sprintf sMinutes d Minutes CorrectTimeFormat sMinutes m_TimeToGo Format Ss s sMinutes sSeconds UpdateData FALSE CDialog OnTimer nIDEvent void CFlickerNoiseMeasurementDlg CountDownSeconds int Seconds int Minutes 0 char sSeconds 2 sMinutes 2 if Seconds 0 KillTimer ID 1 m SecondsCounter Seconds ConvertToMinutesAndSeconds amp Seconds amp Minutes
105. art frequency Hz Stop frequency Hz Number of averages 1 100 50 100 3000 200 3000 100000 500 Table 4 1 Number of averages for different frequency spans Although there is a possibility to store frequently used settings for a measurement in a file and to recall the settings later from disk not all parameters are saved and have to be manually adjusted each time the frequency range is changed This is a possible source of error if the parameters are not carefully examined The conventional way to measure with the SR770 is the following Recall a set of parameters for a certain measurement range from floppy disc and adjust parameters which could not be recalled Carry out the measurement Save the results of the measurement on floppy disc TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 29 141 This had to be repeated for every individual measurement section Afterwards the measurements were combined by hand in Excel with overlapping sections in mind Because a large number of transistors had to be measured for an analysis of the correlation between the flicker and microwave noise the decision was made to rather invest the time to create a program than to take the time to compile all separate measurements by hand This has several advantages The measurement can run unattended and there is no need to interact while the measurement is running Other things can be d
106. astMessage void CFlickerNoiseMeasurementDlg GPIB error char msg char str 100 char tempbuf 20 Start the Message with the failing GPIB call sprintf str msg Add ibsta information to the Message String strcat str ibsta amp H itoa ibsta tempbuf 16 streat str tempbuf strcat str lt if ibsta 6 ERR strcat str ERR if ibsta amp TIMO strcat str TIMO if ibsta amp END strcat str END if ibsta amp SROI strcat str SRQI if ibsta amp ROS Sstrcat str RQS if ibsta amp CMPL strcat str CMPL if ibsta amp LOK strcat str LOK if ibsta amp strcat str REM if ibsta amp CIC strcat str CIC if ibsta 6 ATN strcat str if ibsta amp TACS strcat str TACS if ibsta 6 LACS strcat str LACS if ibsta amp DTAS strcat str DTAS if ibsta amp DCAS strcat str DCAS strcat str gt Add iberr information to the Message String strcat str niberr itoa iberr tempbuf 10 strcat str tempbuf if iberr EDVR strcat str EDVR DOS Error gt if iberr ECIC strcat str ECIC lt Not CIC gt if iberr ENOL strcat str ENOL No Listener gt if iberr EADR strcat str EADR Address error if iberr EARG strcat str
107. asurementDlg DisplayStatus CString StatusMessage UINT Type UpdateData DIALOG TO DATA used to update changes CBitmap StatusIcon if StatusMessage StatusIcon LoadBitmap IDB ICON EMPTY else switch Type case STATUS INFO StatusIcon LoadBitmap IDB ICON INFORMATION MessageBeep MB break case STATUS QUESTION StatusIcon LoadBitmap IDB ICON QUESTION MessageBeep MB ICONQUESTION break case STATUS WARNING StatusIcon LoadBitmap IDB ICON WARNING MessageBeep MB ICONEXCLAMATION break case STATUS ERROR StatusIcon LoadBitmap IDB ICON ERROR MessageBeep MB ICONEXCLAMATION MB ICONHAND break case STATUS SUCCESS TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 93 141 StatusIcon LoadBitmap IDB ICON SUCCESS MessageBeep OK break case STATUS MEASURE StatusIcon LoadBitmap IDB ICON MEASURE MessageBeep MB break case STATUS TRANS StatusIcon LoadBitmap IDB ICON TRANS MessageBeep MB break case STATUS CARD StatusIcon LoadBitmap IDB ICON CARD break case STATUS CARD FIND StatusIcon LoadBitmap IDB ICON CARD FIND break case STATUS CARD ERROR StatusIcon LoadBitmap IDB ICON CARD ERROR MessageBeep ICONEXCLAMATION break case STATUS SAVE StatusIcon LoadBitmap IDB ICON SAVE break case STATUS 5 770
108. ate length Source Fig 2 1 Dimensions of a transistor 2 2 Microwave noise A field effect transistor FET can be considered as a voltage controlled resistor Thermal noise of this resistor the channel plays the dominant role in the microwave region where the flicker and GR noise vanishes The microwave noise is a function of temperature frequency bias condition device geometry or material characteristics and depends further on the source reflection coefficient I The noise level is usually characterized by the noise figure F in dB TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 9 141 2 2 1 Theory of microwave noise measurement The noise figure of linear two ports is a function of the source admittance 13 The general equation expressing the noise figure of a network as a function of source reflection coefficient Z is NEN m F F 20 Lern min where is the complex source reflecting coefficient that results in the minimum noise figure of the network is the reference impedance for defining usually 50 and is called the equivalent noise resistance and is an empirical constant with the dimension of a resistance relating the sensitivity of the noise figure to the source admittance The noise properties of a device are fully characterized by these four noise parameters To find Zop and the noise figure must be measu
109. besseren Befehl FILE Handle asurementDlg OnButtonEditInit Handle fopen FNM InstrumentInit txt rb if Handle NULL no fil xiste Create file with standard init commands Handle fopen FNM InstrumentInit txt W if Handle ULL return CreateInitInstrumentStandardFile Handle fclose Handle system Notepad FNM InstrumentInit txt NOCH esgibt noch einen besseren Befehl tobi WinExec C NwindowsNNotepad exe SW SHOW FILE Handle asurementDlg OnButtonEditAvg Handle fopen FNM AveragingRanges txt rb if Handle NULL no fil xiste Create file with standard init commands Handle fopen FNM AveragingRanges txt w if Handle ULL return CreateAveragingRangesStandardFile Handle fclose Handle system Notepad FNM AveragingRanges txt NOCH esgibt noch einen besseren Befehl FlickerNoiseMeasurementDlg OnRadioAvgAuto m Radio AveragingMode 0 Auto mode for averaging m Button EditAvg EnableWindow FALSE FlickerNoiseMeasurementDlg OnRadioAvgFile m Radio AveragingMode 1 ranges from file for averaging m Button EditAvg EnableWindow TRUE bool CFlickerNoiseMeasurementDlg CalculateAveragingRanges double FSpan log FStart log FStop_log FStep log FStart FStop int CurrentFrequencyINDEX 5 log
110. ce voltage was decreased by 0 5 V from 1 0 V to 1 5 V 1 f noise microwave noise VGz2 1 5V VD 10Vv Fais 2GHZ 160 58 97 40 0 0040 0 58 150 38 91 82 0 0269 15 correlation matrix Table 5 5 Correlation results for an identical bias conditions in saturation Table 5 5 shows an increased correlation result for Sy and Sj As before the other two flicker noise parameters have no meaning and are presented only for completeness 10V dBVrms 4 Hz Sy 100Hz Vo 1 5V Vo e Sy 100Hz Vc 1 5V Vp 10V VIHz TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 49 141 Correlation coefficient 0 52 0 16 Correlation coefficient 0 55 0 15 82 00 5 134 00 84 00 136 00 4 B6 00 138 00 lt 140 00 4 88 00 144200 co 0 00 4 lt 144 00 92 00 4 146 00 4 94 00 4 148 00 4 96 00 4 gt 150 00 98 00 n 152 00 FOr 154 00 4 100 00 gt 156 00 4 102 00 4 S 158 00 104 00 4 amp 160 00 4 106 00 r r 162 00 0 50 1 00 1 50 2 00 2 50 3 00 3 50 0 50 1 00 1 50 2 00 2 50 2GH Vg 1 1 5V Vo 1 OV d B Fmi 2G Hz V5 7 1 a SV V571 OV d B Fig 5 17 Scatter diagrams for a comparison between F nin at 2 GHz and Sy and 5 logarithmic scale at 100 Hz for a altered gate source voltage Fig 5 17 shows again the two scatter diagrams this time for th
111. current noise density S and the Hooge parameter amp p which varied between 132 to 109 dB Hz and 7E 4 to 1E 1 at 100 Hz respectively over all devices The flicker noise versus frequency in the linear region expressed by the relative spectral current noise density S 2 is shown in Fig 3 8 The slope of the curve varied between 0 9 and 1 4 as described in literature 23 but was typically close to unity 15 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 16 141 Low Frequency Noise VG 0 00V VD 0 50V GaNHFET10 F07 Relative spectrum current noise density 1 10 100 1000 10000 100000 Frequency f Hz Fig 3 8 Typical low frequency noise of a GaN HFET biased in the linear region The slope of the red line is ideally proportional to I f The pure flicker noise was often superimposed by generation recombination noise which leads to deviations from a straight line for higher frequencies The spikes are due to disturbances from the power system frequencies 60 Hz and frequencies at 15 kHz of unknown origin and their harmonics The minimum Hooge parameter was found to be low as 7 10 at zero gate bias and a drain source voltage of 0 5 V what is comparable to results reported in 12 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 17 141 4 MEASUREMENT SETUP To characterize noise of transistors specific s
112. d quickly by following messages ibwrt Dev IDN n 71 if ibsta amp ERR no error sending identification string ibrd Dev buffer 100 if ibsta amp ERR no error getting identification string buffer ibcnt 1 0 eliminate the carriage return Asc 10 if strncmp buffer m IdentificationString strlen m IdentificationString 0 sprintf buffer SR770 found at address d pad DisplayStatus buffer STATUS SR770 m PrimaryInstrumentAddress Addr4882 t pad Init the Instrument address m SecondaryInstrumentAddress Addr4882 t sad m InstrumentHandle Dev m GPIB Address SetCurSel m PrimaryInstrumentAddress 1 UpdateData DATA TO DIALOG break quit the for loop End of FOR loop if loop gt num_listeners DisplayStatus FFT Network Analyzer 58770 has not been found nPlease make sure that the device is powered on and is properly connected to the GPIB bus STATUS ERROR return void CFlickerNoiseMeasurementDlg OnSelchangeComboGpibAddr m PrimaryInstrumentAddress m GPIB Address GetCurSel 1 bool CFlickerNoiseMeasurementDlg InstrumentActive char IdentificationString int Dev pad sad DisplayStatus Searching for SR770 on GPIB bus STATUS FIND pause 1000 pad GetPAD m PrimaryInstrumentAddress sad GetSAD m PrimaryInstrumentAddress Dev ibdev BOARD INDEX pad sad m Tim
113. des ClassWizard generated virtual function overrides AFX_VIRTUAL CLPTSwitchRemoteApp public virtual BOOL InitInstance AFX VIRTUAL Implementation AFX_MSG CLPTSwitchRemoteApp NOTE the ClassWizard will add and remove member functions here DO NOT EDIT what you see in these blocks of generated code AFX_MSG DECLARE MESSAGE AFX INSERT Microsoft Visual C will insert additional declarations immediately before the previous line endif defined AFX LPTSWITCHREMOTE 0BD27EC4 5C15 11D4 B354 0006295283FA INCLUDED G 3 1 2 Header file LPTSwitchRemoteDle h LPTSwitchRemoteDlg h header file fif defined AFX LPTSWITCHREMOTEDLG 0BD27EC6 5 15 1104 8354 0006295283FA INCLUDED define AFX LPTSWITCHREMOTEDLG 0BD27EC6 5 15 1104 B354 0006295283FA INCLUDED _ if 5 VER gt 1000 pragma once endif MSC VER gt 1000 CLPTSwitchRemoteDlg dialog class CLPTSwitchRemoteDlg public CDialog Construction public UINT GetLPTPortBaseAddress int PortNumber TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 128 141 int ControlOutput int
114. dialog you will need the code below to draw the icon For MFC applications using the document view model this is automatically done for you by the framework void CFlickerNoiseMeasurementDlg OnPaint if IsIconic CPaintDC dc this device context for painting SendMessage WM ICONERASEBKGND WPARAM dc GetSafeHdc 0 Center icon in client rectangle int cxIcon GetSystemMetrics SM CXICON int cyIcon GetSystemMetrics SM CYICON CRect rect GetClientRect amp rect int x rect Width cxIcon 1 2 int y rect Height cyIcon 1 2 Draw the icon dc DrawIcon x y m hIcon else CDialog OnPaint The system calls this to obtain the cursor to display while the user drags the minimized window HCURSOR CFlickerNoiseMeasurementDlg OnQueryDragIcon return HCURSOR m hIcon void CFlickerNoiseMeasurementDlg GetLastVariableSettingsFromFile FILE Handle char Line 100 Handle fopen FNM LastSettings txt rb if Handle NULL no file existe set options to standard values SetStandardVariableValues else file exists read lines from file fgets Line 100 Handle VERSION_INFO if strncmp Line VERSION STRING strlen VERSION STRING 0 version identical fgets Line 100 Handle skip file name fgets Line 100 Handle m Vg atof Line fgets
115. dual low frequency and microwave noise results for different bias conditions of a single device waver GaNHFETOI In the first two columns the bias condition i e the voltage at the gate and drain is indicated The following three columns hold the individual flicker noise parameters and the last three columns specify the minimum noise figure at three frequencies The correlation coefficient shows no frequency dependence and is close to zero independent of the selected low flicker noise parameter Sy 100Hz dBVrms 4 Hz 5 100 2 dBArms Hz TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 52 141 Correlation coefficient 0 06 0 19 Correlation coefficient 0 00 0 18 148 00 149 00 Vel 150 00 4 151 00 152 00 4 54100 2 dBArmsi4 Hz 153 00 4 154 00 4 155 00 4 i i 156 00 3 00 3 20 950 10001 1050 1100 1150 12 00 Fel 2GHz 88 Fminl 26GHz dB Fig 5 20 Scatter diagrams for a comparison between the minimum noise figure min at a b 2 GHz and c d at 26 GHz and Sy and S at 100 Hz Fig 5 20 shows again the scatter diagrams for a comparison between Sy and 5S measured at 100 Hz and F min bias condition of the same transistor with its corresponding F and flicker noise at 2 GHz and 26 GHz This time a dot in a diagram represents a certain parameter As expected the microwave noise level at 26 GHz is much hig
116. e void CorrectTimeFormat char Element void ConvertToMinutesAndSeconds int Seconds int Minutes int m_SecondsCounter void CountDownSeconds int Seconds bool PollForSettlingCompletion CFlickerNoiseMeasurementDlg pApp int Span char m FileNameRaw 200 int m ComboRefStartValue void SetDlgControlState bool DisableFlag HWND m hWnd int m DataTransferTime bool CreateInitInstrumentStandardFile FILE Handle bool CreateSystemSetupStandardFile FILE Handle bool CreateAveragingRangesStandardFile FILE Handle CWinThread m MeasureThread bool InstrumentInitialized int CalculateAveraging double Start double Stop double Factor standard constructor CFlickerNoiseMeasurementDlg CWnd pParent NULL functions prototypes voidSetStandardVariableValues boolCalculateAveragingRanges boolCheckVariableRanges boolInstrumentActive char IdentificationString boolGetAveragingRangeEntry char Line double Frequency int Averaging int GetNumberOfLinesInFile FILE Handle boolGetAveragingRangesFromFile int GetFrequencySpan double StartFrequency double StopFrequency voidGetLastVariableSettingsFromFile boolGetMeasurementData CFlickerNoiseMeasurementDlg pApp boolProcessData char data double MeasurementData int DataPointer boolPollForCommandCompletion CFlickerNoiseMeasurementDlg boolPollForMeasurementCompletion CFlickerNoiseMeasurementDlg pApp in
117. e oe of ore of ph of eia ore oe of oS ob pj CHD DH DH MEJ DJ DHE E DH of ME CHO CE oe px HE E P 2 E DE pa o ob ob obo XE CHa Dh oe Rob ob ob ob ph ph ode ph pda ph DH OH podes DH S DHE DE Dd T DHE DH DE DE CHO DH DH M DA DH DH H DE CHE DA DH Fig D 3 Waver map of GaNHFETOI 10 1112 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS D 2 GaNHFET04 D 2 1 Typical material properties Mobility 1000 cm2 Vs Carrier sheet density 1 10 cm 2 Aluminum mole fraction 15 20
