Home

Can / Am EMTP News - the European EMTP

1. even after automatic scaling see preceding paragraph This weakness was overcome by forced application of the final point to the extrema vectors Using an old program the user has two alternatives for avoidance of the problem 1 zero the smoothing tolerance or 2 alternatively use FACTOR to manually scale the curves believably News from Outside USA and Canada A 252 page soft bound book about EMTP and specifically and conspicuously the ATP version was received at BPA in a mailing envelope from Doshisha University postmarked Oct 9 The book is a work of the Japanese EMTP Committee published by Ohmsha or Ohm its English symbol or logo with the introduction written by JEC Chairman Akihiro Ametani Your Editor can understand the many figures and diagrams and graphics which contain some English As for the text Dr Liu can read the Chinese if not the Japanese characters On the cover is the Internet address http www ohmsha co jp Spanish and Portuguese are alternative languages for the publication of EMTP news For background see the April 1998 newsletter Prof Juan A Martinez Velasco of Universitat Politechnica de Catalunya in Barcelona Spain made the following announcement in public E mail that was received early in the morning of November 16th not to be confused with the futuristic message date A new issue of the EMTP News in Spanish and Portuguese is already available It can be downloaded from the fol
2. Simulation speed for TACS first was measured on December 4 This involved the same data as for the original TACS ASSEMBLY LANGUAGE TAL as summarized in the January 1998 issue Once again we use the best 5 of 6 consecutive trials of the 60 variable 100K step MATHTACS DAT But this time Salford DBOS is run under DOS of Win95 rather than within a DOS window of Win95 We now can add a row to the preceding tabulation of time step loop seconds using your Editor s 133 MHz Pentium Dube s original TACS 60 variables 106 21 TACS assembly language 60 variables 18 33 New built in compiler 60 variables 18 41 Original TACS one dummy variable 8 46 About time to compile this additional burden would be part of the initial timing figure which measures data input This has not changed significantly Again taking the best 5 of 6 figures gives 2 08 seconds No big deal your Editor s compiler is plenty fast Linux and Programs It Supports C like PL4 files now are offered by GNU ATP thanks to the efforts of Masahiro Kan of Toshiba Corporation in Japan For background see the story entitled Szymanski C like PL4 structure in the preceding issue Your Editor did proceed as planned modularizing and testing using Salford library functions But then came Mr Kan s surprising response Quoting from E mail dated September 9 I looked into the source code and I believe the work I should do is to write the following functions i
3. devices dominant the TACS usage and the difference between compiled and not compiled devices is not great Disk file AUT5 DAT allows the addition of a third row to the April table Times for the time step loop in sec Dube supplemental variables 72 802 Meredith Type 68 device 9 011 Compiled TACS of this 6 209 COMPILED TACS MAKE CTM finally is illustrated by standard test cases A new 6 subcase was added to DCNEW 25 on October 8 for two reasons actually Yes it illustrates CTM important But it also documents the basic pocket calculator formulas that were used many times to illustrate how slow Dube s MODELS and TACS can be for such elementary mathematics Six equations are involved When these are copied 9 times one has the 60 variables mentioned in the January 1998 issue About possible replacement of CTM by COMPTACS MAKE the reader is advised not to try An extraneous error termination resulted due to the stacking Since it seems unlikely that CTM ever would be appended to the end of another subcase for practical use the trouble might not be worth the effort to resolve BEGIN SUPPLEMENTAL ONLY and END SUPPLEMENTAL ONLY are two new declarations to accommodate standard FORTRAN which begins in column 7 To understand the motivation consider some TACS history During 1984 Chinese visitor Ma Renming made several important improvements to the TACS code that originally was written for BPA by contractor Laurent
4. 1997 issue Before some words about hysteresis your Editor made a general observation Jf you or anyone else knows of any other group of Indian users of ATP that information should be mentioned in our newsletter More about the Internet and E mail There is a 72 byte limit on lines of E mail as received by MS Mail running under Windows NT on Dr Tsu huei Liu s Pentium at BPA As a result 80 column EMTP data lines that might be included in messages generally will be broken unfortunately This problem is well known to many but is ignored by some who request help using the Fargo list server On October 20 your Editor reminded readers of the problem once again under a headline that read Periodic advice about broken data lines The following day Orlando Hevia of Universidad Tecnologica Nacional in Santa Fe Argentina explained how he avoids such difficulties Z am using Pegasus mail program and the option Wordwrap long lines can be disabled and the file saved without such modifications On the other hand neither Mr Hevia nor any other respondent explained how to avoid the wraparound using MS Mail Amazon com Books Music amp More is the title of the Web page used by the pioneering Internet retailer Anyone needing to check the name of a book or its authors is advised of this resource which offers a search engine This is better than a card catalog for the electronic book store Millions of titles are included
5. about 50 miles west of the University A block of guest rooms will be available for course participants at the Cedar Cove Beach and Yacht Club This is an environment conducive of diligent and uninterrupted study joke Price is interesting Although the basic fee is higher substantial discounts are offered The base course fee is 1350 There will be a 200 discount for registrants bringing their own portable PC and an additional 100 discount for early registration prior to February 1 1999 University students will also receive a 200 discount For information send E mail to dcarr ece ufl edu Branch Data Input Restructured An error associated with column 80 punches of branch cards was removed by restructuring of GETBUS use on October 15 The trivial little example that was being improperly handled was MRVIC2 DAT as received earlier that same day from Masahiro Kan of Toshiba Corporation in Japan The name of this file is a reference to the person first reporting trouble Jovan Mrvic at the Institute of Nicola Tesla in Belgrade Yugoslavia The combination not being handled properly was a constant parameter 3 phase distributed line followed by a Type 92 nonlinear element with 4 punch in column 80 In turn the nonlinear element this was followed by a series R L C branch with a 1 punch in column 80 Prior to correction there was one output too many in the LIS file and the PL4 file was obviously broken TPPLOT refused to p
6. the date Still when this becomes 2000 the two least significant digits will be 00 and these would be used to name the PL4 file Or would they No we do not have the space We have been saved by lack of space Yes the year 2000 problem should be the same as the year 1990 problem We only use one year digit so we have the problem every 10 years Who can forget the havoc created by the year 1990 problem joke Really all this concern about Y2K seems stupid and wasteful to me Those who inquire typically seem not to recognize the difference between programs that require a current date and programs that do not I have yet to hear from one person I respect who believes Y2K will bring major disruption Consider some of the questions that Prof Prikler was asked Question 1 Will you certify that your product ATP will perform without disruption for all dates of the next century Your Editor s answer No The infamous 1990 problem already has been explained This might plague users well into the next century Question 2 Is there today an upgrade or newer version that can be offered and that is year 2000 secure specify Your Editor s answer I think this comes closer to the real reason so many persons have Y2K interest upgrades can be sold Question 3 Will you describe the consequences of the year 2000 problem Your Editor s reaction The person who asks the questions does not seem to know what the problem is He doe
7. E mail early that same day ATPDraw from Hans Hoidalen Trailing blanks of punched cards might sometimes cause trouble with ATPDraw This was detailed in three public E mail messages dated October 7 with the first coming from Jovan Mrvic at the Institute of Nicola Tesla in Belgrade Yugoslavia This gentleman first provided the surprising explanation that punched cards produced by Salford EMTP have no trailing blanks through column 80 whereas other program versions e g Watcom ATP do Following research your Editor explained the difference later that morning Yes Salford punched cards are missing trailing blanks on the right But full 80 column card images seem to be written It would appear that DBOS or DOS is removing trailing blanks The code clearly orders the output of all 80 possible bytes It is proposed that source code be modified to strip trailing blanks from each punch line Salford does not need this help but it would appear that other program versions would be improved by it As no reader objected within the allotted 24 hours the change was made All punch files DC PCH created by standard test cases then were compared before vs after using Mike Albert s freeware FC and all were found to be identical for Salford Statistical Outputs with ATPdraw was the subject of public E mail from Philippe Bergin of Schneider Electric in Grenoble France He inquired on October 22 about possible support by ATPDraw In h