118. e calloc unsigned m NumberOfAveragingRanges sizeof CAveragingRangeType for CurrentLimitINDEX 0 CurrentLimitINDEX lt 17 CurrentLimitINDEX fgets Line 100 Handle skip line used for comments for CurrentLimitINDEX 0 CurrentLimitINDEX lt m NumberOfAveragingRanges CurrentLimitINDEX fgets Line 100 Handle GetAveragingRangeEntry Line amp m AveragingRanges CurrentLimitINDEX Frequency amp m AveragingRanges CurrentLimit INDEX Averaging if m_AveragingRanges CurrentLimitINDEX 1 Frequency 100000 m AveragingRanges CurrentLimitINDEX 1 Frequency 100000 Last frequency has to be 100000 fclose Handle return TRUE int CFlickerNoiseMeasurementDlg GetNumberOfLinesInFile FILE Handle char Line 100 LastLine 100 FILE BufferHandle BufferHandle Handle int LineCounter 0 while feof Handle fgets Line 100 Handle if strncmp Line LastLine strlen Line if ValidCommand Line LineCounter strcpy LastLine Line Handle BufferHandle return LineCounter TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 99 141 bool CFlickerNoiseMeasurementDlg GetAveragingRangeEntry char Line double Frequency int Averaging char Search Search2 Search Line Search2 Search while Search2 amp amp Search
119. e as well the computer supply to power the microwave switches The parallel port LPT of a computer uses TTL compatible voltages and is thus able to send control information to the switch Unfortunately the Maury Software ATS is protected against illegal copies by a hardware key at the parallel port The following subchapters describe the hardware designed to prevent from interaction between the hardware key and the control commands intended for the microwave switch G 2 1 Schematic The hardware key itself passes all signals from the input to the output To control a microwave switch a positive voltage of 5 V has to be sent to one of its three TTL ports to activate the normally opened corresponding signal path The switch only accepts a single high level at one of the three TTL ports and must be protected against unsuited signals This is what the circuit 1s intended for 7 MW Switch Control 66646 9999 1 TTL 1 5V DC mi fm amp Coil 12V DC 7 Coil 12V DC 8 Coil GND Microuave Suitch Control via LPT Tobias Werner TITLE LPTsuitchctrl Document Numbers REV 171 1 1 Date 27 25 2000 09 53 34a Sheet 1 1 Fig G 3 The schematic of the microwave switch control High 2 0 V to 5 2 V Low 0 V to 0 8 V Automated Tuner System TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 124 141 The heart of the schematic sho
120. e bias condition 1 5 V and 10 V of waver GaNHFETIO As can be seen from the two graphs comparing Sy and 5 with the point constellation shows a clear negative tendency This could indicate that there 1s a possible partly correlation between the flicker and the microwave noise and shows again the sensitivity of the analysis When using a non logarithmic scale for the flicker noise parameters dB is a logarithmic scale the correlation coefficient decreases to 0 25 0 21 and 0 28 0 20 for Sy and S as parameters respectively and is illustrated in Fig 5 18 b Correlation coefficient 0 25 0 21 Correlation coefficient 0 28 0 20 5 00 09 4 Du eos 4 50E 09 4 00 09 4 3 50E 09 4 3 00E 09 4 2 50E 09 4 2 DOE 08 1 50 09 4 1 00E 09 4 5 00E 10 cen 0 00E 00 4 EHI 1 Ke 0 00E 00 0 50 1 00 1 50 2 00 2 50 3 00 3 50 0 50 1 00 1 50 2 50 lt c gt i I c e F D Fmin 2GHz V2 1 5V Vp 10V dB Fmin 2GHz Vo 1 5V Vo 10V dB Fig 5 18 Scatter diagrams for a comparison between F nin at 2 GHz and Sy and 5 non logarithmic scale at 100 Hz for an altered gate source voltage To consider all possible cases and to analyze the deviation of the correlation coefficient for different working points would have required additional measurements TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE M
121. e device geometry or material quality affect the noise level of transistors In order to understand the complex interplay between these parameters and to be able to reduce the noise the underlying physical phenomena have to be studied and understood Often only the assistance of simulation tools can help to gain an idea about the interaction of different noise origins but this requires valid noise models The dependence of two noise types namely the flicker or 1 f noise and the microwave noise of gallium nitride GaN heterostructure field effect transistors HFETs at room temperature was studied TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS 5 141 Whereas the microwave noise originates in thermal fluctuations of electrons experts still lead controversial discussions about the origins of flicker noise For GaN HFETs with the electron transport in the two dimensional electron gas a possible explanation is that traps in the space charge region under the gate generate the flicker noise The trapping and detrapping of electrons gives rise to a 1 f modulation of the channel conductance and this produces flicker noise the drain current 1 The flicker and the microwave noise are the dominant sources of noise in the low frequency and microwave region respectively The demand of industry goes to ever increasing frequencies and one could wonder why it is important to analyze low frequency no
122. e select the file which contains the calibration information WITHOUT FIXTURE n Type the name of the file a peace open and control the selected fil CalFilel fopen file namel r if CalFilel 1 text sprintf File s not found file namel disp text disp EX IT PROGRA M return end NumberOfErrorTerms1 CheckFor8510CalibFile CalFilel if NumberOfErrorTerms1 fclose CalFilel return end 5 55 5 5 5 5 SELECT AND OPEN CALFILE WITH FIXTURE 5 5 5 5 5 95 file name2 file namel while strcmp file namel file name2 file name2 input nPlease select the file which contains the calibration information WITH FIXTURE n Type the name of the file Lp yas if strcmp file namel file name2 text sprintf nERROR You have chosen the same file as befor Try again disp text end end open and control the selected fil CalFile2 fopen file name2 r if CalFile2 1 text sprintf File s not found file name2 disp text disp EX IT PROGRA M return end NumberOfErrorTerms2 CheckFor8510CalibFile CalFile2 if NumberOfErrorTerms fclose CalFile2 ftertufn end enter electrical length if known Electrical Length str2num input nPlease specify the ELECTRICAL LENGTH of the fixture automatically If the electrical length is calculated by the program please mak current directory the fo
123. eOut EOTMODE EOSMODE if ibsta amp ERR SetDlgControlState ENABLED sprintf buffer Unable to open device SR770 at address d m PrimaryInstrumentAddress DisplayStatus buffer STATUS ERROR ibonl 0 0 free GPIB card return FALSE TOBIAS WERNER ibclr Dev Clear the devic if ibsta amp ERR SetDlgControlState ENABLED sprintf buffer Unable to clear device SR770 at address d m PrimaryInstrumentAddress DisplayStatus buffer STATUS ERROR ibonl 0 0 free GPIB card return FALSE ibwrt Dev IDN n 71 if ibsta amp ERR SetDlgControlState ENABLED DisplayStatus ERROR SR770 turned on Connected to GPIB bus Time out Address STATUS ERROR ibonl 0 0 free GPIB card return FALSE ibrd Dev buffer 100 if ibsta amp ERR SetDlgControlState ENABLED DisplayStatus ERROR SR770 turned on Connected to GPIB bus Time out Address STATUS ERROR ibonl 0 0 free GPIB card return FALSE buffer ibcnt 1 0 if buffer IdentificationString strlen IdentificationString 0 sprintf buffer SR770 found at address d pad DisplayStatus buffer STATUS SR770 pause 1000 m InstrumentHandle Dev m PrimaryInstrumentAddress Addr4882 lt m SecondaryInstrumentAddres
124. ear region of the transistor is considerably higher it behaves in reverse with the flicker noise Due to this bias dependence three approaches are chosen to analyze a possible dependence between the two dominant types of noise in the microwave and low frequency domain 1 Lowest noise in saturation is compared to Sy S 91 1 and a in the linear region for several transistors 2 Identical bias condition is compared to Sy and 5 for an identical bias condition 1n saturation for several transistors 3 Entire bias range is compared to Sy and S for an identical bias condition for one transistor and several bias conditions This method was also carried out for F nin at additional frequencies of 10 GHz and 26 GHz With the aid of the correlation coefficient the dependence of the two noise types is evaluated and scatter diagrams are used to verify graphically the calculated result The spectral current noise density S appears to be the most expressive parameter to refer to the flicker noise and correlation coefficients between 0 and 0 3 with an absolute error of 0 2 due to a small number of observations are obtained The small correlation coefficient is supported by a widely spread point constellation covering the entire plot area of the scatter diagrams which shows no dependence between the flicker and the microwave noise TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 1
125. ectively Associated Gain G_ dB a e TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 15 141 b 2000 Tobias Wemer 2000 Tobias Werner 15 92 dB 0 5V max 1 0 Maximum Available Gain 6 8 8 0 Drain voltage 5 0 10 12 Drain voltage Vp V 10 12 Fig 3 7 Bias dependence of a the associated gain at minimum noise figure G and b the maximum available gain Gmax G The difference between and Gmax gets smaller and smaller as Vp increases This means 1 always smaller as G decreases quickly for decreasing drain source voltages that the tradeoff between highest power and lowest noise gets more favorable as Vp increases Comparing Fig 3 6a and Fig 3 7a we can observe that the bias condition for a negligible deviation between and Gay and the optimum coincide at the same bias condition This is important for low noise applications in first stage amplifiers since not only lowest noise but also highest gain matters as we can see from the Friis s cascade equation 20 1 3 3 with F the noise figure of the i th stage G the gain of the i th stage and the overall noise figure The noise level of the flicker or 1 f noise in the linear region of a transistor can be conveniently described by two parameters the relative spectral
126. ed as COV y y gt x Qui C 2 and the standard deviations defined as indicates the number of points in data range TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 64 141 Equations C 2 and C 3 contain the average or mean x and y of the two data ranges which can be calculated by _ I Xi 1 Ms Xx Mes i C 4 n 1 C 2 Least Square Regression To visualize a dependence of two data ranges one range is plotted versus the other in a so called scatter diagram Fig C 1 Correlation coefficient 0 647 3 50 4 3 00 e gt z 1 S 2 50 z i gt gt 2 00 I e 150 4 e 1 00 0 50 0 50 1 00 1 50 2 00 2 50 3 00 3 50 20 1 5 210 dB Fig C 1 Example of a scatter diagram An intuitive idea of finding a linear approximation of the relationship between the independent and dependent variables can be obtained by drawing a line that is chosen to minimize the squared distances between the points and the line Fig C 2 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 65 141 Correlation coefficient 0 647 3 50 4 Fmin 2GHz V z 1V Voz8V dB 0 50 T T T T 0 50 1 00 1 50 2 00 2 50 3 00 3 50 Fnn 2GHz Vg 1 5V Vp 10V dB
127. ed range In this case a misinterpretation of the obtained result gets very likely C 4 The Coefficient of Determination The regression line is defined to be the line lying closest to the data points It is often useful to be able to evaluate how close the line approximate the points what 15 often referred to as the goodness of fit of the regression line Fig C 4 shows a comparison of two scatter plots with regression lines with the same slope Although both diagrams have the same regression line the left 1s referred to have a better fit than the other because the points are closer to the line Correlation coefficient 0 990 Correlation coefficient 0 820 Coefficient of determination 0 981 Coefficient of determination 0 672 3 50 3 00 2 20 8 dB 8 48 Fmin 2GHz V z 1V Vn Fminj2 GHz Vn 1 00 0 50 0 50 1 00 1 50 2 00 2 50 3 00 3 50 Frnnt2GHz V 1 5V p 10V dB Frnint2GHz V 1 5V Vp 10V dB Fig C 4 Comparison of goodness of fit for two regression lines with the same slope TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 67 141 The measure of relative closeness for evaluating the goodness of fit is called the coefficient of determination Because of its relationship to the correlation coefficient this measure is generally referred to as the 2 The coefficient of determination is the square of the correlation coefficient It descr
128. efine STATU define STATU define STATU define STATU UESTION 1 RNING 2 RROR 3 UCCESS 4 zi D define EOSMODE define COMMAND define MEASUREMENT IC CES Ae volatile bool ThreadExitFlag char buffer 200 ad ndthread S S S S S define STATUS MEASURE 5 define STATUS TRANS 6 define STATUS CARD 7 define STATUS FIND 8 define STATUS CARD 9 define STATUS SR770 10 define STATUS SR770 OK define STATUS SAVE 12 define STATUS CARD ERROR 13 define STATUS STOPPED 14 define VERSION STRING V1 4 Version of parameter file define BOARD INDEX 0 Board index define EOTMODE 1 Enable the END messag Disable the EOS mode Used for poll function Used for poll function MILII PMEIEIMM MI AAA aH II PP MM II PP MUI ATTA TTT ATA TT CAboutDlg dialog used for App About class CAboutDlg public CDialog public CAboutDlg Dialog Data AFX DATA CAboutDlg enum IDD IDD ABOUTBOX DATA ClassWizard generated virtual function overrides AFX VIRTUAL CAboutDlg protected virtual void DoDataExchange CDataExchange pDX DDX DDV support AFX VIRTUAL Implementation protected AFX MSG CAboutD1g AFX MSG DECLARE MESSAGE MAP CAboutDlg CAboutDlg CDialog CAboutDlg
129. ent Phase INDEX 2 130 if Phase jump counter Start phase Fixture S S21 Current Phase 1 2 Start frequency Fixture S S21 Current Phase 1 1 end Phase jump counter Phase jump counter 1 End phase Fixture S A S21 Current Phase INDEX 2 Stop frequency Fixture S A S21 Current Phase INDEX 1 end end delta phase Phase jump counter 2 180 Start phase End phase degree delta frequency Stop frequency Start frequency 1E9 Frequency in Hz slope delta phase delta frequency Velocity of light 3E8 m s electrical length slope Velocity of light 2 180 y electrical length return FUNCTION RangePhase Phase INPUT Phase in degree OUTPUT Phase ranged on 180 180 AUTOR Tobias Werner Universitaet Stuttgart DATE March 02 2000 function RangePhase Phase in degrees Sign of phase sign Phase Phase mod Phase 360 Number of phases size Phase 1 for Current Phase INDEX 1 Number of phases if Phase Current Phase INDEX gt 180 RangedPhase Current Phase INDEX Sign of phase Current Phase INDEX 360 Phase Current Phase INDEX else RangedPhase Current Phase INDEX Sign of phase Current Phase INDEX Phase Current Phase INDEX end end y RangedPhase return TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MIC
130. erential resistance r in the bias point P The resulting differential resistance for the bias point can be calculated by the relation ala Ga B 19 with suo V pi V pii a a m 20 21 Ipi41 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 63 141 Appendix CORRELATION If a dependence of two data ranges in our case the parameter representing the noise level of the flicker noise and the microwave noise is not directly perceptible a mathematical approach can help to identify the degree of reliance between the two ranges The correlation coefficient is such a mathematical aid and a short overview will be given how it was applied to analyze the data C 1 The Correlation Coefficient The correlation coefficient is a measure to compare if two ranges of data X and Y are dependent on each other That 1s whether large values of one set are associated with large values of the other positive correlation correlation coefficient ry y 1 whether small values of one set are associated with large values of the other negative correlation ry y 1 or whether values in both sets are unrelated correlation near zero ry y 0 The correlation coefficient is defined as the covariance of two data sets X and Y divided by the product of their standard deviations 29 PX Y Oy with the covariance defin
131. etups with various instruments have to be used The low frequency noise has to be measured with a different system as the microwave noise In the succeeding chapters we will discuss what instruments were used for the measurements and how they were assembled with each other for both systems the low frequency noise and the microwave noise setup 4 1 Microwave noise setup A picture of the microwave setup is shown in Fig 4 1 noise figure meter ____ power meter parameter analyzer oscillator control network analyzer SOUTCE Tuner load turer extender oe probe station computer control Fig 4 1 The microwave setup to characterize DC characteristics S parameters noise and power of transistors in the microwave frequency range The instruments of the microwave setup were already available but it was necessary to assemble and calibrate the system Design decisions concerning the connections had to be made and several problems to calibrate the system had to be solved Most important 1 to make the system insertable compare Appendix F Fixture Characterization what helps to keep the calibration process simple This 1s not always possible and techniques like Adapter Removal have to be used TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 18 141 If even this method cannot be applied custom solutions have to be developed For this purpose a program was written in Matlab Appendix
132. etween the results for an S parameter determination of a commercial system blue and the presented Matlab program red On the left side is the magnitude and on the right side the phase of the DG OCIO e e Dd 115 The new microwave setup to characterize DC characteristics S parameters power and noise of transistors in the microwave frequency range from 2 GHz to 18 GHz Microwave switches at the input and output allow to switch between different setup configurations without having to COMO HOLE Te 121 left The two microwave switches with indicator LEDs connected to the control cable Rigid microwave cables at the inputs and outputs are mounted to connect to the system components right A mounted microwave switch on the load side of the 1 122 The schematic of the microwave switch control esses 123 The layout left and the component placement right of the microwave COMI ON ete delli Oa cata gt 125 Top and bottom view of the printed circuit board to control the microwave EM MEI E 126 The user interface to control the microwave switches 1 127 H 2 Table index Table 4 1 Number of averages for different frequency 5 26 Table 5 1 Correlation results for bias conditions of lowest noise
133. g DoDataExchange pDX AFX DATA CFlickerNoiseMeasurementD1lg DDX Control pDX IDC COMBO LOG SAMPLE m ComboLogSample DDX Control pDX IDC STATUS ICON m StatusIcon DDX Control pDX IDB Measure m Button MeasureStop DDX Control pDX IDC BUTTON EDIT AVG m Button EditAvg DDX Control pDX IDC COMBO GPIB TOUT m Combo GPIB Timeout DDX Control pDX IDC COMBO GPIB ADDR m GPIB Address DDX Text pDX IDC STATUS m Status DDX Text pDX IDC EDIT VR m Vr DDX Text pDX IDC EDIT VG m Vg DDX Text pDX IDC EDIT VD m Vd DDX Text pDX IDC EDIT R m R DDX Text pDX IDC EDIT P2 m P2 DDX Text pDX IDC EDIT Pl m P1 DDX Text pDX IDC EDIT FSTOP m FStop DDX Text pDX IDC EDIT FSTART m FStart DDX Check pDX IDC CHECK AVG m CheckAverage DDX Radio pDX IDC RADIO AVG AUTO m Radio AveragingMode DDX Check pDX IDC CHECK LOG AVG m CheckLogAverage DDX Text pDX IDC TIMETOGO m TimeToGo AFX DATA MAP MESSAGE CFlickerNoiseMeasurementDlg CDialog AFX MSG MAP CFlickerNoiseMeasurementDlg ESSAGE WM START COUNTDOWN OnStartCountdown ESSAGE WM DISPLAY STATUS OnDisplayStatus ESSAGE WM SAVE DATA OnSaveData ESSAGE WM MEASUREMENT STOPED OnMeasurementStoped SYSCOMMAND PAINT ERY DRAGICON ICKED IDB Measure OnMeasure LFOCUS IDC EDIT OnKillfocusEditR ETFOCUS IDC EDIT FSTART OnSetfocusEditFstart ETFOCUS IDC EDIT FSTOP OnSetfocus
134. gs have no influence on the measurement process and are only written to the data file for further reference The value of VG and VD will be used as a part of the file name Measurement range N The start and stop frequency determine the frequency range of the measurement The start and stop frequency will be used as a part of the file name e g 3 004 Vg 7 00V start 1 0Hz stop 100000 0Hz_raw txt The other file contains converted data which is adjusted to equidistant steps in the logarithmic frequency scale The file name is created as before except the appended text Vd 3 00V Vg 1 00V start 1 0Hz stop 100000 0Hz txt The data files are stored in the same directory as the program is GPIB settings located The time out indicates the time the GPIB interface will wait for a command to succeed If the instrument does not respond the program will be aborted and an error message will be shown The standard value for this parameter is 10s The address should be adjusted to the GPIB address of the instrument The standard setting is 10 Use the Find button to scan for the address of the SR770 Automated Flicker Noise Measurement Wemer 2000 BEY Initialization file If you push this Edit button the initialization file will open This file contains GPIB commands for the SR770 refer manual chapter 6 to E initialize the instrument You can sat 10 Hz Stop 500 He adjust the initial
135. hases if abs Fixture S A S21 Current Phase INDEX 2 Ideal phase Current Phase INDEX 160 Fixture S A S21 Current Phase INDEX 2 Fixture S A S21 Current Phase INDEX 2 180 if Fixture S A S21 Current Phase INDEX 2 180 Fixture S A S21 Current Phase INDEX 2 Fixture S A S21 Current Phase INDEX 2 360 end end end y Fixture 5 521 2 return TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 120 141 0000000000000000000000000000000000000000000000000000000000000000000000000 00000000000000 FUNCTION CalculateElectricalLength Fixture S MA 2 INPUT Fixture S MA 1 Frequency Fixture S MA 2 uncorr phase of 12 OUTPUT Electrical length AUTOR Tobias Werner Universitaet Stuttgart DATE arch 02 2000 000000000000000000000000000000000000000000000000000000000000000000000000000000000000000 5 55 5 5 5 5 5 5555 5 55 5 5 5 5 5 5 5 55 5 55 5 5 55 5 5555 55 5 5 5 5 55 5 5 5 5 55 5 595 95 9 function CalculateElectricalLength Fixture S 521 Get array size Number of phases size Fixture S A 521 1 Phase jump counter 0 for Current Phase INDEX 1 Number of phases 1 if Fixture 5 S21 Current Phase 1 2 Fixture S A S21 Curr
136. her than at 2 GHz compare Fig 5 5 The constellation of points at 26 GHz seems to be shifted towards higher noise figures and cover a larger range 9 7 dB to 12 5 dB in comparison to 2 GHz 2 7 dB to 3 6 dB All diagrams a to d show a point constellation which support the correlation coefficient close to zero and show no association between the two noise types The individual results for the three presented methods will be compared to each other and a general statement concerning the topic of this investigation if the flicker and the microwave noise of GaN HFETs is correlated will be given in the following chapter the conclusion c Correlation coefficient 0 09 0 18 Correlation coefficient 0 03 0 19 54 00 84 00 4 Va V i 95 00 55 12 95 00 1 eum e523 96 00 ioi 5 2 96004 4 At gt rte x 97 00 5 97 004 at 98 00 4 4 98 00 99 00 99 00 100 00 100 00 101 00 4 is 101 00 102 00 i 102 00 4 2 60 2 80 3 00 320 3 40 3 80 9 50 10 00 10 50 11 00 11 50 12 00 12 50 Fe 2GHz dB F4 26GHz dB 12 50 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 53 141 6 CONCLUSION The correlation between the flicker and the microwave noise of GaN HFETs at room temperature was investigated For this purpose a microwave system had to be assemb
137. ialog is dismissed with OK else if nResponse IDCANCEL TODO Place code here to handle when the dialog is dismissed with Cancel Since the dialog has been closed return FALSE so that we exit the application rather than start the application s message pump return FALSE TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 129 141 G 3 1 4 Implementation file LPTSwitchRemoteDle cpp LPTSwitchRemoteDlg cpp implementation file include stdafx h include LPTSwitchRemote h include LPTSwitchRemoteDlg h include lt conio h gt include lt dos h gt ifdef _DEBUG define new DEBUG NEW undef THIS FILE static char THIS FILE _ FILE endif define LPT1 0x378 lptl define STATUS LPT1 1 lptl define CONTROL LPT1 2 lptl define ERROR 0 define SUCCESS 1 define SWITCH_ENABLE 236 11101100 define SWITCH DISABLE 228 11100100 define SWITCH_OFFLINE 000 DBO DB1 DB2 DB3 DB4 DB5 DB6 DB7 high 255 define SWITCH S PARAMETER 001 DBO low DB1 DB2 DB3 DB4 DB5 DB6 DB7 high 254 define SWITCH POWER 002 DB1 10w DBO DB2 DB3 DB4 DB5 DB6 DB7 high 253 define SWITCH NOISE 004 DB2 low DBO DB1 DB3 DB4 DB5 DB6 DB7 high 251 define LPT2 0x278 lpt2 CString MessageControlString CAboutDlg dialog used for App Abo
138. iated gain at minimum noise figure and D The maximum available COIN Gi o 15 Fig 3 8 Typical low frequency noise of a GaN HFET biased in the linear region The slope of the red line is ideally proportional to I f esses 16 Fig 4 1 The microwave setup to characterize DC characteristics S parameters noise and power of transistors the microwave frequency range 17 Fig 4 2 a Typical DC measurement result obtained with the aid of a b parameter analyzer Hewlett Packard HP4145 18 Fig 4 3 The microwave setup configured for S parameter measurements 19 Fig 4 4 a Typical result of a S parameter measurement obtained with the aid of a b vector network analyzer Hewlett Packard 8150 20 Fig 4 5 Picture of two mechanical tuners MT982A401 with a tuner controller J0 Maury 20 Fig Fig Fig Fig Fig Fig Fig Fig Fig Fig Fig Fig Fig Fig Fig Fig Fig Fig Fig TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 133 141 4 6 a Probe tip VHT40A GSG 150P and b probe tips contacting a through onacdlibration sic eae es dee Re d ase Ste 4 7 The representation of the S parameter setup in the ATS software of Maury MTOW Rae Re p ENC SC Reh eter ten
139. ibes the proportion of variance in common between the two variables If we multiply this by 100 we get the percentage of variance in common C 5 Error due to small number of observations The error due to a small number of observations can be calculated 30 by r Seide C 8 with n the number of observations The correlation coefficient can thus be written as C 9 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 68 141 Appendix D WAVER INFORMATION D 1 GaNHFETO1 The high frequency properties of these devices were measured at Raytheon Corporation in a frequency range from 2 GHz to 26 GHz 1 1 Typical material properties Mobility 1000 cm2 Vs Carrier sheet density 1x10 cm 2 Aluminum mole fraction 15 Surface passivation Process MOCVD low pressure 1 2 Layer structure SOURCE WIDE Alp 55 300 A 2DEG interface NARROW GaN 1 5 um E sapphire Fig D 1 Layer structure of GaNHFETOI TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 69 141 D 1 5 Typical DC characteristics A typical output characteristic and typical gate leakage current is shown in Fig D 2 for a transistor 6 3 on waver GaNHFETOI Ip Vc and vs Vp GaNHFETO1 6 3 25 t Gate Source Voltage 0 20 4 Gate Soutce
140. ibration standard connected to the reference plane of the VNA 38 These error coefficients represent the S parameters of the network between the VNA and the reference plane at which the calibration was carried out The network is considered passive and reciprocal Fig F 4 shows an illustration of the so called error adapter calculated from the error coefficients Reference Plane 1 Vector Error terms Network Analyzer Hewlett Packard HP8510C Fig F 4 The error model for a one port calibration with the HP8510C The error adapter can be described by the S parameter matrix As F 1 Two calibrations have to be performed to calculate the S parameters of the fixture One calibration at the coaxial reference plane with a 3 5 mm calibration kit the other with the fixture connected at the previous reference plane with an on waver calibration substrate like demonstrated on the left and right side of Fig F 5 TOBIAS WERNER Reference Plane Vector Network Analyzer Hewlett Packard HP amp 8510C Open short Load 3 5 mm Calibration Reference Plane Vector Error terms Network m Analyzer Hewlett Packard HPEBE510C CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 110 141 Reference Plane j Fixture Vector Network Open um Analyzer Load Packard HP8510C On Waver Calibration Reference Plane Vector Err
141. icated file AUTOR Tobias Werner Universitaet Stuttgart 5 February 29 2000 5 function ReadFrequencyList CalFile exit_counter 1 current String fgetl CalFile while isempty findstr current String LIST exit counter exit counter 1 if exit counter 19 disp ERROR in file format disp EX IT PROGRAM 0 return end current String fgetl CalFile end if strncmp current String VAR LIST BEGIN 14 555555555555 Read the frequencies for VAR LIST text sprintf Reading frequencies n disp text FREQUENCY counter 0 Current Frequency fgetl CalFile while strncmp Current Frequency VAR LIST 12 FREQUENCY counter REQUENCY counter 1 Frequency Array FREQUENCY counter 1 str2num Current Frequency Current Frequency fgetl CalFile if FREQUENCY counter 800 nj disp ERROR The number of frequencies in the file is bigger than the maximum number for the HP8510 801 disp EX IT PROGRA M y 0 return end end y Frequency Array elseif strncmp current String SEG LIST BEGIN 14 5555555555555 Read the frequencies for SEG LIST text sprintf Reading frequencies n disp text Current Frequency fgetl CalFile Right Current Frequency 4 size Current Frequency 2 frequency range str2num Right Frequency
142. icient of data set X and data set Y coefficient of determination goodness of fit covariance of data set X and data set Y average of data set X average of data set Y slope of the regression line intercept of the regression line with the y axis velocity of light bias point 7 HP8510 error coefficient representing the frequency response HP8510 error coefficient representing the directivity HP8510 error coefficient representing the source match S parameter matrix describing the error adapter of the first calibration chain matrix transformed matrix As S parameter matrix describing the fixture chain matrix transformed matrix S parameter matrix describing error adapter of the second calibration chain matrix transformed matrix lt electrical length of a line reference impedance of the system 50 2 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 56 141 Abbreviation Meaning ALFNM Automated Low Frequency Noise Measurement AlGaN Aluminum Gallium Nitride AIN Aluminum Nitride ATS Automated Tuner System Maury Microwaves DUT Device Under Test FET Field Effect Transistor FFT Fast Fourier Transformation GaAs Gallium Arsenide GaN Gallium Nitride GPIB General Purpose Instrument Bus HBT Heterojunction Bipolar Transistor HEMT High Electron Mobility Transistor HFET Heterostructure Field Effect Transistor HJFET Heterojunction Field Effect Transistor JFET Junction
143. icrowave noise bias condition e Sy 100Hz V s0V 0 5 dBVrmsi SI 100Hz V s0V Vo 0 5V dB Hz TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 38 141 Sy represents the measured quantity which is only specified for reference Sy showed negative correlation coefficient of about 0 2 what indicates no significant dependence considering the small number of available devices A different correlation close to zero 1s obtained if the spectral current noise density is normalized to the drain current Sr 2 and additionally the source drain spacing is taken into account It is difficult to judge which parameter is the most suitable to express the flicker noise level however taking into consideration that for the microwave noise no normalization was applied the spectral current noise density S appears to be the most expressive parameter The scatter diagrams for 5 therefore are framed To make it possible to draw own conclusions and to compare the parameters to each other all parameters however were listed The following scatter diagrams will help to get a graphical idea of the dependence and to better judge the bare correlation result 2GHz V 7 1V V5 10V dB Correlation coefficient 0 01 0 20 Correlation coefficient 0 19 0 19 116 00 4 118 00 eu 120 00 122 00 5 124 00