8. are 3664 and 82216 respectively for use with and without RWC More details are ready for the next issue Bernd Stein of FGH in Mannheim Germany was mentioned in the October newsletter His data too continues to grow He reported progress in E mail dated November 9th The data case now encloses 3121 nodes up to DU2664 automatically serialized 6570 branches and about 19000 data lines after inclusion of all DBM data The square of the maximum number of coupled phases LSIZ26 is still 5476 as we do not yet model coupling between the three phases Brain Damaged MS Windows John Kappenman is mentioned in a later story He too seemed to be a victim of Bill G s incompatibility with himself Kappenman wrote I am having a problem getting my hands on a printable licensing form This is the first complaint your Editor can recall about MS Word not importing the user group s standard WP 5 1 disk file At BPA in Portland using Word under WinNT on Dr Liu s Pentium there is no problem But a copy of the same file sent to Duluth was rejected Why About software being used Mr Kappenman wrote J did try it on both the old license agreement as well as the new Both are WPS I can at least get an old WPD file to open though it is garbage in the translation still no go I am operating in Win95 with a Word version from Office97 Like I mentioned before I can open and translate most WPD files but apparently not this WP5
9. but now an explanatory error message will precede a civilized halt As background it should be explained that ATPDRAW overflow was the Subject of the original complaint That s right it illustrates how data assembly isolates the user from ATP data Mr Esztergalyos later explained JLINE in columns 51 52 defaulted to a number I instead of a O that requires an input of the surge impedance Z in column A and propagation v in column B The User s Manual for ATPDraw version 1 0 for Windows is now available on the ftp ee mtu edu server This according to author Hans Kr Hoidalen who issued a public announcement using the Fargo list server on December 7th Continuing The manual is on a zipped pdf format for Acrobate Reader 3 0 or later The manual contains elementary information on how to use ATPDraw but also information for more advanced users along with several useful examples The sample circuits in the manual will be distributed with ATPDraw A parallel announcement the same day was about ATPDraw itself Author Hoidalen provided an impressive explanation from which the following has been extracted as a brief outline Version 1 3 includes several extensions and useful updates as will be documented below but is not totally debugged yet Version 1 3 of ATPDraw for Windows is capable of reading the MOD file a text file starting with MODEL lt lt ModelName gt and ending with ENDMODEL directly examine its input o
10. dimensional parameter variation is possible for LINE CONSTANTS and probably also for CABLE CONSTANTS or CABLE PARAMETERS if one of the two parameters to be varied is frequency This innovative idea came from BPA s Dr Tsu huei Liu Upon seeing the just described variation of conductor height she suggested the switch from a single fixed frequency to a scan of frequency as used by the JMARTI SETUP fitter It worked perfectly on the first try and is illustrated by a new 11 subcase of DC 59 Note that variation of the raw data geometry is the outer loop whereas variation of frequency is the inner loop Both the LIS file and the PCH file will include all results in this same order with comment cards of the latter clearly delineating the start of each new pass of the outer loop Any STATISTICS or SYSTEMATIC data case is incompatible with the PCVP declaration This observation about your Editor s programming came from BPA s Dr Tsu huei Liu during debugging of PCVP use with LINE CONSTANTS It is true at least one variable KNT is used by both features but with different meanings Is this limitation worth removing It seems to your Editor that simultaneous use would be impractical It would be placing too many eggs in one basket By itself STATISTICS or SYSTEMATIC is enough of a challenge generally What reader disagrees and why Timing of PCVP use with a supporting program passed through simulation formulas prior to correction
11. file To solve the problem we in Portland exported as a Word file and sent the result End of difficulty Corrections to This Newsletter Note that Prof Hermann Dommel s contribution to the preceding issue was not published under the present title It was not submitted along with any claim that your Editor s writing was in any way in error Prof Dommel did apologize but not for his misleading association of BPA with the error it is to be noted Prof Dommel s characterization of DCG EPRI EMTP version 1 1 as used by TEPCO as a very old version is revealing The professor indicates that corrections were made in EMTP96 Version 3 But according to EPRI s own projection see the October 1996 issue this was not available to members two years prior to Prof Dommel s writing In round numbers of the 14 years of DCG EPRI EMTP development the error had been corrected for two years at most Rather than refer to a version that is uncorrected as very old it might be more accurate to refer to the corrected version as very new Of course this would be more embarrassing to both the professor and his sponsor It also makes more preposterous the association of BPA with the error Not only did BPA have no special responsibility during the mid 80s it has supplied to DCG or EPRI no EMTP work since then as far as your Editor is aware The error clearly is DCG s Variable Dimensioning of ATP The Type 50 SCE
12. from a user Eng Eduardo Lalla surely the biggest user in Argentina about a bigger TPBIG I sent to him the gnu TPBIG but there is a problem with the time to load about a minute for each run So you Editor quickly redimensioned Salford EMTP using Mr Hevia s dimensions preserved at BPA in disk file LISTSIZE 20K which is used by VARDI20K BAT The 20K reflects the maximum number of nodes or branches Total table size is staggering LABCOM tables 10694972 INTEGER words which is just under 43 Mbytes Yet Salford EMTP simulated DC 6 without any apparent hesitation About possible table dumping your Editor warned Mr Hevia God help the person who tries to dump tables Beware of STATISTICS etc Well if you had 64 Mbytes of RAM and this all were available to DBOS I guess you could dump tables But beware of Bill G and the games he will play with Windows memory management Improvements to Salford TPPLOT AUTO SCALE is the keyboard equivalent of a mouse click on the AUTO button of the Multipliers heading within the CURVE window This service began November 7 when your Editor realized that a user without a mouse was deprived of an easy way to display both power and energy on the same graph The new command should be honored even though it will not be found in any of the colored pull down menus In order that it not be forgotten it will be added to the HELP file however Even users with mice should find occasi
13. introduced in ATP and are supported by ATPDraw as well A new harmonic source is also introduced Handling of transposition for objects having several 3 phase nodes is improved The icon of each component can be edited individually The high precision option VINTAGE 1 has been extended and improved All models are written directly in the ATP file instead of using INCLUDE TACS Device 50 has received an extra variable max acceptable change The ATP Settings dialog box has changed a lot Frequency Scans and Harmonics Nonlinear frequency dependence within HFS first was used realistically by Gabor Furst who reported success in E mail dated October 2 I attach the file DUMF DAT which is a semi realistic HFS file with some of the sources made near zero to be able to use PCVP I have a line from GEN to TRAN with 2 j6 1 ohms and a load from TRAN to TERRA called RESIST This resistance is varied by the Funk Hantel formula It works very nicely I checked the results manually and they are correct I think we have something really neat now The following day Mr Furst s data was added to DCNEW 26 as a new 4 subcase Neatness continued to improve too Important progress of that morning included elimination of the need for all harmonics note that only numbers 1 3 5 7 and 11 are involved That afternoon Mr Furst s request for polar output POLAR OUTPUT VARIABLES was activated previously any of
14. no different than for ordinary FREQUENCY SCAN see the illustrations of DC 51 and 52 If nonzero beware of voluminous output Success superimposing two frequencies on a multiphase Pi circuit occurred November 12 A new 6 subcase of DCNEW 26 documents this higher frequency 400 and 800 Hz balanced multiphase use that involves that famous 3 kHz line section from DC 3 One end is excited by balanced 3 phase voltage sources while the other is connected to ground by 3 equal R L branches that draw plenty of current The period is 2 5 msec and agreement of unknown voltage nodes for the first two cycles is obvious The amplitude of each harmonic is unity and the LIS file shows the following for unknown voltages Step Time TRANA TRANB TRANC 0 0 0 6411782 655262 0054071 100 0025 6410699 655039 0052891 200 0050 6410606 655011 0052750 Nonlinear reactors having type code 93 true nonlinear 10 or 98 pseudo nonlinear were made compatible with superposition on November 16 and 18 Prior to this Orlando Hevia of Universidad Tecnologica Nacional in Santa Fe Argentina included such elements in his simulations and seemed to be producing results that were useful for engineering However if one looked closely at time steps 1 and 2 discontinuities were obvious Mr Hevia s use of AVERAGE OUTPUT was an indication that initialization was imperfect Well no longer A new 9 subcase of DCNEW 26 demonstrates a smooth start