144. ind of noise with a focus on flicker noise and microwave noise Chapter 2 will give a brief overview about the analyzed transistors their characteristics and typical device performance To characterize the noise properties of the transistors in the low frequency and the microwave region two different measurement setups were used and will be presented in Chapter 3 In Chapter 4 the obtained measurement results for the flicker and the microwave noise will be discussed and examined for a possible correlation In Chapter 5 we will compare and summarize the obtained results and draw a conclusion In the Appendix explanations of the used symbols and abbreviations long derivations specific waver information program listings and the references can be found III Vs vol 11 No 1 p 7 Jan Feb 1998 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 6 141 2 NOISE Noise arising in an electronic component or a device is historically classified as thermal noise Shot noise or flicker noise Thermal noise occurs in any conductor due to the random motion of electrons caused by thermal agitation In 1928 Johnson experimentally verified Schottky s theory that the thermal noise voltage depends only on resistance temperature and bandwidth and is independent of material and shape 3 Because of this discovery thermal noise is also called Johnson noise At the same time Nyquist showed that the available
145. ints in Fig 5 12 however is very sensitive to errors because only a few entries represent the microwave noise with elevated noise figures To illustrate this sensitivity the comparison in Fig 5 13 shows that the correlation coefficient changes considerably if only one point is inserted into the data range Fig 5 13 Dependent on the constellation of the points in the scatter diagram the correlation coefficient is very sensitive to errors Only one additional point can change the correlation coefficient from 0 564 a to 0 223 b Unfortunately the analyzed devices were the only correctly working transistors on the waver GaNHFET08 Other transistors on the waver seemed to work but could not be pinched off and were not considered for this examination The results for this waver could not be regarded as reliable due to the presented constellation sensitivity and the low number of analyzed devices This example visualizes that the bare correlation coefficient 15 only qualified expressive and that a look at the point constellation in a scatter diagram provides important information to judge the reliability of the correlation result Correlation coefficient 0 564 Correlation coefficient 0 223 124 4 124 125 4 125 4 ee 126 126 m i 127 4 B 9E ds 5 128 lt 128 1 m 129 4 129 4 2 TT b or 130 4 x 4130 4 131 4 en 134 is 7414 132 132 4 133 4 e
146. ion to an intermediate frequency within the frequency range of the instrument is required 32 a b Ren m H Fig 4 9 a Noise source MT7618E and 5 Noise Gain Analyzer MT2075C TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 23 141 For this purpose a frequency extender MT868C of Maury Microwave is inserted at the output of the system The noise signal is down converted by a modulation with a signal generated by an oscillator HP83752A of Hewlett Packard to a frequency range which 15 suitable for the noise figure meter For the noise measurement the source and load tuner have different responsibilities The impedance setting of the load tuner has no influence on the noise level of the measurement but it can be necessary to adjust its impedance to prevent the setup from oscillating In some cases the system can be even more stable without a tuner at the load position 34 To find the output impedance however at which the transistor is power matched a load tuner must be present Noise Measurement 2 0000 GHz Source Fig 4 10 Example of source and load positions for a noise measurement The light blue crosses in the smith chart indicate the range which can be covered by the mechanical tuners at 2 GHz The red crosses in the source chart are the positions where the noise figure is measured and will be used to determine the noise parameters The red cros
147. ise This is due to the fact that the low frequency noise is up converted to a higher frequency band and can there interfere with the main signal if for example a transistor is applied in an oscillator circuit 2 GaN is a new promising material combination especially for high power applications due to its higher breakdown voltage in comparison to the state of the art technology gallium arsenide GaAs For GaN semiconductors a share of the market of 3 billion dollars was forecasted for the year 2006 The expression heterostructure denotes that a wide band gap semiconductor layer AlGaN is brought together with a narrow band gap semiconductor layer GaN Due to the difference in electron affinities of these two layers a two dimensional electron gas 2DEG 15 formed at the so called heterointerface what results in an improved electron mobility and electron velocity With the result of this investigation important conclusions can be drawn concerning the noise mechanisms in semiconductors This can help to improve noise models and thus to better predict the noise performance of semiconductors The measurement process to determine the minimum noise figure is more complex and thus more time consuming than the characterization of the flicker noise Therefore if dependence between the flicker and the microwave noise could be found quicker noise determination of analyzed devices could be performed In Chapter 1 we will introduce different k
148. ise figure F nin at 2 GHz and Sy 8 Sy and at 100 Hz GaNHFETO4 Fig 5 14 shows again the four scatter diagrams for the bias condition with the lowest averaged minimum noise figure 3 79 dB for waver GANHFET04 The correlation coefficient for all four low frequency noise parameters compared with the minimum noise figure resulted in a negative correlation between 0 24 and 0 27 Even so the points are arbitrarily scattered and do not lie close to the regression line An increased number of transistors on the waver GaNHFET04 would have been desirable but only devices with widely differing properties were left on the waver and were not usable When comparing the correlation coefficient between Fin and the spectral current noise density Sj obtained for GaNHFET04 and GaNHFETIO we can notice that both are negative and show about the same value of 0 2 0 2 No uniform correlation result is obtained when other flicker noise parameters are taken to compare to In all analyzed cases for method J the scatter diagrams showed a cloud of points that were widely scattered from the regression line The correlation coefficient in general was small and no clear tendency for the dependence of the flicker and the microwave noise was found 6 00 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 45 141 5 5 Method B Same bias condition This method was carried out for 21 devices on waver G
149. ization procedure by adding or deleting commands The commands are executed once if the Measure button is pushed PIEI E FNM_Instrumentinit tet Event bar Displays help the status of the program and the time until end of measurement The SR770 can only measure at equidistant frequencies on a linear scale The Averaging ranges You can either select automatic or manual ranges for the averaging of the 58770 In the automatic mode the program will estimate the frequency steps and applied number of averages automatically depending on the selected frequency range The File option is more flexible and you can decide in how many frequency steps the measurement range should be divided and what averaging factor should be used in these ranges The format should be frequency averaging It expresses what averaging should be used below the indicated frequency FHM Averagingh anges File Edit Search Help Measure Stop button start and interrupt the measurement with the this button Autorange If selected the autoranging will only be performed once before measuring Setup file If you push this Edit button the SystemSetup file will open The File contains GPIB commands for the SR770 refer manual chapter 6 to put the instrument in the setup state This state is usually used before starting measurement to control and arrange the setup You can adjust this file by adding
150. k default essageBox Wrong argument in modul ControlOutput return ERROR return SUCCESS UINT CLPTSwitchRemoteDlg GetLPTPortBaseAddress int PortNumber return unsigned int far MK 0 40 6 2 PortNumber return 0 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 132 141 Appendix H INDEXES H 1 Figure index Fig 1 1 Noise distorting a an analog signal b regeneratable digital signal and a non regeneratable digital pectet ttt Perte aen dtd 4 Dios qd UN CNSIONS Oo us un 8 Fig 3 1 Examples of modulation doped layers for HFETS 222 0 0000000000000 11 Fig 3 2 Examples of GaN with T shaped 11 Fig 3 3 Typical current voltage characteristic of a GaN HFETs comparison to D OAS PIE 12 Fig 3 4 S parameters of a GaN HFET versus frequency at a certain bias condition Typically the S21 is smaller than unity at a reference impedance of 50 UBI ai 13 Fig 3 5 Frequency dependence of the minimum noise figure and the associated cain of GAN at room TemperalWfe eee e 13 Fig 3 6 Bias dependence of the minimum noise figure and 5 normalized equivalent noise resistance GAN HFETS sss 14 Fig 3 7 Bias dependence of a the assoc
151. larations immediately befor INCLUDED the previous line endif defined AFX FLICKERNOISEMEASUREMENT FF726924 2 4 11D4 B354 0006295283FA INCLUDED 4 2 Header file FlickerNoiseMeasurementDlg h FlickerNoiseMeasurementDlg h header file if MSC VER gt 1000 pragma once endif MSC VER gt 1000 define DATA TO DIALOG FALS define DIALOG TO DATA TRUE const WM DISPLAY STATUS WM USER 100 benutzerdefinierte Nachricht const WM SAVE DATA WM USER 101 benutzerdefinierte Nachricht const WM MEASUREMENT STOPED WM USER 102 benutzerdefinierte Nachricht const WM START COUNTDOWN WM USER 103 benutzerdefinierte Nachricht UINT MeasureThread LPVOID pParam Thread is globally defined CFlickerNoiseMeasurementDlg dialog class CAveragingRangeType public double Frequency int Averaging class CFlickerNoiseMeasurementDlg public CDialog Construction public if defined AFX FLICKERNOISEMEASUREMENTDLG FF726926 2 4 11D4 B354 0006295283FA INCLUD define AFX FLICKERNOISEMEASUREMENTDIG FF726926 2 4 11D4 B354 0006295283FA INCLUDED ED PAGE 86 141 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 87 141 int CalculateMeasurementTim
152. ld The FET is a three terminal device in which the current through two terminals source and drain 1s controlled by a voltage at the third terminal gate It is a majority carrier device and is therefore often called a unipolar transistor The FET comes in several forms In a Junction Field Effect Transistor JFET the control gate voltage varies the depletion width of a reverse biased p n junction A similar device results if the junction is replaced by a Schottky barrier MEtal Semiconductor Field Effect Transistor MESFET Alternatively the metal gate electrode may be separated from the semiconductor by an insulator Metal Insulator Semiconductor Field Effect Transistor MISFET The Metal Oxide Semiconductor Field Effect Transistor MOSFET is a common special case of this type and uses an oxide layer as the insulator Since the MESFET is compatible with the use of III V compounds it is possible to exploit the band gap engineering available with heterojunctions in these materials In order to maintain high transconductance in a MESFET the channel conductivity must be as high as possible Increasing the doping in the channel and thus the carrier concentration increases the conductivity However increased doping also causes increased scattering by ionized impurities which leads to a degradation of mobility What is needed 1s a way of creating a high electron concentration in the channel of a MESFET by some other means than doping An approach t
153. led and put into service The comparison of the minimum noise figure level with the spectral current noise density 9 which was considered to be the most which expresses the microwave noise expressive parameter to refer to the flicker noise level allows to draw the conclusion that the flicker and the microwave noise are not correlated Method A Method B Method C GaNHFETIO GaNHFETIO GaNHFETO01 2 GHz Correlation coefficient 0 19 18 Correlation coefficient D 3 021 Correlation coefficient 0 06 0 19 130 00 A32 00 4 134 00 5 125 00 138 00 140 00 3 142 00 144 00 3 146 00 148 00 150 00 152 00 154 00 176 00 156 00 0 50 100 1 50 200 0 50 1009 5 20 260 200 3 20 151 00 V D VIVI 158 00 4 161 00 4 168 00 5S4100Hz 2 171 00 S4 100Hz Vice Vie 83 1 dBArmshs Hz SA 100Hz v z2 1 09V Vos dBiA mss Hz F Ves dB Fel GHE DV Vos Inv dB F 2GHz 48 GaNHFETO04 GaNHFETIO worst case 01 26 GHz Correlabon coefficient D 24 0 23 Correlabon coefficient D 85 0 15 Correlation coefficient 0 00 0 18 134 00 148 00 135 00 4 133 00 140 00 142 00 4 444 00 14656 00 1 00 150 00 152 00 154 00 4 Vp 145 00 A50 00 151 00 152 00 4183 00 4 Ri 154 d 158 00 4 458 00 4 455 00 00 162 00 i
154. llowing calibration files were found Mn if known sets has been chosen small enough Otherwise the program will calculate a incorrect value n Type in the electrical length in m or press to continues 5 55 55 5 5 5 Read the frequencies of both files and compare them 5 disp disp PROCCESSING DATA disp 7 text sprintf Proccessing error correction file WITHOUT FIXTURE s file namel disp text FrequencyArrayl ReadFrequencyList CalFilel if FrequencyArrayl 0 return end NumberOfFrequenciesl size FrequencyArrayl 1 text sprintf Proccessing error correction file WITH FIXTURE s file name2 disp text FrequencyArray2 ReadFrequencyList CalFile2 if FrequencyArray2 return end NumberOfFrequencies2 siz FrequencyArray2 1 Control if frequency ranges are th if FrequencyArrayl FrequencyArray2 sam Sur or press ENT ENTE ER to let the program calculate it that the frequency spacing of the two calibration R TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS disp You have to use the same frequency spacing for the two calibration files disp EX IT PROGRA M return end 555555555555555555 Read th rror terms from the two files 5 5 5 55 55
155. ly when the application s main window is not a dialog SetIcon m_hIcon TRUE Set big icon SetIcon m_hIcon FALSE Set small icon UINT PortBaseAdress PortBaseAdress GetLPTPortBaseAddress 1 m SetupConfig ReadPortStatus UpdateData false TODO Add extra initialization here return TRUE return TRUE unless you set the focus to a control void CLPTSwitchRemoteDlg OnSysCommand UINT nID LPARAM lParam if nID amp OxFFFO IDM ABOUTBOX CAboutDlg dlgAbout dlgAbout DoModal else CDialog OnSysCommand nID lParam T you add a minimize button to your dialog you will need the code below to draw the icon For MFC applications using the document view model this is automatically done for you by the framework void CLPTSwitchRemoteDlg OnPaint if IsIconic CPaintDC dc this device context for painting SendMessage WM ICONERASEBKGND WPARAM dc GetSafeHdc 0 Center icon in client rectangle int cxIcon GetSystemMetrics SM CXICON int cyIcon GetSystemMetrics SM CYICON CRect rect GetClientRect amp rect int x rect Width cxIcon 1 2 int y rect Height cyIcon 1 2 Draw the icon dc DrawIcon x y else CDialog OnPaint The system calls this to obtain the cursor to display while the user drags the minimized window HC