15. on October 25 Sometimes it died on DC 59 with division by zero time This was using BPA s shared 90 MHz Pentium across the hall Why the trouble was not seen at home on your Editor s 133 MHz Pentium or on Dr Liu s 486 at BPA is not known Salford DBOS is newer of course If one symbol contains another the shorter one must follow the longer one rather than precede it This was a new restriction beginning October 23 when the illustration of DCNEW 25 was changed This became necessary following permission of one or more blanks preceding the equal sign When such a blank was part of the symbol it distinguished MILLIH from MILLIHENRIES for example But once this distinction ended once trailing blanks were ignored as they should be the shorter symbol would be located where the longer was intended In fact both would be found if the shorter had been declared first ATP now checks for such conflicts and will issue a message such as the following Notice Symbol 1 is contained within symbol 2 Swap these two and try again Use of FMTPL4 widexx with a PCVP loop over simulation resulted in missing heading lines of the 2 and later PL4 files I e the 001 file was correct and complete whereas the 002 etc files were missing the first 3 lines This was prior to corrections in TSHEAD and PLTFIL late on December 3 Orlando Hevia of Universidad Tecnologica Nacional in Santa Fe Argentina first reported the problem in
16. standard test cases on September 19 in order to document three different types of PCVP usage related to phasor solutions By design all three have comparable output The first subcase creates its own loop over frequency whereas the second relies upon a familiar FREQUENCY SCAN declaration and the third upon a HARMONIC FREQUENCY SCAN declaration to 12 do the looping This is in addition toa PCVP declaration Plotting as a function of frequency is permitted of course and the associated PL4 file can be saved for later use by another program For the F scan user the past 3 months have brought an important expansion of capability Supporting programs are the third class of PCVP studies Operation began October 2 with only details differing from usage of the other two classes A new 10 subcase has been added to DC 59 to illustrate use with LINE CONSTANTS to vary the height of a transmission line About differences there is printout control JIOPCVP which has only limited effect as initially coded Of course full output requires no explanation But reduced output is not reduced as much as for the other classes of PCVP use Only printout outside the supporting program in question will be suppressed About the PCH file punched output DC59J PCH will contain the card images for all passes concatenated in a single file I e such output is unified As usual comment cards document which geometry produced which set of branch cards Two
17. the three alternatives to magnitude only led to a later error Supercaps i e super capacitors possess frequency dependence that might sometimes require representation This detail from the cutting edge of harmonic analysis was learned from Gene Porter of Power Profiles Inc in Knoxville Tennessee The following summary was received in E mail that was dated October 5 If you have some time visit http wwwl psi ch www_f5_hn 14 Electrochemistry supercaps html to see the frequency response of a few supercaps A practical example we investigated was the use of supercapacitors on the dc side of a pwm drive to increase the ride through capability of the drive during sags and such A boost regulator was used in conjunction to further increase the ride through capability The simulations showed that it worked but we were skeptical about the model In the end we suggested that a frequency dependent model of the capacitor was necessary for accurate analysis My portion of the work with Ferraro Oliver and Associates amp Technology Insights ended and I dont know what has happened since Well ATP would seem to be ready The founding father of HFS Gabor Furst assured Mr Porter that we have now the ideal tool for such frequency dependent simulation The ATP PCVP_ is now working fine I will write to you in more detail Limits of LINE MODEL FREQUENCY SCAN LMFS were removed October 26 following a report of trouble from Pr
18. Can Am EMTP News Voice of the Canadian American EMTP User Group Publishers and Mailers Drs Kai Hwa Ger and Tsu huei Liu 3179 Oak Tree Court West Linn Oregon 97068 United States of America Table of Contents Salford Compiler and DOS Extender 0 1 Improvements to Salford TPPLOT seese 2 News from Outside USA and Canada ue 2 More about the Internet and E mail oe 3 Watcom ATP for MS Windows ou eee 4 News About TACS and MODELS seee 4 Line and Cable Constants oo eeeeeseeeeseeeeseeeees 5 Higher Order Pi Circuits eee ee eseeeeseeeeeeeeeeeeeee 5 Brain damaged MS Windows ou cceeseeseeeseeeeeeeeees 6 Corrections to This Newsletter cess eseeeeeeees 6 Variable Dimensioning of ATP seese 6 Evaluating Analytical Functions I oe 6 Linux and Programs It Supports essees 8 Compiled TACS amp Assembly Language 0 0 9 Superposition of Phasor Solutions oe 9 Stu Cook Uses Apple Macintosh oo eee 10 Comings and GOINGS 00 eeeeseeeeseeeeeeeseeeeseeseees 11 Florida Resort Course March 8 11 wee 11 Branch Data Input Restructured ce ceeeeeeeseeseeteeees 11 Parameter Variation Studies by PCVP 11 ATPDraw from Hans Hoidalen neeesser 13 Frequency Scans and Harmonics s s s 14 Changes to Universal ATP Graphics eee 15 Year 2000 Compliance of ATP woes 15 Data Sorting by Class eee eseeeeseeeeteeeeeeeees 16 Use of Initial Condition Cards oes 17 Partial Table Dumping o0 eee eeeeereeeeeeeeeeeee 18 Int
19. Dube Probably few readers recall the separation and ordering of data by TACS class with each terminated by its own blank card Thanks to Mr Ma s creative work this no longer was necessary in the M39 version dating to July of 1984 But the price that was paid for such progress was reliance upon a unique type code in columns 1 2 for all TACS data The card requesting TACS output variables became type 33 etc Well standard FORTRAN which uses columns 1 through 5 for a statement number collides with this assumption Standard FORTRAN can be allowed only if some means other than the type code of columns 1 2 is used to determine TACS data type The request words that began this paragraph are one such means and they always works There is no limit on the number of times such switches can be used either or where they can be placed within TACS data They work for TACS data much as VINTAGE or UNITS does for branch data They are needed for standard FORTRAN if the equal sign is not preceded by one blank and followed by one blank This latter rule is simple and probably will be used more even though it is not perfectly general Superposition of Phasor Solutions The proper way to handle multiple frequency sinusoidal excitation of the steady state phasor solution is to use superposition Or is it For the first time in two decades minor progress has been made Superposition of phasor solutions has been demonstrated for linear networks of s
20. It remained broken until October 6 when it too was modified to handle 132 column names This was the day following the first complaint of trouble Hegal file name Try again was the wrongful complaint reported by Prof Juan Martinez Velasco of the Polytechnic University of Catalunya in Barcelona Spain in E mail dated October 5 Whether DBOS actually will return names in excess of 80 characters has not yet been tested but at least now the source code should be ready Those long file names see preceding issue never were tested using Salford DBOS although the code seemed to execute correctly for the usual short names That was using ver 2 66 and 2 67 of DBOS Upon switching to ver 3 50 compilation linking and execution on October 24 it was found that execution sometimes would end prematurely always near the start of any second or later subcase of simulation The problem was zero rather than blank fill of the uninitialized end of ANSI32L in SYSDEP This trouble already was explained in the July issue the only difference this time was 80 rather than 32 bytes as the start of the garbage Really big ATP networks have been simulated even without large numbers of coupled coils see high order in recent issues Such cases make the superiority of Salford DBOS obvious One request was from Orlando Hevia of Universidad Tecnologica Nacional in Santa Fe Argentina He wrote as follows on November 30 I received a requirement
21. Only if a name has not been defined by the user will ATP tables be searched for it Quotation marks ceased to be a requirement and were removed from the 1 subcase of DCNEW 19 on November 28 Changes to Universal ATP Graphics PostScript output changed in small ways September 28 in response to experimentation by Robert Meredith of the New York City area In E mail dated September 20 he supplied changes to three modules Most adjustments are related to the spacing of text Jt works better with lower 15 case descenders I found that there is an inconsistency in how plotted title lines are spaced and how the legend lines are spaced The header of a widexx PL4 file might be mildly corrupted if load flow FIX SOURCE is involved Attached to E mail dated October 26 Prof Mustafa Kizilcay of FH Osnabrueck in Germany supplied disk file DC25_1 DAT to illustrate the problem An extra unintended and corrupted 4 heading line of FS or HES not being used by the data was being produced due to a variable conflict between KTAPER in FXSOUR and MODHEFS used elsewhere Correction was made later that same day MODHFS was mentioned in the October issue as a possible appendage to PL4 files On December 12 there was a modification to the circumstances under which such information is appended The creation of a PL4 file now is required although one or more comment is not for the case of simulation as opposed to frequency scan The
22. PL4 file and shaft data from an auxiliary file The shaft section data may be in metric or English units PLOTXY comes from Massimo Ceraolo of the University of Pisa in Italy as explained in the July 1998 issue December 14 BPA s Dr Tsu huei Liu and your Editor verified use of this separate interactive plotting program under Windows NT for plotting of an ongoing simulation This development which promises to be important as a replacement for SPY PLOT should be the subject of a separate story in the next issue It also should justify a new improved PL4 file format Miscellaneous Intel PC Information The Asian financial crisis now a year and a half old has resulted in dramatic reductions in prices of PCs in many other countries such as the USA High end workstations of unprecedented power have become amazingly affordable Consider the following October quotation to BPA by a local company named Affordable Technology For 1818 one can purchase a 400 MHz Pentium II PC having 128 Mbytes of RAM a 6 4 Gbyte hard disk fancy output card 8 MB ATI EXPERT WORK PRO AGP floppy disk keyboard mouse 100 Mbyte Iomega ZIP drive CD ROM 32X LITEON sound card and speakers 17 inch monitor 26 PROVIEW DIG and Windows NT That was at the high end For the low end consider an offering on page 8 of the November mail order catalog from MEI Micro Center of Columbus Ohio Look what 518 will buy 233 MHz Cyrix MediaGX MMX process
23. active plotting program as explained in the July 1998 issue Well it began as an alternative to Salford TPPLOT which will not run under MS Windows NT But it had the advantage of reliance upon a graphics library that was available for many platforms so extension to completely different Linux was just a matter of time This important milestone is documented by E mail from the author to Masahiro Kan of Toshiba Corporation in Japan on October 24 I send the source code of GTPPLOT for Linux the makefile the binary the ini and hlp files Some work remains Mr Hevia lists 5 things Please send your opinion if you can run the program The following day author Hevia added For Linux the best option for hard copy are the Postscript files I dont know if hp2xx is available for Linux Three days later Mr Kan responded with a favorable evaluation It also worked fine at home without the libdislin so 7 file Gtpplot Linux is ready to publish Anyway you did a great job Three days later Mr Hevia acknowledged the importance of Mr Kan s support for Linux He also related interesting details about his recent acquisition Thanks with the help of a good friend By the way the RedHat Linux 5 1 took 3 days to travel from the USA to Santa Fe The package cost 49 95 and included Installation guide 2 CD ROM I CD ROM as catalogue of Linux products and demos Delivery by Federal Express added 32 and the Argentine governm
24. al now execution runs to completion Let me know what more might be wrong if anything and I will work on it One or more blanks separating a parameter name from its following equal sign were not generally permitted prior to October 22 The preceding day Gabor Furst had noticed that column boundaries for R and L of a series R L C card were not being properly observed This was for his induction motor model within an HFS study The error was traced to an extra blank left of the equal sign of a substitution variable But blanks provide useful visual separation and should be encouraged So the logic of S N 1094 and below was added to MATDAT in order to ignore any number of blanks preceding an equal sign The Type 10 electrical source card carries an analytical expression to the left of familiar T start in columns 61 70 Unfortunately an easy error for the user to make was a formula that was too long If the formula spilled over into T start the result generally would be unreadable and execution would die for Salford EMTP with DBOS opening a debugger error window Beginning October 28 ATP will check for legal T start and T stop and will end the subcase in civilized fashion KILL 236 if an error is found This is illustrated by a new 2 subcase of DCNEW 19 the former 2 has become the 3 While few users care about Type 10 sources the improvement is believed to be noteworthy as a test of procedures that might be ap
25. and the user is able to search for key words more powerful than simple alphabetical sorting of titles Prices are attractive with discounts from prices of the publishers clearly stated Notices of undeliverable mail occur from time to time For an address that is valid typically a 4 hour warning is to be expected But what about the following explanation from BPA s post office Subject Undeliverable Mail Date Sent Monday November 30 1998 12 14PM The following recipient s could not be reached Gabor Furst SMTP gfurst istar ca That user s mailbox is busy Who has ever heard of a mailbox being busy All IEEE papers are available from IEEE according to Marjan Popov who on November 17 wrote public E mail from TU Delft in The Netherlands You can visit http www ieee org power There you can find information about ordering of all IEEE publications from conferences transactions and proceedings The World Wide Web and high energy physics is the title of a story by Bebo White in the November 1998 issue of Physics Today a publication of the American Institute of Physics This article traces present WWW use from the Internet which was begun in the late 1960s by the US Department of Defense which wanted to develop a communications network that could survive a nuclear war In particular messages had to be able to reach their destinations even if significant communication routes were destroyed The solution the pa
26. anta Fe Argentina In E mail dated October 2 Mr Hevia reported The case is an HFS with a filter for the 7 harmonic isolated by a set of switches with T close 1 0 Yes gtpplot fails The Salford and gnu djgpp versions produce the same result If you delete the 7 harmonic filter branches the problem disappears I think that the same error is in the LIS file The code was corrected later that same day and an unexcited branch was added to the 3 subcase of DCNEW 21 to make sure such branches continue to be handled correctly in the future The trouble was traced to conflicting demands of the new output by columns like the time step loop as opposed to the original early 70s output that was limited to branch flows Today one can have both of course The original program output preceded not only all frequency scans but also the use of computer monitors at BPA Real paper was being printed and branches that carried zero current represented considerable waste So such output was suppressed While no problem for the table of branch flows positive KSSOUT this omission upset recent tabular output of the LIS file and PL4 files of all types the error was in fact independent of plot file type There is not yet any provision for the omission of zero signals of the LIS or PL4 files If any reader can imagine a workable procedure the idea will be considered Meanwhile FS and HFS solutions will disable the long establ
27. appear after the blank card ending node voltage and possibly other output requests and before any request for statistical tabulation Provided the user has not disabled plotting with negative IPLOT the interpretation of plot cards will be seen in the LIS file following the output of extrema If the user has requested a PRINTER PLOT this output too will be seen in the LIS file But for the CALCOMP PLOT alternative nothing more will be seen on the LIS file although plots will be seen on the screen if BOTH is used For the user who disables plotting nothing of plot cards or the blank card terminating them will be seen in the LIS file For an illustration see the ever changing DCNEW 25 Use of PCVP for frequency scans began September 12 This was in response to Gabor Furst s reaction to the preceding use for time domain simulation Quoting from his E mail dated September 11 This raises an exciting possibility Why not use it in HFS We can then vary the value of a resistor as a function of harmonic number h using a user defined function The message from your Editor the following day began as follows J have put in about 6 hours adapting PCVP to loop over frequency I have my first results and will attach them Note that I vary both frequency and inductance using the pocket calculator I do believe there is going to be yet another type of F scan for you to illustrate at your course in Prague DCNEW 26 entered the set of