156. m Calculate Fixture Tobias Werner February 2000 1 fprintf SParamFilel Fixture S Parameters n fprintf SParamFilel Date time s n datestr now fprintf SParamFilel GHZ S 50 n fprintf SParamFilel Scattering Parameters n fprintf SParamFilel 1 5f 1 5f 8 6g 1 5f 8 6g 1 5f 8 6g 1 5f 8 6g n Fixture S fclose SParamFilel y 0 return 5 5555555555 FUNCTION CorrectPhase Fixture S MA 2 Electrical Length INPUT Fixture S MA 1 Frequency Fixture S MA 2 uncorr phase of 812 OUTPUT Corrects the phase of S 21 AUTOR Tobias Werner Universitaet Stuttgart DATE March 02 2000 function CorrectPhase Fixture 5 S21 Electrical Length text sprintf Applying phase correction to 512 and 521 disp text Velocity of light 3E8 m s if isempty Electrical Length Electrical Length CalculateElectricalLength Fixture 5 521 text sprintf The electrical length of the fixture was calculated by the program 2 6f m n Electrical Length else text sprintf The electrical length of the fixture was entered by the user 2 6f m n Electrical Length end disp text Ideal phase RangePhase 2 180 Fixture S A S21 1 1E9 Electrical Length Velocity of light Number of phases size Ideal phase 1 for Current Phase INDEX 1 Number of p
157. method but not only for one bias point and several transistors but covering the entire bias range of one transistor Fig 5 7 5 4 Method A Bias condition for lowest noise Method Al bias condition for lowest noise was carried out for 26 transistors on waver GaNHFET10 15 transistors on GaNHFETOS and 18 transistors on GaNHFETOA as depicted in Fig 5 8 bias condition for lowest noise 2 GHz 100 Hz GaNHFET04 GaNHFET08 microwave noise flicker noise GaNHFET10 Fig 5 8 Compared parameters and number of devices on wavers for method A linear region saturation TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 37 141 5 4 1 Results for GANHFETIO The minimum noise figure was measured at 2 GHz in saturation at an optimum bias n condition for lowest microwave noise and was compared to Sy Sj 91 1 2 and at 100 Hz in the linear region V5 0 V 0 5 V To ensure reliable results and to better detect measurement errors the microwave noise was measured at four different working points Although the flicker noise measurement is sensitive to the contact quality between the probe tips and the transistor pads the measurement process is less complex and one measurement is sufficient to ensure a consistent measurement result 1 f noise microwave noise VGz 1V 1 VGz 15V 0 1 5 vD 8V 10 20 2 Fminf2GH2Z Fma444260H2
158. mponent of this type of noise and used to be known as semiconductor noise since the noise in early semiconductor devices was dominated by this type 6 Numerous scientists have investigated the phenomenon of flicker noise in semiconductor devices but the details of the mechanisms involved are still not entirely understood There 15 no consensus to the origin of the 1 f noise and it is very likely that there exist more than one mechanism giving rise to the same noise characteristics TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 7 141 It appears that 1 f noise can be a surface effect and a bulk effect and the physical origins of the noise presumably being different in the two cases According to MeWhorter s number fluctuation theory 7 1 f noise is attributed to the trapping and detrapping processes of the charges in the oxide traps Hooge s empirical model 8 considers the 1 f noise as a result of carrier mobility fluctuation due to lattice scattering It has been reported that both the carrier number fluctuation and the mobility fluctuation are possible mechanisms which lead to the 1 f noise The 1 f noise in HFETs was at first rather large extending to 1 GHz It varied as at low drain currents The 1 f noise seems to be generated by traps the space charge region under the gate The trapping and detrapping of electrons gives rise to a 1 f modulation of the channel conductance and this
159. mum complex source reflection coefficient complex source reflection coefficient minimum noise figure noise figure total noise figure noise figure of the DUT associated gain at minimum noise figure maximum available gain MAG available gain of the DUT gate source voltage drain source voltage drain source breakdown voltage measured voltage at the drain of the DUT supply voltage voltage drop over the resistor battery voltage drain source current gate current saturated drain source current variation of the drain current equivalent noise resistance normalized equivalent noise resistance resistance of the resistor at the drain of the DUT mean resistance of the DUT Q IOS ND ua b m 8 E 3 x lt 996035045 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 55 141 Unit O O V lHz A IHz cm Hz m s O Description resulting bias network resistance differential resistance of the DUT differential resistance for bias point 7 variation of the DUT s differential resistance DC resistance of the of the DUT resistance of the potentiometer at the drain of the DUT resistance of the potentiometer at the gate of the DUT spectral voltage noise density spectral current noise density total number of charge carriers electron sheet density frequency gate width gate length data set data set correlation coeff
160. n Microsoft Visual C 6 0 and communicates with the instrument over the GPIB It is designed as a multithreading application and the measurement can be as a result interrupted at any time This is important due to the fact that measurements can take a long time and there should be a possibility to stop and restart it if a wrong setting or a sudden perturbation is discovered Very convenient is that the time until the predicted end of the measurement is displayed in the status bar of the program Especially if long measurements are taken this makes it easy to leave the setup unattended and to come back right when the measurement has finished All parameters are examined for wrong settings and items in the interface are enabled or disabled depending on the interaction with the user to make wrong program handling impossible A main rule was that the program had to be intuitive Therefore the user interface was designed with great and be seen in Fig E 3 Automated Flicker Noise Measurement 9Tobias Wemer 2000 Voltages WG 1 Wwo sv WRe ay Resistors Pi 10 kOhm A 10 kOhm P2 10 Frequency range Stark 1 Hz Stop J 100000 Hz Averaging ranges Automatic ranges v File initial autorange GPIB Setup Time out 10s Addr 10 Find Comand files Initialization file Edit Setup system file Ed Setup system Data file Start sample for lag c
161. n header file for the FLICKERNOISEMEASUREMENT application if MSC VER gt 1000 pragma once endif MSC VER gt 1000 ifndef _AFXWIN H error include stdafx h before including this file for PCH endif include resource h main symbols include Decl 32 h CFlickerNoiseMeasurementApp S FlickerNoiseMeasurement cpp for the implementation of this class class CFlickerNoiseMeasurementApp public CWinApp public CFlickerNoiseMeasurementApp Overrides ClassWizard generated virtual function overrides AFX_VIRTUAL CFlickerNoiseMeasurementApp public virtual BOOL InitInstance AFX VIRTUAL Implementation AFX MSG CFlickerNoiseMeasurementApp NOTE the ClassWizard will add and remove member functions here DO NOT EDIT what you see in these blocks of generated code AFX MSG DECLARE ESSAGE AFX INSERT LOCATION if defined AFX FLICKERNOISEMEASUREMENT H FF726924 2 4 11D4 B354 0006295283FA _ define AFX FLICKERNOISEMEASUREMENT FF726924 2 4 11D4 B354 0006295283FA INCLUDED Microsoft Visual C will insert additional dec
162. n voltage Vp V 10 12 Drain voltage Vp V 10 12 D Ni 7 PES DA NR a Relative Noise Resistance r Fig 3 6 Bias dependence a the minimum noise figure min and b the normalized equivalent noise resistance of GaN HFETs As be seen decreases for increasing drain source voltages and decreasing gate source voltages The tendencies for approaching the pinch off voltage and approaching the drain source breakdown voltage 15 not shown in the picture In the first case increases sharply after passing a minimum In the second case also a minimum can be found and than rises again slowly The optimum bias condition for was typically found at about a third of the saturated drain source current and about half of the drain source breakdown voltage Vps expressed as equations Ip 1 3 ps and 3 1 1 2 V psg 3 2 The normalized equivalent noise resistance 7 was found to be the most sensitive parameter of the four noise parameters It showed large variations when including or excluding impedance points from the least mean square algorithm 7 showed no clear tendency dependent on the bias condition Fig 3 65 The bias dependence of two kinds of gains is shown in Fig 3 7 The associated gain at minimum noise figure or the maximum available gain Gmax or MAG occurs if the input of the device 15 noise or power matched resp
163. noise level was much higher than the noise floor and only some spikes which can easily be separated from the noise contribution of the transistor interfered with the measurement Bias Unit Digital Multimeter Hewlett Packard 1 1 GSD d 1 1 20001 D 10001 4 5 Parameter Analyzer 30001 G Hewlett Packard 55 9 GPIB j S Control Program I V Tobias Werner Automated Low FFT Spectrum Analyzer Noise Measurement GSD 5 3 Stanford Research Systems Tobias Werner 8770 Automated IV Curve I 31 Measurement 1 I 1 1 1 1 1 I Fig 4 16 Schematic view of the low frequency noise setup Stanford Research Systems 1290 D Reamwood Avenue Sunnyvale 94089 USA Phone 1 408 744 9040 Fax 1 408 744 9049 ww w stanfordresearch com TOBIAS WERNER K 57 CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 27 141 As can be seen from the schematic view of the low frequency noise setup in Fig 4 16 the input of the spectrum analyzer is connected to the drain of a biased transistor The gate source voltage and the drain source voltage Vp can be adjusted by potentiometers 100 and P 2 which are connected to bat
164. nt SR770 measurement interrupted SR770 measurement completed SR770 setup ready SR770 transferring data general information general question 84 Fig F 1 Example of an insertable system component with APC3 5 female connectors at the input and APC3 5 male connectors at the output 107 Fig F 2 Example of a non insertable system component with APC3 5 female at INE INDULGE ig pO e 108 Fig F 3 Example of a device with an APC3 5 male connector on one side and a connector less end the other SIC sic tta oe bert Lon lie ce 108 Fig F 4 The error model for a one port calibration with the 851 109 Fig F 5 The error terms of two one port calibrations can be used calculate the S parameters or the oodd otia iul 110 Fig F 6 Uncorrected phase of 812 and S21 The dots represent the discrete measured values and the line is the linear interpolation between them 113 Io d Corrected Phase Of STZ QW E Ea 114 Fig F 6 The slope of the line segment can be easily calculated in the left example but will lead to a wrong result for the electrical length in the right ODIT 114 Fig F 9 Fig G I Fig G 2 Fig G 3 Fig G 4 Fig G 5 Fig G 6 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 136 141 Comparison b
165. o interfaces in InP In 5Ga55As InP HFETs by means of digital signal processing of measured low frequency noise spectra Conference Proceedings Sixth International Conference on Indium Phosphide and Related Materials Cat No 94CH3369 6 p 656 415 18 1994 Buckingham M J Noise in electronic devices and systems Ellis Horwood Limited 1983 A Balandin S Morozov Wijeratne S J Cai Li J L Wang Viswanathan and Yu Dubrovskii Effect of channel doping on the low frequency noise in GaN AIGaN heterostructure field effect transistors Applied Physics Letters vol 75 no 14 p 2064 6 4 Oct 1999 N Pala R Gaska S Rumyantsev M S Shur M Asif Kahn X Hu G Simin and J Yang Low frequency noise in AlGaN GaN MOS HFETS Electronic Letters vol 36 no 3 p 268 70 3 Feb 2000 Fukui H Available Power Gain Noise Figure and Noise Measure of Two Ports and Their Graphical Representation IEEE Trans on CT vol CT 13 No 2 pp 137 142 14 15 16 17 ma m oo 23 24 25 26 27 28 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 140 141 H T Friis Noise Figures of Radio Receivers Proceedings of the IRE Vol 32 pp 419422 July 1944 N X Nguyen M Micovic W S Wong P Hashimoto P Janke D Harvey and C Nguyen Robust low microwave noise GaN MODFETs with 0 60 dB noise figure at
166. o this requirement is to grow a thin undoped well e g GaN bounded by wider band gap doped barriers e g AlGaN This configuration called modulation doping results in conductive GaN when electrons from the doped AlGaN barrier fall into the well and become trapped there Since the donors are in the AlGaN rather than the GaN there is no impurity scattering of electrons in the well This device is called a MOdulation Doped Field Effect Transistor MODFET or High Electron Mobility Transistor HEMT It 1 also called Heterostructure Field Effect Transistor HFET Two Dimensional Electron Gas Field Effect Transistor TEGFET or Selectively Doped Heterostructure Field Effect Transistor SDHFET When two different semiconductors layers are joined together the atoms at the interface have to form chemical bonds Due to a difference in the electron affinity of these two layers a two dimensional gas 2DEG is formed at the so called heterointerface In HFET a wide band gap semiconductor layer separates the gate from the channel Fig 3 1 At the heterointerface between the wide and narrow band gap semiconductor layers a 2DEG is formed and carries the drain to source current TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 11 141 WIDE 15Gaq g5N 300 WIDE Alo 4Gag 600 2DEG hetero interface 2DEG doped GaN 6A hetero interface NARROW GaN 1 5um NARROW GaN 0 5 um sapphire in
167. of the correlation coefficient results from the comparison between and the spectral current noise density S but a correlation coefficient of about 0 2 still does not indicate a significant dependence TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 40 141 In summary it may be concluded for waver GaNHFETI0 that correlation coefficient is quite small and the points show no clear tendency This indicates that there appears to be no dependence between the low frequency noise and the microwave noise for this waver 5 4 2 Results GANHFETOS The minimum noise figure for this waver was measured at 2 GHz for six different bias points in the saturation region and was compared as before to Sy Sr 91 1 and a at 100 Hz for one bias point in the linear region 0 V Vp 0 5 V 1 f noise microwave noise 1 1 5 1 2 0 2 0 2 VD 2 VD 4 VD 6 VD 2 VD 4 lara Device dB dB dB dB dB dB ADS 6 36 6 30 6 49 6 67 6 65 6 52 A12 2 58 2 10 223 1 63 1 66 1 75 A15 3 07 2 07 2 12 1 65 1 56 1 58 B01 4 10 3 38 3 43 2 55 2 43 2 45 802 2 49 1 65 1 57 1 57 1 41 1 42 212 2 30 1 71 1 85 1 21 1 20 1 24 Cis 2 59 1 80 1 87 1 39 1 36 1 39 005 3 05 2 04 2 07 1 51 1 47 1 47 010 3 10 2 05 2 08 1 51 1 45 1 48 017 5 44 4 41 4 25 5 16 4 96 4 40 202 3 97 2 10 2 11 1 48
168. of two data ranges in our case the parameter representing the level of the flicker noise and the microwave noise is not directly perceptible a mathematical approach can help to identify the degree of reliance between the two ranges The correlation coefficient is such a mathematical aid and a short overview is given in Appendix C Correlation To investigate a possible correlation between the low frequency 1 f noise and the microwave noise three different approaches A B and C were chosen as illustrated in Fig 5 7 This current occurs when the gate to source voltage is held to zero and the drain to source voltage Vps is set to a specified value usually about 3 V TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 36 141 same bias point i 2 I 4 fo fo o Uu microwave noise microwave noise flicker noise flicker noise Fig 5 7 Different approaches to analyze the correlation between the flicker and the microwave noise Method compares the noise dependence at the bias condition for lowest noise This means in the linear region for the low frequency noise and in saturation for the microwave noise Fig 5 7A Method compares the flicker and microwave noise for an identical bias condition in the saturation what is the common working point for transistors Fig 5 7B Method examines the dependence also for a similar bias condition like in