28. ards Included is possible load flow initial condition and output variable data The new sorting logic requires that all data following miscellaneous data cards be sorted The user is not allowed to sort only part of his data For example note that just BRANCH and SWITCH declarations are used in DC 8 The new logic will fail on such partially sorted data The new logic requires that all data be sorted if any is sorted This is the key to simplicity which allows but does not require the omission of all blank cards Blank separator cards need not be supplied by the user because they will be generated internally by the program Such ATP generated lines will be easily recognized by their inimitable English language explanation NEWSORT and OLDSORT are requests in data for use of the new or the old sorting logic respectively The first of these is illustrated by a new 4 subcase of DCNEW 25 the former 4 of 4 subcases became the 5th of 5 on September 30 It is assumed SZBED is positive for correct handling of all test cases by RUN BAT This preserves continuity with the past It is the way STARTUP is being maintained by the user group no change For additional explanation see comments in the new data which is a modification of Prof Martinezs FILE6 DAT illustration NORUN is the switch that allows the user to inspect data after sorting recall This was described in a story in the January 1995 newsletter Use is highly recomm
29. are taken and these approximate a continuation from the sinusoidal steady state All modes are strongly excited operation is not balanced in order to be sure that nothing is overlooked Smoothness of the solution is documented on comment cards for those who might not otherwise know the answer All linear branches were interpreted the same way prior to improvement on November 24 Now transmission lines are distinguished by their own interpretation that includes the display of the fourth previously missing floating point number the imaginary part of I mk See DCNEW 27 output for many illustrations of this improvement JMarti frequency dependence was not originally provided with manual ac phasor initialization it is curious to note Instead there was manual dc zero frequency initialization Rationale behind this peculiarity has long been forgotten and dc initialization never was used as far as developers in Portland know or can remember If any reader knows otherwise or would like to make a case for the importance of true dc as opposed to low frequency ac initialization it is requested that he do so using the Fargo list server For background of first use consult page IONM 1 onward of the Vol XI EMTP Memorandum dated 3 October 1981 Meanwhile the dc capability has been changed to ac and first illustrations of DCNEW 27 see subcases 3 and 4 for single phase and 3 phase respectively involve use at the power frequency But lat
30. ational approximations The test code we use from Argonne National Labs shows an average error of 1 bit or less There aren t many values that you can give the cosine function that have an exact result in any case but 0 0 certainly is one of those few As to why this created an ATP problem optimal encoding was involved Curiously this number that is one bit short of unity failed to satisfy the D9 LT 1 0 test that dates to the beginning of ATP A correction from Mr Cook involves one trivial instruction no arithmetic and should be independent of computer This was installed in FLTOPT on December 11 Comings and Goings The University of Pittsburgh in Pennsylvania is one place where EMTP has been used for decades The Rule Book dated September 1980 had a PDP 10 page near the beginning Although only the computer specialist Prof J Tom Cain was mentioned on that page x k Prof R Gerald Jerry Colclaser seemed to be the main engineering user After 15 or more years of non contact E mail from Jerry Colclaser was received September 15 This referred to a graduate student in my power system analysis class I introduce the students to ATP and would like to get the latest version As has happened so many times during recent weeks BPA s Dr Tsu huei Liu responded by supplying login information for those password protected Japanese storage sites on the Internet John Kappenman remains in Duluth but no longer is an employee o
31. cket switching network was designed so that the particular route a message takes to reach its destination is unimportant The original Internet was named Arpanet after its sponsor the Advanced Research Projects Agency But that was E mail and TCP IP for file transfer It did not provide hyperlinks for documentation This was a European extension that came from the physics establishment rather than the defense establishment CERN located on the Franco Swiss border near Geneva is one of the most successful high energy laboratories in the world The subtitle of the article explains The Web what you may now be using to buy an airline ticket or look up Miles Davis s discography began life as a tool for improving the flow of information at CERN Fig 1 shows Tim Berners Lee the Oxford educated physicist who invented the World Wide Web Fig 3 shows the actual NeXT computer that ran the original World Wide Web server and browser at CERN This dates to November of 1990 It is quite a story and another good illustration of how necessity drives invention In the beginning was chaos of information exchange Unlicensed subscribers to the Fargo list server remain a concern as this issue goes to press during mid December Look for details in the following issue Watcom ATP for MS Windows The DIR command in response to the opening prompt was mentioned in the first Salford story of this issue Upon consulting th
32. commercial exploitation should have closed So according to this line of thinking secrecy could be relaxed But there is another viewpoint Once a free compiler is available what user needs to see TAL It will be used internally but no longer will be required externally Should user supplied TAL survive availability of the free compiler Is the burden of compilation large enough so that users will want to save the assembler Answers are not obvious The Type 68 Meredith hysteresis device was described in the April 1998 issue It was not supported by compiled TACS until September 28 however when operation first was tested New disk file DCN25A DAT was added to the set of compiled TACS illustrations to ensure that Meredith s model will not be overlooked again Prior to modification of the code the Salford compiler warned that variable A was undefined For completeness DCN210 DAT is a second disk file that was added at the same time In that execution halts prior to entry into the time step loop compiled TACS really is not involved and little is demonstrated But TACS is involved so DCN210O the 15th subcase of DCNEW 21 corresponds to letter O was added just to prove that it too can be handled without difficulty About AUTS Meredith s benchmark mentioned in the April newsletter compiled TACS did not make a great deal of difference Meredith s Type 68 device already had saved most of the inefficiency Following that reform
33. d October 2 was a 4 page report entitled Report of the recent modifications of Aki Cable Parameters June to Oct 1998 Near the top of page 3 is the section heading 5 Mutual impedance between overhead and underground cables At the end of the section Prof Ametani wrote J am going to present the above modification and calculated results in comparison with the accurate result at the EEUG98 meeting in Prague The Czech Republic The following modification was entitled 6 Accurate formula of space impedance self term The professor may now be a dean see the July 1998 issue but he seems able to remain productive in the area of research that made him internationally famous Higher Order Pi Circuits Real data for 400 coupled coils were received from Dr Hiroshi Arita of Hitachi Ltd on October 23 For back ground see the April 1998 issue Full size data was received in the form of HIT6 DAT which consisted of 31884 lines and required 2613 Kbytes to store Yet formatting continues to expand such data unnecessarily Dr Arita has not yet switched to the special C like alternative HIGH ORDER PI CIRCUIT or HOPC in abbreviated form he continues to use conventional Type 51 52 branches which have become overpowering There are 3736 branches of which only 400 are the famous high order Pi Next issue more statistics might be provided RENUMBER WITHOUT COUPLING RWC isa new request word that was introduced Octob
34. despite high harmonic content in a nonlinear reactor Actually there are two of them There are two independent parallel circuits with the Type 93 model used in one and the Type 98 model in the other Both produce identical answers Of course the reactors are not saturated saturation is a different problem that has nothing to do with multiple frequency excitation and is not helped by it A comment card was allowed immediately before a declaration of phasor superposition beginning November 20 Before that there was no problem if all comments were destroyed NOCOMM 1 But the user who retained his comments and placed one immediately before a declaration for superposition was treated badly ATP failed to recognize the declaration The burden on List 13 did not correctly reflect use of SUPERIMPOSE prior to correction on November 29 In fact usage was not being considered at all So if there was a plot it was only the burden of plotting the original meaning that was reported If there was no plot as for the 9 subcase of DCNEW 26 the undefined indicator 9999 was reported Following correction this changed to 24 An error in the initialization of JMarti line modeling for a second or later subcase that is unexcited was removed December 3 Although unrelated to use of superposition the error was discovered during such work For that single phase example in the 25 subcase of DC 9 answers changed only very slightly The fir