169. one in the meantime The data is directly transferred to the controlling computer No floppy disc has to be used The individual sections are combined automatically and the program eliminates overlapping sections Combination errors of the numerous files are prevented Eliminating unnecessary redundant samples and restricting the data to logarithmic equidistant steps can reduce the number of measurement points Thus file size is decreased while the qualitative meaning is preserved Wrong parameter settings and with it wrong measurement results can be prevented because the program takes care of setting all parameters to the correct values The measurement conditions and parameter settings are written directly into the file for further reference and no additional notes are needed The measurement situation can therefore be reconstructed The measurement file is named with the corresponding parameter settings and is not restricted to 8 characters This helps to keep the work organized and to locate the files quicker The measurement takes less time 1s more flexible TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 30 141 Automated Low Frequency Hoise Measurement Werner 2000 MA Voltages WG 4 v Resistors E 10 k hm R k hm RAS Frequency range Start 1 Hz Stop 100000 Hz Averaging ranges Automatic ranges
170. onversion 20 Average GPIB card was found on this computer Fig E 3 The user interface of the automated low frequency noise measurement program TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 84 141 The interface of the dialog based program 1 divided into the categories Voltages Resistors Frequency range Averaging ranges GPIB Setup Command files and Data file A schematic diagram of the measurement setup shows where the corresponding voltages are measured and resistors are located The measurement range can be defined with a start and stop frequency and will be split up into several sections depending on the setting in the category Averaging ranges An automatic or a file based solution can be chosen and will decide how many sections will be used and what averaging will be applied in these sections In the next category the GPIB time out and GPIB address can be adjusted To set up the system easier a scan function has been implemented which scans the GPIB bus for the name string of the SR770 and changes the address accordingly if the instrument was found The initialization and setup procedure of the analyzer is not fixed and can be altered in a file To reduce the overhead of redundant information an algorithm was implemented to convert data points from an equidistant linear to an equidistant logarithmic scale The start sample for this conversion and if the values
171. or a altered gate source voltage 49 Fig 5 18 Scatter diagrams for a comparison between at 2 GHz and Sy and 5 non logarithmic scale at 100 Hz for an altered gate source voltage 49 Fig 5 19 Compared parameters and number of devices on wavers for method C 50 Fig 5 20 Scatter diagrams for a comparison between the minimum noise figure F min at a b 2 GHz and c d at 26 GHz and Sy and S at 100 Hz sess 22 Fig B 1 Simplified circuit to derive the relation between 5 and suus 97 Fig B 2 Differential resistance in the linear and saturation region suus 59 Fig Derivation of all parameters by the measured voltages and known resistors 60 Fig B 4 Calculation of the differential resistance r in the bias point P 2 2 2 2200040 62 dido koc x daten dido adea 64 Fig C 2 The least squares line minimizes the squared distances between the line and TIO DOM 65 Fig C 3 Different correlation results dependent on the chosen range a Entire quc ACH 66 Fig C 4 Comparison of goodness of fit for two regression lines with the same slope 66 pis Di Layerstructure of GaANHAFE TO ecce t ete t p ptas iet rte 68 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 135 141 Iug D2 Typical lV Curve
172. or by which it can be easily mounted inside a standard PC enclosure G 2 2 Layout The layout and schematic was developed with an educational version of Eagle V3 55r3 of CadSoft The layout from the component view of the microwave switch control and the placement of the components can be seen in Fig G 4 50 12U 02 O14 O2 T92 pets L1 1 em Lo Va Ll 3 mj 1 74 245 Fig G 4 The layout left and the component placement right of the microwave switch control TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 126 141 In the left most part of the circuit board the two connectors used to pass and access the parallel port signal In the middle the transceiver circuit above it the block capacitor and underneath to save space the three pull down resistors In the upper right area the two connectors for the power supply and on the right most side the female 9 pin sub D connector G 2 3 Printed circuit board Fig G 5 illustrates the technical realization of the printed circuit board from the component and the solder side Fig G 5 Top and bottom view of the printed circuit board to control the microwave switches G 3 Software The program to control the hardware via the parallel port of the computer was written in Microsoft Visual C 6 0 Fig G 6 shows the user interface of the program The microwave measurement setup can be
173. or deleting commands The commands are executed if the Setup system button is pressed F FNM SystemSetup txt Mi gt File Edit Search Help data is normally displayed with a logarithmic frequency axis To reduce the overhead of information in the upper frequency range the measurement values are converted to an equidistant logarithmic frequency scale beginning atthe indicated sample position The start sample positions 2 5 10 20 30 50 100 200 300 and 400 can be selected The two pictures beside show examples for the third sample as start sample If averaging is selected the algorithm to reduce the number of measurement points calculates the mean of all samples until the last taken point and adjusts the frequency to the one in the middle between the last and the current taken frequency The right picture shows the original blue and the converted sequence red with averaging If no averaging is selected only the value at the taken frequency is chosen and the other values are discarded The left picture shows the original blue and the converted sequence red without averaging 2000 Tobias Werner TOBIAS WERNER E 4 Program Code CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS The main source code can be found in the following subchapters 4 1 Header file FlickerNoiseMeasurement h FlickerNoiseMeasurement h mai
174. or due to a small number of observations which is introduced in Appendix C 5 Error due to small number of observations For all analyzed wavers the error was in the order of about 40 2 100 Hz are plotted versus the minimum noise figure For a significant linear correlation the points in the diagrams should lie close to the regression line what is not the case for this waver The dots rather show a wide distributed pattern covering the entire plot area The normalization of 5 to the drain current leads to a slight displacement of the points Fig 5 9b versus Fig 5 9c This indicates that the correlation of with does not in only reflect the influence of a different source drain spacing but also the effect of the normalization to the drain current which alters the correlation result The different point constellation between Fig 5 9c and Fig 5 9d is mainly due to the changed non logarithmic scale used for the Hooge parameter A comparison if also the Hooge parameter is displayed in dB is shown in Fig 5 10 and demonstrates a better agreement between the points Still there is a deviation what can be ascribed to the different source drain spacing of the transistors Correlation coefficient 0 04 0 20 Correlation coefficient 0 16 0 19 Fig 5 10 Comparison between the point constellation of S F filled dots solid line and the Hooge parameter in dB non filled dots dashed line The highest value
175. or terms Network m Analyzer Hewlett Packard Fig F 5 The error terms of two one port calibrations can be used to calculate the S parameters of the fixture For the first calibration we get a set of coefficients expressed by the matrix 2 The second set of error coefficients 1s expressed by the matrix 22 a Cc and consists of the same matrix Ag as in F 2 in chain with a matrix expressing the S parameters of the fixture To calculate the matrix we transform the two S parameter matrices and into chain matrices and C by the conversion 11 X 22 7 512 X Xoo F 4 X T21 252 ui s21 l 22 Xy TOBIAS WERNER 1 and 2 2 52 Ep E En Ed gt ER ER For the chain matrix can be written Cr Or zac With F 5 and F 6 in F 8 we get Eg2 Ep Es Epi Es B Ego Eg2 Er Egi En After converting F 9 back to S parameters by the formula ASSAT UM and taking into account that 51 equals S5 for the fixture we get CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS Es EpotER Ep PAGE 111 141 F 5 F 6 F 7 F 8 9 10 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND
176. orintf Handle GERMANY n fprintf Handle n forintf 1 fprintf Handle Frequency list n fprintf Handle 100 50 n fprintf Handle 3000 200 n fprintf Handle 100000 500 return TRUE TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS bool CFlickerNoiseMeasurementDlg CreateSystemSetupStandardFile FILE Handle fprintf Handle SPAN14 t span 3 125kHz n fprintf Handle STRF100Nt start frequency 100Hz n fprintf Handle OVLPO t overlapping 0 pecent in fprintf Handle NAVG200 t number of averages 200 5n fprintf Handle AVGOlNt averaging on n fprintf Handle ARNG1 t autoranging on n fprintf Handle ARNGO t autoranging off n fprintf Handle AOFF t perform offset calibration n fprintf Handle AUTSO t auto scale Last command no carriage return return TRUE bool CFlickerNoiseMeasurementDlg CreateInitInstrumentStandardFile FILE Handle fprintf Handle RST t reset the instrument n fprintf Handle STOP t stop measurement n fprintf Handle FMTS1 t display dual n fprintf Handle DISPO 0 t display log mag n fprintf Handle XAXSO 1 t x axis log n fprintf Handle MEASO 1 t measurement PSD n fprintf Handle UNITO 3 t dBVrms
177. orrelation results for an identical bias conditions in saturation GaNHFETIO Table 5 4 shows the individual low frequency and microwave noise results for each device on the waver GaNHFET 10 The correlation coefficient shows a large variation dependent on the applied flicker noise parameter if either Sy S Sy I 2 a are used In saturation the flicker noise is no longer proportional to a and the electron distribution in the channel gets inhomogeneous Therefore the two flicker noise parameters Sy and have no significant meaning and are only presented in the table for completeness y 100Hz Vo 1 0V Vo 10V dBVrmsiyHz TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 47 141 Correlation coefficient 0 24 0 21 Correlation coefficient 0 32 0 21 80 00 4 82 00 4 3 8400 4 4 86 00 88 00 4 gt 90 00 4 T 92 00 gt 94 00 4 e 96 00 gt 98 00 4 wi 100 00 4 n S 102 00 i 0 50 1 00 1 50 2 00 2 50 1 00 1 50 2 00 Fminl 2GHz V 7 1 0V 1 dB GHZ V 7 1 0 V dB Fig 5 16 Scatter diagrams for a comparison between the minimum noise figure F nin at 2 GHz and Sy and S at 100 Hz Fig 5 16 shows the scatter diagrams for a bias condition Vg 1 V and Vp 10 V of waver GaNHFETIO Only the graphs comparing Sy and 5 with F are of importance and show
178. pensate the loss phase error and frequency response The calibration can only remove systematic errors of the system For the on waver calibration a substrate CS 5 Industries with a full two port LRM line reflect match technique was utilized For the reflect the short standards were used As recommended by a Maury engineer the averaging factor of the VNA was 128 for the reflect match 50 Ohm load and line through standard and 1024 for the isolation calibration with the load If S parameter measurements of active devices are made it is important to adjust the output power of the VNA to 0 dBm because otherwise the device will be additionally biased by the VNA and not only by the bias control as intended The S parameter setup like it is depicted in the Maury software can be seen in Fig 4 7 Fig 4 7 The representation of the S parameter setup in the ATS software of Maury Microwave Also called TRM Through Reflect Match TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 22 141 4 1 2 Noise configuration The core of the system formed by the bias control the bias Ts both tuners and the two probes remains unchanged and is assembled like described in the previous chapter A detailed view of the system configured for noise measurement can be seen in Fig 4 8 Noise Setup Control Program Maury Microwave T993C Tobias Werner Automated IV Curve Measurement
179. put If the S parameters of a non insertable system component has to be measured the adapter removal technique 35 can be used Fig F 2 shows an example of a non insertable system component TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 108 141 Characterisation Fig F 2 Example of a non insertable system component with APC3 5 female connectors at the input and output If even the connector family is different for a device there is still a way to measure the S parameters as described in 35 Fixture Characterisation Fig F 3 Example of a device with an APC3 5 male connector on one side and a connector less end on the other side In our case we had the problem to characterize the high frequency properties of a fixture consisting of a low loss cable and a microwave probe tip shown in Fig F 3 The following chapter describes how this problem was solved by performing two one port calibrations and applying the two sets of corresponding error terms of the HP8510C to calculate the S parameters of the fixture TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 109 141 F 1 Systematic approach If a one port calibration is performed the error coefficients representing the frequency response the directivity and the source match are calculated from solving an equation for an open short and load cal
180. r network analyzers VNA are used for this purpose and measure the S parameters which describe these properties To get accurate measurement results for the S parameters the VNA has to be calibrated A measurement calibration is a process which mathematically derives the error model for the VNA This error model is an array of vector coefficients used to establish a fixed reference plane of zero phase shift zero magnitude and known impedance The array coefficients are computed by measuring a set of known devices connected at a fixed point and solving the vector difference between the modeled and the measured response These known devices are called calibration standards and they have a precisely known magnitude and phase response as a function of frequency In general a microwave measurement setup consists of devices with the same connector family and insertable connections i e a female contact on one side and a male contact on the other side In this case it is very easy to measure the S parameters of this device because it is the configuration intended for a VNA measurement process Fig F shows an example of an insertable system component Tuner Characterisation gt Bias 1j source Tuner 115 im oW M TN Maury Microwaves e MT982A01 3 Fig F 1 Example of an insertable system component with APC3 5 female connectors at the input and APC3 5 male connectors at the out