35. e corresponding Watcom code it was noted that DIR was not supported Yet GNU ATP for Linux offered directory output although not interactive selection and this seemed preferable to nothing So DIR output was added to Watcom ATP on October 8 Optimized compilation was described in the January 1998 issue which reported use of ot on 5 fpc fp5 with the final three of these a replacement for Robert Meredith s ox full optimization This resulted in just over 9 seconds for time spent inside the time step loop of DC 1 At the time this seemed good But then Meredith published his own benchmarks on October 8 and these were Clearly faster As reported in the October newsletter there was an increase in the local paging file size to 100 Mbytes This was the only known change since a year ago Anyway October 9 optimized compilation as recommended by Meredith using ox succeeded at BPA and all test cases except DC 56 were handled correctly by the resulting Watcom TPBIG But what about DC 56 Dr Tsu huei Liu had the idea to omit ox from just the compilation of that chunk of fortran that handles SPY HELP and this resulted in correction of the DC 56 solution This is the significance of new COMP6 BAT rather than the previous COMP5 as explained on rem DOS comment cards Protection against illegal arguments of library functions might be one reason to avoid Watcom compiler optimization BPA s Dr Tsu huei Liu located details a
36. ended if problems are suspected Remember that an ATP error message generally will be wrong if data sorting is wrong I e GIGO To understand what has happened the user should look at the output of sorting as a complete file Use of Initial Condition Cards A new KILL 237 error termination was added to prevent the manual specification of initial conditions for a constant parameter distributed line when two circumstances are satisfied 1 there is no phasor solution and 2 the simulation time TMAX is less than the travel time of the line Previously execution would continue but the solution was wrong because line history had been omitted in order to save space Since this logic has been in place for more than 25 years and probably forever it is being left unmodified But an error stop is being added November 20 to warn the user of the problem For more information about the manual specification of initial conditions see the separate story about INITIAL and data sorting by class JMarti line modeling did not handle manually supplied initial conditions properly in some cases Weakness of the logic which dates to the early 80s M35 idents was discovered October 22 while working on partial table dumping see separate story This provided initial inspiration for strenuous work on initial condition cards List 8 storage of past history has been saved in certain situations This first was demonstrated November 18 before sol
37. ent added 35 in taxes So Linux in Argentina is more expensive than in the USA but still a very good deal Consider the alternative from Bill G Compare with Windows Fortran Graphics package f About use by others J send the document Igtpplot doc in word 6 0 for Windows 3 1 format Please review the text I converted Word to PostScript and PostScript to PDF with very good results using Linux The PDF format is nearly universal Can Am EMTP News could be printed using Linux too WordPerfect for Linux was the subject of E mail from Masahiro Kan Dated October 27 this explained The following is an abstract from http www corel com Mmews 1998 octoberNinux htm Corel Announces No Charge Linux Download Special version of Corel WordPerfect8 for Linux Personal Edition to be available at no charge Information about the download version will be available from http linux corel com Interested users can preregister at this site for the application which is expected to be available in November Compiled TACS amp Assembly Language Secrecy surrounding TAL has not yet been relaxed There are two schools of thought and no final decision has yet been made about the question Recall that secrecy was declared in order to prevent commercial exploitation That was explained in the January 1998 issue before work on a compiler had begun Once the free compiler becomes available to the general public any window for
38. er one will find use of frequency 01 Hz in the 8 and 9 subcases and this should be good enough for engineering Who might disagree and why Multi frequency initialization began Thanksgiving Day For years see page MSPR 6 of the Vol XI EMTP Memorandum dated 5 August 1981 the phasor solution has allowed multiple frequencies provided each was isolated from all others in a disconnected subnetwork But there never was a complementary extension to manual initialization Today there is as first illustrated by the 6 subcase of DCNEW 27 Read data comments for details about how the program knows which branch is excited by which frequency an important detail obviously Noda frequency dependence is not involved in any of the extensions mentioned in this story unfortunately This is a consequence of the difference of language C rather than FORTRAN If manual specification of initial conditions later might be required for Noda modeling programming in C will be required This provides yet another illustration of the disadvantages of mixed language programming Today this complication is being avoided Partial Table Dumping Partial Table Dumping is an extension to Turbo Table Dumping by Robert Schultz of the New York City area For background on Schultz s quantum leap forward see newsletters beginning with the October 1993 issue Partial or selective table dumping began early in the morning of October 12 when MODELS tables we
39. er 24 in order to avoid the burden of transient renumbering Illustration began the following day with the addition of a new 5 subcase of DC 5 see comment cards for further explanation In case the reader does not understand the need consider preliminary results of that 400 coil example from Hitachi As reported by your Editor in E mail that same day the burden of renumbering is reflected in the second component of time Seconds for overlays 6 11 and this was overpowering 112 527 sec for your Editor s 133 MHz Pentium The preceding data input required only 15 seconds using LISTOFF to suppress nearly all output and the following preparation for the time step loop only 41 seconds Transient renumbering there was no phasor solution clearly was the bottleneck Using RWC the 112 sec was reduced to 3 132 seconds Furthermore there was no noticeable change in resulting simulation speed Storage of List 5 which includes triangularized Y was unchanged Typically it will change of course For normal data John Walker s Scheme 2 renumbering is of value If it is largely avoided simulation speed generally should suffer But not for the 400 coupled coils of HIT6 Why Perhaps each node already is connected to every other node so order of elimination might be immaterial About the burden on List 23 the saving is striking although small when viewed from the perspective of List 5 As reported in case summary statistics the figures
40. eractive Plotting Programs 0 ee eeeeeeeeeeeeee 18 Miscellaneous Intel PC Information 20 00 ee 19 Miscellaneous Small Items oo eeeeeeeeeeeeeseees 19 Authorized by Co chairmen Dr W Scott Meyer Editor Dr Tsu huei Liu E mail thliu bpa gov Vol 99 1 January 1999 Salford Compiler and DOS Extender READ PL4 HEADER was removed as a special request word for C like PL4 files on August 7 This was the three day Labor Day weekend of work preparing for GNU C like files when it was discovered that no standard test case DC DAT used the request So entire REQUEC was removed for Salford leaving an unsatisfied external The structure is being left to allow use by other compilers After all of these years it has been forgotten which compilers required the request word for which type of PL4 files and why If any DBOS user stumbles over the change he is asked to contact program developers with an explanation or illustration of his use It would be easy to restore the C like service and this will be done if need can be demonstrated soon For file types other than C like it should be mentioned that nothing has changed About illegal library function usage see story in the preceding issue no changes have been made to Salford EMTP Note that DBOS ver 3 5 provides a good error message It also documents the location subroutine name and hexadecimal address of the infraction and will provide a traceback of locations thr
41. ere deleted while many new ones were added The total number of SPY symbols increased from 820 to 877 But 20 separate from this update operation was corrected by the addition of spelling protection to SPYING earlier that same day After all even if the symbol list in SPY were complete there would remain the problem of misspelling by the user Zero LU6VRT will send some diagnostic output to the LIS file rather than the DBG file This was rediscovered again by mistake on November 22 The differences comparison of all DC LIS output with stored known solutions SAL were surprisingly large so were worked on for an hour or so In most cases changes were made to force diagnostic output back to the DBG file when LU6VRT is not positive i e when output is not being cached in RAM prior to being written to disk But in two cases non blank lines of production output that belonged in the LIS file had been hidden in DBG files The first was a warning message from Prof Hian Lauw about U M use in DCNEW 16 Warning Unbalanced electric power network elements or sources Second there was parameter tabulation of TEPCO s Type 58 S M optional output that follows the phasor solution The latter has been properly converted to lower case and removed from the code for reasons of multilinguality At the time of the original installation see the April 1997 issue your Editor was busy so he postponed the conversion Then appa