181. rOfAveragingRanges AveragingRangeINDEX switch ControlFrequencyRange Start Stop m AveragingRanges AveragingRangeINDEX Frequency case 0 start and stop are below the frequency limit FrequencySpan 100000 pow 2 19 GetFrequencySpan Start Stop return EntireTime case 1 only start is below the frequency limit Start m AveragingRanges AveragingRangeINDEX Frequency FrequencySpan 100000 pow 2 19 GetFrequencySpan Start m AveragingRanges AveragingRangeIND SettlingTime int 400 FrequencySpan asurementTime int 400 FrequencySpan m AveragingRanges AveragingRangeINDEX Averaging EntireTime SettlingTime MeasurementTime TransferTime OverheadTime SettlingTime int 400 FrequencySpan asurementTime int 400 FrequencySpan m AveragingRanges AveragingRangeINDEX Averaging EntireTime SettlingTime MeasurementTime TransferTime OverheadTime break default start and stop are above the frequency limit break do nothing return EntireTime Frequency PAGE 106 141 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 107 141 Appendix F FIXTURE CHARACTERIZATION In a microwave noise measurement setup all elements of the system have to be characterized concerning their high frequency properties in order to get accurate results Vecto
182. rating in the linear region it can be considered as a voltage controlled resistor r If we consider the voltage and the value of the resistor R as constant a variation Or of the resistor r will lead to a variation 07 due to Ohm s law 0 V or Fig B 1 Simplified circuit to derive the relation between S and Sy The resistor network in Fig 1 shows a voltage divider formed by and V meas can be expressed by TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS If we calculate the derivative of B 1 with respect to r we get Onus Vo pu Or tr R res r Res t r Res Solved for or and divided by equation B 2 can be written as OV meas Res T V Ries If we insert in B 3 we get or OV ets Rove us meas res For the short circuit current fluctuation we can write or 2 or OV meas r I r y R meas res If we solve B 5 for with y meas R a l res square both sides of the equation and divide by Af we get E uu tr 7 Af Af res The expression in B 7 can be written as r PAGE 58 141 B 1 B 2 B 3 B 4 B 5 B 6 B 7 B 8 with S the spectral current noise density and Sy the spectral voltage noise density 1 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOIS
183. red at a variety of source impedances In principle since four scalar variables 1s complex are to be found only four measurements are required In practice however more points are measured and a least mean square technique 1s applied to extract the parameters from the over determined data Noise characterization also requires the parameters of the device under test DUT to be separated from the parameters of the measurement system to which the DUT is connected To do this the system must be calibrated to compensate the system contribution The noise contribution of the system often called the second stage since it follows the DUT will vary with its source impedance according to 2 4 Therefore the complete noise and gain parameters of the system must be known to determine the system noise contribution when a particular DUT is connected This consists of measuring the total noise figure with several different source impedances The noise figure of the DUT for each position is then given from the Friis s cascade equation 14 in 2 5 Fo 1 Faye 9 2 ut tota 2 5 is the available gain of the DUT and is resolved from the DUT s S parameters and the source reflection coefficient TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 10 141 3 ANALYZED TRANSISTORS The first Field Effect Transistor FET was already proposed in 1930 by Lilienfe
184. rentSeriesValueINDEX 0 CurrentSeriesValueINDEX lt 400 CurrentSeriesValueINDEX CurrentValueINDEX CurrentFrequencyRangeINDEX 400 CurrentSeriesValueINDEX output to raw data file entire samples fprintf Handle2 6 6f t 6 6f n m MeasurementDataX CurrentValueINDEX m MeasurementDataY CurrentValueINDEX if m MeasurementDataX CurrentValueINDEX LastFrequency amp amp m MeasurementDataxX CurrentValueINDEX gt 0 if CurrentValueINDEX lt FirstRefValue take the first FirstRefValue values and write it to the file fprintf Handlel 6 6f t 6 6f n m MeasurementDataX CurrentValueINDEX m MeasurementDataY CurrentValueINDEX LastFrequency m MeasurementDatax CurrentValueINDEX else if CurrentValueINDEX FirstRefValue TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 96 141 fprintf Handlel 6 6fNt 6 6f n m MeasurementDataX CurrentValueINDEX m MeasurementDataY CurrentValueINDEX RefFrequencyStep log log m MeasurementDataX CurrentValueINDEX log m MeasurementDataX CurrentValueINDEX 1 NextFrequency log log m MeasurementDataX CurrentValueINDEX RefFrequencyStep log LastFrequency m MeasurementDataX CurrentValueINDEX else ThisFrequency log log m MeasurementDataxX CurrentValueINDEX MeanCo
185. rms into a straight line TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 28 141 The logarithmic scale however interferes with the measurement procedure of the FFT spectrum analyzer The SR770 divides the measurement range into 400 equidistant spaced steps and takes a measurement after each segment For a measurement range from 1 Hz to 100000 Hz this means that every 250 Hz a measurement is taken 3000 Range 1 Range 2 3 gt Noise dBV Noise dBV 10 100 1000 10000 100000 10 100 1000 10000 100000 Frequency f Hz Frequency f Hz Fig 4 18 The measurement has to be split into several measurement sections because an insufficient number of samples represent the lower part of the spectrum Fig 4 18 shows that in this case the lower part of the spectrum is only represented by two samples at 1 Hz and 251 Hz Fig 4 18 a This is to imprecise and therefore the measurement range has to be composed by several individual measurement sections Fig 4 18 b For all low frequency noise measurements the frequency range was divided into three spans from 1 Hz to 100 Hz from 100 Hz to 3000 Hz and from 3000 Hz to 100000 Hz Because noise 15 a fluctuating quantity a noise measurement consists of an averaged number of single measurements Table 4 1 shows which number of averages was used for the measurement depending on the frequency span St
186. s Addr4882 t sad return TRUE SetDlgControlState ENABLED sprintf buffer SR770 not found at address d Use Find to search for the instrument m PrimaryInstrumentAddress DisplayStatus buffer STATUS ERROR return FALSE void CFlickerNoiseMeasurementDlg OnSetupSystem FILE Handle char Command 100 int AveragingFlag if FindGBIPCard return if InstrumentActive m IdentificationString return if InitInstrument return DisplayStatus Prepare instrument for setup STATUS SR770 sprintf Command FNM SystemSetup txt Handle fopen Command rb if Handle NULL no fil xiste Create file with standard init commands Handle fopen Command w if Handle NULL SetDlgControlState ENABLED DisplayStatus Error creating file FNM SystemSetup txt Write protection Disc space STATUS return CreateSystemSetupStandardFile Handle fclose Handle Handle fopen Command rb if SendDataSR770 ALRMO this return disable sound while feof Handle fgets Command 100 Handle get Command if ValidCommand Command if SendDataSR770 Command this DisplayStatus ERROR preparing the SR770 for Setup STATUS ERROR return if strncmp Command ARNG1 6 BEGIN if SendDataSR770 AVGO this return Query Averaging if
187. s SIC eee Fig 3 1 Examples of modulation doped layers for HFETs All analyzed devices were AlGaN GaN HFETs with an aluminum mole fraction of 15 to 20 percent and were grown on a sapphire or SiC substrate The devices were processed in collaboration with Cornell and the University of California santa Barbara and were fabricated with a Metal Organic Chemical Vapor Deposition MOCVD The surface was non passivated The carrier sheet density was in the order of 1 1 13 and the mobility varied between 1000 and 1400 cm V s All transistors had a T shaped gate Fig 3 2 with a quarter micron gate length and 100 micron gate width Fig 3 2 Examples of GaN HFETs with T shaped gate The waver maps which describe the location of the analyzed transistors on different wafers and specific layer structures and parameters can be found in Appendix D Waver information a 60 50 40 30 20 Drain Source Current lp mA 10 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 12 141 3 1 Typical device performance Measurement results vary from transistor to transistor and from waver to waver due to the complex fabrication process Numerous transistors on four different wavers were examined and the following graphs represent typical measurement results In Fig 3 3 a current voltage characteristic of a GaN HFET 1 compared to that of a GaAs Pseudomorphic
188. s in the load chart point to the position of the load impedance The noise level of the DUT expressed by the noise figure F depends on the impedance at its input The role of the source tuner is to adjust this impedance level in order to find the position where the lowest noise 15 exhibited by the DUT Fig 4 10 shows a typical selection of source and load impedances for a noise measurement A typical noise measurement result obtained with the ATS 15 shown in Fig 4 11 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 24 141 EN Snpw Swept Display 68 21 File Edit View Window Help Noise Parameters vs Frequency Source Fmin 0 45 dB D DD to 2 50 dB rn 4 45 0 00 to 2 50 0 9186 lt 31 29 Assoc gain 14 03 dB 0 00 to 20 00 dB 16 04 dB 0500 to 20 00 dB Marker Frequency 2 00 GHz Label Frequency UU to 2 U0 GHz GaNHFET10 F07 Vds 10V Vg 1 0V 2GHz cal 07 08 2000 1 Frequency Fmin rn Topt Assoc_gain Gmax GHz dB Mad Phase dB dB 2 00 0 452 4 445 0 9186 31 28 14 028 16 039 Fig 4 11 Typical noise measurement screen in the ATS of Maury Microwave The optimum source reflection coefficient is indicated by the square in the smith chart Circles of constant noise figure for different source impedances are plotted in red color into the smith chart The noise setup like it is depicted in the Maury software can be seen in Fig 4 12 Fig 4 1
189. sg void OnSet EditVg x msg void OnSetfocusEditVr msg void OnFind msg void OnSelchangeComboGpibAddr msg void OnSetupSystem msg void OnButtonEditSetup x msg void OnButtonEditInit x msg void OnButtonEditAvg x msg void OnRadioAvgAuto msg void OnRadioAvgFile msg void OnSetfocusComboGpibAddr msg void OnSetfocusComboGpibTout msg void OnSelchangeComboGpibTimeOut x msg void OnKillfocusEdit x msg void OnDestroy x msg void OnSelchangeComboLogSample msg void OnSetfocusComboLogSample msg void OnTimer UINT nIDEvent AFX MSG DECLARE MESSAGE MAP Fh Fh Fh Fh h th rh Fh Fh Fh a a a a a a a a a a a a a a a a a a a a a a a a a AFX _ Microsoft Visual will insert additional declarations immediately befor endif defined AFX FLICKERNOISEMEASUREMENTDLG FF726926 2 4 11D4 B354 0006295283FA INCLUDI the previous line CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS ED PAGE 88 141 4 3 Implementation file FlickerNoiseMeasurement cpp FlickerNoiseMeasurement cpp Defines the class behaviors for the application include stdafx h include FlickerNoiseMeasurement h include FlickerNoiseMeasurementDlg h
190. should be averaged can be selected d We gh m cy 11 17 Fig 4 Expressive icons are used to reflect the status of the program Their meaning from left to right GPIB card search GPIB card found GBIP card failure SR770 search SR770 found SR770 failure SR770 warning SR770 save data SR770 measurement SR770 measurement interrupted SR770 measurement completed SR770 setup ready SR770 transferring data general information general question Help messages for all parameters and the progress of the measurement are displayed in the status bar of the program Icons help to illustrate if the messages are related to the GBIB card the SR770 or if it 1s a general hint or question Some of the icons can be seen in Fig E 4 TOBIAS WERNER E 3 Manual CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 85 141 Automated Low Frequency Noise Measurement V1 0 General This program is used to control the Spectrum Analyzer SR770 of Stanford Research Systems It communicates with the instrument by sending commands and receiving data over the GPIB bus After measurement is completed it stores the measurement data into two files One file is a raw data file and contains the untreated measurement values Its file name consists of the voltages VG and VD the measured frequency range and the appended text raw Static Settings The voltage and resistor settin
191. stMessage void CFlickerNoiseMeasurementDlg OnSetfocusEditR m LastMessage Please enter the resistor value for the drain resistor R in kOhm DisplayStatus m LastMessage void CFlickerNoiseMeasurementDlg OnSetfocusEditP2 m LastMessage Please enter the resistor value for the potentiometer P2 in kOhm DisplayStatus m LastMessage void CFlickerNoiseMeasurementDlg OnSetfocusEditFstart m_LastMessage Please set the start frequency in Hz 0 Hz lt start lt 100000 Hz DisplayStatus m_LastMessage void CFlickerNoiseMeasurementDlg OnSetfocusEditFstop m_LastMessage Please set the stop frequency in Hz 0 Hz lt stop lt 100000 Hz DisplayStatus m_LastMessage void CFlickerNoiseMeasurementDlg OnSetfocusComboGpibAdar m_LastMessage Please select the GPIB Adress of the SR770 Standard 10 DisplayStatus m LastMessage void CFlickerNoiseMeasurementDlg OnSetfocusComboLogSample m_LastMessage Please select the start sample for the logarithmic data conversion Standard 20 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 94 141 DisplayStatus m LastMessage void CFlickerNoiseMeasurementDlg OnSetfocusComboGpibTout m_LastMessage Please select the timeout for the SR770 Standard 10 s DisplayStatus m_L
192. t ControlFrequencyRange double Start double Stop double Limit boolReceiveDataSR770 char data int count bool exit CFlickerNoiseMeasurementDlg pApp boolSendDataSR770 char command CFlickerNoiseMeasurementDlg pApp boolDisableBoard boolDoMeasurement double StartFrequency double StopFrequency int Averaging CFlickerNoiseMeasurementDlg pApp boolInitInstrument voidGPIB error char msg boolFindGBIPCard boolFindInstrument CString IdentificationString voidDisplayStatus CString StatusMessage UINT Type 0 voidCalculateCurrentID boolSaveMeasurementDataToFile voidSaveLastVariableSettingsToFile voidpause long delay boolValidCommand char Command Member Variables int m DataPointerX int m DataPointerY double m MeasurementDataX double m MeasurementDataY charm FileName 200 charm IdentificationString 50 boolm FirstAverage int m NumberOfPassedFrequencyRanges int m NumberOfAveragingRanges CAveragingRangeType m AveragingRanges int m InstrumentHandle int m PrimaryInstrumentAddress int m SecondaryInstrumentAddress int m TimeOut CString m LastMessage CString m MessageBoxCaption Dialog Data AFX DATA CFlickerNoiseMeasurementDlg enum IDD IDD FLICKERNOISEMEASUREMENT DIALOG CComboBox m_ComboLogSample CStatic m StatusIcon CButton m Button MeasureStop CButton m Button EditAvg CComboBox m Combo GPIB Timeout CComboBox m GPIB Address CString m Status
193. te Source Voltage Ve 4 00 V Ve 0 00 V Ve Ve 1 00 V 2 00 V T me Vg 2 00 V 1 00 V Gate Current uA Drain Source Current b mA e e N Ve 3 00 V 0 Vs 0 00 VvN 4 00 V 0 2 4 6 8 10 12 Drain Source Voltage Vp V Fig 4 2 a Typical DC measurement result obtained with the aid of a b parameter analyzer Hewlett Packard HP4145 For S parameters and noise measurements different instruments are used and assembled in a different way These setups will be explained in the next two chapters 2 The system is actually designed for a frequency range of 2 GHz to 18 GHz but is limited by the two bias Ts at the input and output of the system that are only specified up to 12 4 GHz All other components meet the requirements of 18 GHz as an upper frequency limit TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 19 141 4 1 1 S parameter configuration The configuration of the system for S parameter measurement is shown in Fig 4 3 S Parameter Setup Control Program Maury Microwave MT993C Tobias Werner Automated IV Curve Measurement 0000 00000 0 045 110 Ghz GPIB Network Analyzer go Hewlett Packard pis HP8510C 3 0 45 12 5 Ghz 0 8 18 Ghz 0 8 18 Ghz 0 45 12 5 Ghz Adapter c