42. eries R L C branches But where do we go from here and why What are practical uses How much more effort is warranted to handle other branches What about nonlinear elements Your Editor had no shortage of questions DCNEW 26 has been given a new 5 subcase to document the first successful superposition which occurred November 8 ATP licensed readers who have interest are encouraged to study comments of this data and to verify smoothness of the solution The screen plot using Salford EMTP is perfectly repetitive and smooth and the HP GL output looks similarly perfect using WP 7 The LIS DAT and HPG files that document operation were sent to Prof Mustafa Kizilcay and Gabor Furst in Prague early the following morning This message had Subject Another new type of F scan Your Editor concluded as follows There are lots of questions that need to be thought through You Gabor and your short course seem like a good forum for the discussion Within each phasor solution two or more frequencies are allowed as long as they do not overlap This is no different than it was in years past It is the user s responsibility to ensure that subnetworks involving different frequencies are not connected The new SSOMIT switch see mention elsewhere in this issue might be of assistance in this regard Alternatively if there is a connection the two different frequencies now can be separated by placement in different phasor solutions The user
43. es For such data the correction of the preceding paragraph seemed to be correct But for multiple 2 or 4 output parts associated with either polar or rectangular variables or both the printed heading was in error This first was reported by Prof Kizilcay in an E mail message dated October 30 Printed variable headings but not the following variables themselves incorrectly were repeated for MODELS variables prior to correction later that same day Now each MODELS variable never has more than one part ABSOLUTE TACS DIMENSIONS allows the user to size TACS tables arbitrarily within the total available working space of VARDIM List 19 The fourth of eight table sizes is for the number of TACS sources and this is confused by the fact that 11 such sources are supplied internally and automatically Recall names such as TIMEX for simulation time and PI for the geometric constant The user s size is supposed to include space for these but what if he forgets and requests a number less than 11 It was found that ATP was protected provided data defined at least one source of its own Otherwise there was no protection and results were erroneous prior to the addition of special error checking to SUBR1 on October 30 Line and Cable Constants The mixture of overhead and underground cables is one of the latest innovations of Prof Akihiro Ametani of Doshisha University in Kyoto Japan Attached to a letter to BPA s Dr Tsu huei Liu date
44. f Minnesota Power This change was learned in E mail from jkappenma aol com dated November 2 Metatech Corporation is the new company name and circumstances are interesting J have recently left Minnesota Power for the world of consulting Metatech is a California firm with offices in Santa Barbara and Albequerque New Mexico They have about 20 full time employees in scientific and engineering areas with specialties in difficult electromagnetic coupling problems They were nice enough to allow me to open a Duluth office which will be developing products for power systems applications We are currently specializing in geomagnetic disturbances impacts on power systems and storm forecasting services We will also be working in the lightning area in which I will be 11 using ATP from time to time My office is only about 2 blocks west and one block north of the MP offices so no big change in that routine Florida Resort Course March 8 11 Prof Dennis Carroll s ATP short course in Florida is scheduled for March 8 11 with faculty unchanged from last year This according to E mail of the Fargo list server dated December 15 Listed faculty consist of Prof Dennis Carroll outside experts Drs Kurt Fehrle and Yin Yuexin and graduate student Carlos Mata Course location has changed drastically From the campus in Gainesville the course has been moved to the small resort town of Cedar Key which is located on the Gulf of Mexico
45. first to request such improvement of rules was Orlando Hevia of Universidad Tecnologica Nacional in Santa Fe Argentina although nothing was done until Massimo Ceraolo of the University of Pisa in Italy wrote with a question related to more work on his own plotting program PLOTXY see the July 1998 issue Year 2000 Compliance of ATP The alleged year 2000 problem popularly referred to as the Y2K problem first was mentioned in the October issue Since then several collaborators have reported nuisance inquiries that reveal the mindlessly stupid fashion with which some most bureaucracies seem to be responding to the alleged threat The humor continues At the risk of overlooking some good report several that come to mind will be summarized First Robert Meredith of the New York City area shared his own response to a Y2K inquiry in E mail dated October 5 An inquiry about Y2K compliance of Watcom ATP had been received from the headquarters building of Northern States Power Company in downtown Minneapolis Minnesota A questionaire was attached and prior to the Y2K questions there were others such as 1 Briefly what does the app do 2 Is it NT 4 0 compliant If not when is an NT 4 0 compliant version coming out What is the cost 3 How frequent are upgrades When is the next one 4 Is it network compliant 5 Is it meterable More tolerant than your Editor Mr Meredith did answer the NSP questions quickly He also co
46. generator model preceded the Type 59 model of Dr Vladimir Brandwajn recall Although the SCE code itself was removed by Dr Brandwajn during the early 80s a remaining trace has lingered on without anyone noticing or caring all of these years Your Editor is referring to dependent List Size 53 which sized compensation current vector CURSUB using both the number of nonlinear elements List 9 and the number of S M components List 17 But the Type 59 code does not use compensation as the SCE model once did On the other hand the U M does Rather than replace List 53 by List 9 your Editor decided to replace List 17 in the formula by an estimate of the number of U M machines based on default proportions This was done October 18 when it was assumed that one machine exists for each 112 cells of List 25 space see DC34 LIS which assumes the default value of NUMFIX 3 machines Several COMMON blocks were found to be unused so were removed This was the most surprising discovery during research about table dumping and restoring COMMON blocks that had not been used for years The heading that begins the LIS file changed as follows on October 17 from size of LABCOM tables 233316 words to 230308 for 3 times default dimensioning This may be the first decrease in history Evaluating Analytical Functions III Exponentiation is requested by the of FORTRAN and it was supposed to be a supported function of course But as Gabo
47. has complete control over which phasor solution involves which excitation Although most users probably will order the solutions so frequency increases in fact there is no such need Order is arbitrary except for one detail All phasor solutions preceding the last one are added to the last one for continuation into the time step loop As a result the final excitation must include sources for the time step loop All previous groups are used only for the associated phasor solutions The final group serves an additional function so is fundamentally different Perhaps it is more useful to have the final solution include the power frequency Additional storage is required for the new superposition and an existing working vector of List 13 has been used for this There should be protection against overflow of course although such trouble seems unlikely for networks that are not huge Storage does depend on the number of nodes and branches but thus far is independent of the number of harmonics or phasor solutions involved Integer miscellaneous data parameters IDOUBL and KSSOUT are important to the user of superposition About the former value zero will result in no output of the connectivity table of course On the other hand value unity will result in just one copy of the table which will be seen as part of the first phasor solution For the second and later solutions such constant output is suppressed automatically As for KSSOUT this is
48. hat might be useful to you But who might be willing to contribute data illustrating such ATP simulation
49. he problem is related only to the file size However if this is the result of your tests let me know Frequency spreading is the procedure that is used to determine possible overlap of phasor solutions having different frequencies The operation begins at sources and expands through branches and closed switches keeping track of which node is excited by which frequency If a collision overlap of two different frequencies ever might be noted ATP should halt with a KILL 222 error message That is the theory In practice the logic was imperfect prior to a correction in OVER8 on November 16 Any older program will not produce a correct solution to the 10 subcase of DCNEW 26 which was created at the same time Note clarification of the line that precedes the error message This is new SSOMIT is a special request that is the opposite of SSONLY see the October 1994 newsletter ATP recognition began November 6 following the composition of E mail to Prof Akihiro Ametani of Doshisha University in Kyoto Japan The combined use of 50 and 60 Hz in Japan had been mentioned as a problem in E mail dated November 1 and your Editor and Dr Liu proposed the following solution What you write is true any one subnetwork is limited to a single frequency Thus one can not mix 50 and 60 Hz However one subnetwork can have 50 Hz and the other 60 Hz Thyristors must not provide a connection of the two subsystems however With all thyris