194. teries Batteries are used instead of power supplies because the voltage is free from any harmonic distortions caused by multiples of the fundamental power system frequency 60Hz spikes and electrical noise due to radio transmitters power electronic circuits arcing loads and switching power supplies To determine the drain source current Ip a resistor R 100 or 220 is placed between the potentiometer and the drain contact and 75 can be calculated by the expression V O 4 1 All voltages were measured with a digital multimeter HP34401A from Hewlett Packard The quality of measurement relies to a great degree on the contact between the probe tips and the transistor pads Therefore it is indispensable to clean the tips with alcohol proceeding to each measurement The contact also depends on the applied pressure on the needles and should be varied until no significant alteration of the noise spectrum is noticed 4 2 1 Measurement with the SR770 The power spectral density Sy of flicker noise is proportional to 1 f S gt S o o N z 2 9 4 5 6 7 8910 Frequency f Hz Frequency f Hz Fig 4 17 Comparison between the same function proportional to 1 f plotted versus a linear left and a logarithmic right axis If Sy is plotted versus a logarithmic scale Fig 4 17 b instead of a linear Fig 4 17 a the dependence on 1 f is more perceptible because the graph transfo
195. th the help of the Automated Tuner System Software of Maury Microwave VS frequency f Ve 5 V Vp 8 V GaNHFETO1_6 3 Minimum Noise Figure dB 0 2 4 6 12 16 6 10 20 2 M 20 3B Frequency f GHz Fig 5 5 Frequency dependence of the minimum noise figure F min TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 35 141 Fig 5 5 shows typical noise measurement result of F in versus frequency from 2 GHz to 26 GHz for a transistor biased at Vg 5 V and Vp 8 V saturation region is also function of the voltage applied to the gate and drain V of the transistor Linear Saturation Int region region Va 0V 2 5 5 2 1 2 Bias condition for lowest noise Vose A3 loss Drain Source Voltage Vp Ve VA Fig 5 6 Bias conditions for lowest microwave noise The minimum noise figure of the examined GaN HFETs was found to be minimal for a bias condition of approximately a third of the saturated drain source current Ips a and half of the drain source breakdown voltage Vps thus in the saturation region Fig 5 6 5 3 Correlation between the flicker and the microwave noise A graphical way of analyzing a dependence of two data ranges is to plot them versus each other in a so called scatter plot and to examine a visible tendency of the points in the graph If a dependence
196. the discrete measured values and the line is the linear interpolation between them The slope of a line segment can be calculated by taking two discrete points with frequency f and corresponding phase 2 qx uer F 16 If we compare F 15 and F 16 and solve for the electrical length we get 92 f2 ah F 17 72 71 This electrical length can now be inserted in F 13 to calculate the theoretical value of phase at each frequency of and 5 1 Than the theoretical value can be compared to the calculated value of the phase from measurement and can be corrected by 180 degrees if necessary Fig F 7 shows the corrected phase information for 512 and 5 TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 114 141 812 821 Phase correted Phase degree Frequency f GHz Fig F 7 Corrected phase of S12 and S21 The algorithm to calculate the electrical length will only work if the frequency spacing 15 close enough related to the electrical length of the fixture Fig F 5 shows a good left and an unsuitable right selection for the frequency spacing 512 521 Phase narrew frequency spacing 512 H Phage wide frequency spacing 180 i 1 11111 4 E f 4 i 44 a a a Frequency f GHz Frequermcy 552 512 Fiked Fl 2
197. thod B Same bias condition Correlation coefficient 0 46 0 20 Correlation coefficient 0 39 0 22 124 00 4 z 125 00 4 gt 126 00 E 127 00 4 D 128 00 gt 129 00 4r wu 130 00 4 E 131 00 4 gt 13200 A C12 133 00 4 d z 134 00 o 1 00 2 00 3 00 4 00 5 00 6 00 7 00 3 00 4 00 5 00 Frin 2GHz 2 dB Fein 2GHz V57 2V dB Correlation coefficient 0 43 0 21 Correlation coefficient 0 38 0 22 116 00 4 6 00E 03 4 118 00 4 21 5 500 03 T 120 00 gt 4 00 03 T 122 00 4 gt 3 00 03 A 5 E 124 00 2 00 03 i B EN ci e 77 415 1265 00 4 o 1 00E 03 e Tis 2 128 00 T i 0 00 00 1 00 2 00 3 00 400 5 00 6 00 7 00 1 00 2 00 3 00 4 00 5 00 6 00 7 00 Fmin 2GHz V 2 2V Vp 6V dB Finin 2GHz Vg 2V Vp 6V dB Fig 5 12 Scatter diagrams for a comparison between the minimum noise figure F nin at 2 GHz and Sy 8 Sy and at 100 Hz e Sy 100Hz Vp 0 5V dBVrms Hz TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 42 141 The four scatter diagrams in Fig 5 2 are arranged in the same way as described before for the bias condition with the lowest averaged minimum noise figure 2 11 dB The constellation of the po
198. tude and Phase Fixture S 2 abs Fixture S RI 1 overwrite the second column with the magnitude values of 511 Fixture S MA 3 angle Fixture S RI 1 180 pi S12 Magnitude and Phase Fixture S MA 4 abs Fixture S RI 2 Fixture S MA 5 was assigned before S21 S12 Fixture S MA 6 Fixture S MA 4 Fixture S MA 7 Fixture S MA 5 S22 Magnitude and Phase Fixture S MA 8 abs Fixture S RI 4 Fixture S MA 9 angle Fixture S RI 4 180 pi y Fixture S MA return FUNCTION StoreFixtureFile Fixture S INPUT Fixture S MA Fixture file with frequency list in Magnitude and Angle OUTPUT S parameter file named fixture s2p AUTOR Tobias Werner Universitaet Stuttgart function y StoreFixtureFile Fixture S File namel File name2 Create output file name fixture file name sprintf Fixture 5 5 52 File namel 4 size File 1 2 File name2 4 size File name2 2 text sprintf Storing the calculated S parameters in the file s fixture file name disp text SParamFilel fopen fixture file name w if SParamFilel 1 text sprintf ERROR Can not write file output file s Write protection fixture file name disp text disp E X IT PROGRAM fclose SParamFilel 1 end fprintf SParamFilel S parameter file n fprintf SParamFilel Matlab program
199. ub DC 40 Ghz id RF Probes Bias EL Source Tuner LED RA Ze of LoadTuner 1 7 Bias J Gi G E foto An ur g g 5 BNC FG SS Maury Microwave BNC T U S 55 Maury Microwave 2 SS SS BNC mo MT982A01 MT982A01 BNC m g GGB Industries 40A GSG 150P C ul PIB Tuner Control Maury Microwave GPIB MT986B02 Bias Control Parameter Analyzer Hewlett Packard HP4145 Fig 4 3 The microwave setup configured for S parameter measurements The Automated Tuner System ATS software MT993C from Maury Microwave controls the entire setup via the General Purpose Instrument Bus GPIB For S parameter measurements the vector network analyzer VNA HP8510C from Hewlett Packard Fig 4 4b 1s used Its input and output ports are connected to bias Ts HP11590B of Hewlett Packard which are used to apply the gate and drain voltage provided by a parameter analyzer HP4145 of Hewlett Packard Fig 4 2a The bias Ts are necessary to adjust the working point of the analyzed transistor and to prevent interference with the measurement signal of the VNA Maury Microwave 2900 Inland Empire Boulevard Ontario California 91764 USA Phone 1 909 987 4715 Fax 1 909 987 1112 www maurymw com Hewlett Packard Agilent www agilent com TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 20 141 Fig
200. unter tt if MeanCounter 1 LastMeanFrequency m_ MeasurementDataX CurrentValueINDEX MeanAccuY m MeasurementDataY CurrentValueINDEX if ThisFrequency log gt NextFrequency 109 if m_CheckLogAverage take the mean for the last values MeanFrequency LastMeanFrequency m MeasurementDataX CurrentValueINDEX LastMeanFrequency 2 0 fprintf Handlel 6 6f t 6 6f n MeanFrequency MeanAccuY MeanCounter else take only the value at this frequency fprintf Handlel 6 6fNt 6 6f n m MeasurementDataX CurrentValueINDEX m MeasurementDataY CurrentValueINDEX MeanCounter 0 MeanAccuY 0 NextFrequency log RefFrequencyStep log LastFrequency m MeasurementDataX CurrentValueINDEX else if CurrentValueINDEX m NumberOfPassedFrequencyRanges 400 1 store last value in series fprintf Handlel 6 6fNt 6 6f n m MeasurementDataX CurrentValueINDEX m MeasurementDataY CurrentValueINDEX fclose Handlel fclose Handle2 return TRUE bool CFlickerNois asurementDlg DisableBoard ibonl BOARD INDEX 0 SendDataSR770 LOCLO this set to local free allocated memory NOCH Wieder aktivieren und ueberlegen ree m MeasurementDataX ree m MeasurementDataY ree m AveragingRanges return TRUE bool CFlickerNoiseMeasurementDlg SendDataSR770 char command CFlickerNoiseMeasurementDlg
201. ury 717 EH Phasa n rM JL INDEM o PM Fig F 9 Comparison between the results for an S parameter determination of a commercial system blue and the presented Matlab program red On the left side is the magnitude and on the right side the phase of the S parameters TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS PAGE 116 141 F 5 Matlab source code function y CalculateFixture Funktion Calculate Fixture Author Tobias Werner Universitaet Stuttgart Date Februar 2000 DESCRIPTION Function opens two calibration sets of the HP8510C and calculates the S Parameters of the fixture clear all clc home disp disp disp ELX TURE CALCULATION disp xj disp disp This program calculates the S parameters of a disp fixture from two HP8510 error correction terms disp Februar 2000 disp Tobias Werner disp University of Stuttgart disp disp Ask for calibration file without fixtur text sprintf nIn th disp text eval ls CS to open 5 55 5 5 5 5 SELECT AND OPEN CALFILE WITHOUT FIXTURE file namel input nPleas
202. ut class CAboutDlg public CDialog public CAboutDlg Dialog Data AFX DATA CAboutDlg enum IDD IDD ABOUTBOX ClassWizard generated virtual function overrides AFX VIRTUAL CAboutDlg protected virtual void DoDataExchange CDataExchange pDX DDX DDV support JAFX VIRTUAL Implementation protected AFX MSG CAboutD1g AFX MSG DECLARE MESSAGE MAP CAboutDlg CAboutDlg CDialog CAboutDlg AFX DATA INIT CAboutDlg JAFX DATA INIT void CAboutDlg DoDataExchange CDataExchange pDX CDialog DoDataExchange pDX AFX DATA MAP CAboutDlg AFX DATA MAP BEGIN MESSAGE CAboutDlg CDialog AFX MSG MAP CAboutD1g No message handlers AFX MSG MAP END MESSAGE MAP CLPTSwitchRemoteDlg dialog CLPTSwitchRemoteDlg CLPTSwitchRemoteDlg CWnd pParent NULL CDialog CLPTSwitchRemoteDlg IDD pParent AFX DATA INIT CLPTSwitchRemoteDlg m SetupConfig 1 JAFX DATA INIT Note that LoadIcon does not require a subsequent DestroyIcon in Win32 m AfxGetApp LoadIcon IDR MAINFRAME void CLPTSwitchRemoteDlg DoDataExchange CDataExchange pDX CDialog DoDataExchange
203. uted to this work Jianyu Deng Doctoral candidate at the RPI has been an inexhaustible source for discussions on all kinds of microwave noise related aspects Serguei Roumiantsev provided many comments and suggestions concerning the low frequency noise setup and the mechanisms of flicker noise I also extend my appreciation to all staff members of the institute for their assistance and help It was a great pleasure for me to work with them and the friendly atmosphere in the SDG turned work into pleasure Finally I owe special gratitude to my family for their continuous support and to my fianc e Sabine for her patience understanding and love ABSTRACT The correlation between the low frequency flicker noise and the microwave noise at room temperature of gallium nitride GaN heterostructure field effect transistors HFETs 1s investigated The study is based on a comparison between the noise levels of the microwave and the flicker noise The minimum noise figure at 2 GHz represents the microwave noise and the spectral voltage noise density Sy the spectral current noise density Sy the relative spectral current noise density S or the Hooge parameter at 100 Hz expresses the flicker noise Both noise types are functions of the applied voltages at the gate and drain of the transistor but show different behavior Whereas the optimum bias condition for lowest microwave noise is in saturation and the noise in the lin
204. uto scale if pApp SendDataSR770 STRT return FALSE Start measurement if pApp gt SendDataSR770 AUTSO pApp return FALSE Auto scale if pApp gt PollForMeasurementCompletion pApp return FALSE if pApp gt SendDataSR770 AUTSO pApp return FALSE Auto scale PostMessage pApp gt m_hWnd WM START COUNTDOWN WPARAM 10 0 if pApp gt GetMeasurementData pApp return FALSE return TRUE LONG CFlickerNoiseMeasurementDlg OnDisplayStatus WPARAM Message LPARAM Status DisplayStatus char Message UINT Status return 0 LONG CFlickerNoiseMeasurementDlg OnSaveData WPARAM wParam LPARAM lParam if SaveMeasurementDataToFile SetDlgControlState ENABLED CountDownSeconds 0 DisplayStatus Measurement finished successfully STATUS SUCCESS SendDataSR770 ALRM1 this SendDataSR770 MSGS C TOBIAS WERNER 2000 this else SetDlgControlState ENABLED TOBIAS WERNER CORRELATION BETWEEN THE FLICKER AND THE MICROWAVE NOISE OF GAN HFETS DisplayStatus Measurement was canceled STATUS STOPPED DisableBoard return 0 LONG CFlickerNoiseMeasurementDlg OnMeasurementStoped WPARAM wParam LPARAM 1Param SetDlgControlState ENABLED DisplayStatus Measurement stopped STATUS STOPPED DisableBoard return 0
205. wn in Fig G 3 is the 74HCT245 a 3 state octal bus transceiver In the selected configuration it passes the data byte at port A to Port B if the enable port G 1s activated low active The connector labeled LPT IN receives the parallel port signal from the motherboard and passes it directly to the LPT port on the rear of the computer via the connector labeled LPT OUT The data byte of the parallel port signal is also passed to the transceiver data input ports The printer enable signal nSelect Printer is connected to the enable port of the transceiver and activates the setting of the microwave switches The port assignment of the parallel port and corresponding signal name is outlined in Table G 2 D Type main board Centronics In out Inverted 228 pP J pu 355 B paal pu ___ ___ 5 p 6 1 6 ___ 7 3 7 ___ S 15 8 Qu ___ 9 7 9 1 pl 1 Busy Paper Out Paper End n Out 15 B B2 pEmor nFalt fn 16 _ _ Gl plniialize p6 pmSelectPrinter nSelect In Out Q5 Ground Gnd 25 4 6 0 pe NO p7 f Lp p _____ po Ground Table G 2 Parallel port LPT assignment and signal names ot un tatus m m m tatus
206. ystemMenu FALSE if pSysMenu NULL TOBIAS WERNER CString strAboutMenu strAboutMenu LoadString 105 _ABOUTBOX if strAboutMenu IsEmpty pSysMenu gt AppendMenu MF_SEPARATOR pSysMenu gt AppendMenu MF_STRING IDM ABOUTBOX strAboutMenu Set the icon for this dialog The framework does this automatically when the application s main window is not a dialog SetIcon m_hIcon TRUE Set big icon SetIcon m hIcon FALSE Set small icon TODO Add extra initialization here m TimeToGo m MessageBoxCaption Automated Flicker Noise Measurement sprintf m IdentificationString Stanford Research Systems SR770 s n24484 ver091 GetLastVariableSettingsFromFile if m Radio AveragingMode 0 m Button EditAvg EnableWindow FALSE m GPIB Address SetCurSel m PrimaryInstrumentAddress 1 Set GPIB address m Combo GPIB Timeout SetCurSel m TimeOut Set time out m ComboLogSample SetCurSel m ComboRefStartValue CalculateCurrentID UpdateData DATA TO DIALOG InstrumentInitialized FALSE return TRUE return TRUE unless you set the focus to a control void CFlickerNoiseMeasurementDlg OnSysCommand UINT nID LPARAM lParam if nID amp OxFFFO IDM ABOUTBOX CAboutDlg dlgAbout dlgAbout DoModal else CDialog OnSysCommand nID lParam T you add a minimize button to your

correlation between the flicker and the microwave noise of gallium

Contents

Download Pdf Manuals

Related Search

Related Contents