50. ible the only thing you can do is to calculate H both ways and see which value is reasonable for the given machine In most cases you will find that it is the 1 37 multiplier which gives the correct conversion DEPOSIT was mentioned in the October 1995 newsletter and is illustrated by use near the top of DC 6 Operation remains generally correct it would seem since no user yet has complained However one concurrence of unlikely circumstances was discovered October 9 as Masahiro Kan s famous benchmark case TEST_ATP was modified by the addition of DEPOSIT use Dimensions already had been adjusted within data NEW LIST SIZES but that eliminated only one of two changes that once were required prior to timing The second was a change of KOMPAR from the usual value 4 for program developers to unity An attempt to add this assignment to data led to an infinite loop within the time step loop when storage of plot points within RAM 240K words had filled and SPY was being used to flush RAM to disk The trouble occurred because 1 SPY symbols did not include KOMPAR and 2 the plot cache had filled The first problem was solved October 10 when SPY symbols were updated for the first time in years the defective storage dates to October of 1996 Although work was limited to EQUIVALENCE statements of BLKCOM many changes were made Included should be most names that have been mentioned in newsletters the past 2 or 3 years Some old symbols w
51. is not obvious that others have the knowledge Partial table dumping prohibits the application of Schultz s logic to those portions of ATP tables that are known not to be in use Although details are secret the 18 concept is easily understood Why dump and restore what is not being used Not only is this demanding of memory it also wastes time unnecessary indexing in RAM may be very fast but is not free It is more efficient to apply the programmer s knowledge to omit such unused regions of memory from consideration Upon reading of the idea pioneering table archivist Schultz seemed highly pleased On October 12 he wrote I just received and read your email Bravo Clean and simply based on using one s brain The only pitfall might be the special exceptions you mentioned in which case Incidentally I got to touch an actual jet fighter engine at the San Diego Aerospace Museum fully equipped with afterburners I have pictures of it at home Nice experience PARTIAL TABLE DUMPING PTD is a new special request word that is required for use of the new logic Initially the user is being forced to declare his desire explicitly This is to prevent unintended use while debugging is underway Later if and when usage might be understood to be extremely reliable the need for explicit declaration could be removed One might later add a switch to STARTUP Only time and lots of practice by small burrowing animals guinea
52. is response the following day author Hoidalen confirmed trouble handling such details ATPDraw handles statistical output rather poorly Only output of node voltages is possible with no BASE control Special request STATISTICAL DATA is not supported It looks like the user has to change the ATP 13 file manually I have just looked into the problem of statistical and systematic switching since I m introducing the concept of a master and slave switches However at this moment I see no immediate solution to the statistical output problem I am planning to release a new version of ATPDraw around the EEUG meeting in Prague This new version 1 5 will remove several bugs and introduce some more general components smoother handling of MODELS direct execution of ATP etc Too long a constant parameter distributed transmission line was a problem for Watcom ATP prior to the addition of special protection on November 19 Without any indication of location ERR KO 06 integer overflow was the uncivilized message that terminated program execution This was seen by BPA s Jules Esztergalyos after phasor branch flows so he submitted the case for analysis Dr Tsu huei Liu provided explanation later that same day The problem is with the branch cards contained in your data The first line mode has a travel time of 6173 sec Divided by a step size of 1 E 6 the 32 bit integer limit of 2 048E6 was exceeded Execution still ends
53. ished omission of output of unexcited branches If new DCNEW 21 is used with an old program execution will die in plotting due to the inconsistent PL4 file Also the column heading of NONE TERRA will have no numbers below it except for extrema which come from different code Frequency dependence of saturable TRANSFORMER models first was demonstrated by Orlando Hevia of Universidad Tecnologica Nacional in Santa Fe Argentina His E mail dated November 13 had Subject Funk Hantel functions for parameters It explained I tried a case with the Funk Hantel functions for transformer parameters resistance and inductance I applied the parameters to the TRANSFORMER fields directly not to external branches I changed only the primary branches The case ran OK The transformer parameters can be varied as a function of frequency in a very easy manner This PCVP is a wonderful tool if one knows how to use it Quotation marks may be used to distinguish EMTP symbols from data symbols defined by the user This was an early decision associated with PARAMETER data of the Powell pocket calculator as explained in the April 1997 issue Quotation marks should continue to be honored but are they necessary Your Editor thinks not Although some risk of false identification is involved ATP should be capable of telling the difference between the user s data symbol names and EMTP symbol names All user symbol names will be searched first
54. lot because some time values were negative To ensure continued correct operation MRVIC2 DAT was added as a new 9th subcase of DCNEW 24 Parameter Variation Studies by PCVP POCKET CALCULATOR VARIES PARAMETERS PCVP is a request word that will result in a family of simulations For background see the beginning of this story in the preceding issue The feature evolves Statistical tabulation of extrema is a concept familiar to those who perform Monte Carlo simulation As illustrated by standard test cases DC 24 and DC 40 extrema not only can be tabulated extrema also can be saved on disk for later combination and tabulation Well this concept of STATISTICS studies was extended to PCVP studies on August 29 and the 4 subcase of DCNEW 25 was enhanced to illustrate such output There are no new rules of data input since all the old ones of STATISTICS use continue to apply Included among alternatives are the use of FIND and EXCLUDE to locate extrema that are not the worst the only ones revealed by the extrema of extrema output Batch mode plotting was added to the PCVP data structure on August 30 How practical or popular this will prove to be is not known however One problem is that ATP screen graphics can not be scrolled today as they could using Apollo workstations of a decade ago In any case graphics now exist giving data the same structure as a STATISTICS data case that includes a base case solution Plot cards are to
55. lowing web site http Nitree ing unlp edu ar estudios caue revista htm This issue closes the 1998 volume also called Volume 0 which has been edited without a fixed periodicity As usual the new issue includes three papers which deal with different ATP applications Editors started this business with the aim of learning from experience From the next volume on we hope four issues will be released every year on March June September and December As mentioned in the previous message those interested in publishing their work can submit their documents in either PDF format or in a popular text processor A file with Instructions for Authors can be downloaded from the web site Indian ATP use remains confused with the latest indication being an E mail message to Prof Bruce Mork of Michigan Tech Dated December 3 Ashok Maitra of B E College D U B Garden Howrah 711 103 wrote from a Hotmail address He began his inquiry about hysteresis modeling as follows Our institute and I myself are members of Indian EMTP User Group We are also licensed user of EMTP software We sent drafts to Mr I S JHA Secretary Indian EMTP User Group on 03 8 98 for renewal of our membership for 1998 and to obtain password for accessing ATP software through Internet but no acknowledgement or reply so far been received in this respect Prof Mork responded with a brief reference to Prof M V Hariharan in Bombay see the Januay
56. mong disk files that accompany the ver 11 compiler One bit of writing has headline What you should know about optimization The first paragraph contains the following warning when you are using the OM option to generate inline math functions no argument validation will be done for the intrinsic math functions such as sin or cos Consider the needs of your application and select the optimization options that best meet your requirements The 11 and final subcase of DC 59 illustrated two dimensional parameter variation with PCVP providing one of the dimensions and LINE CONSTANTS itself providing the other frequency This was handled correctly by Salford EMTP when data was in a separate disk file but it was wrong as positioned in DC 59 The trouble first was recognized during Watcom ATP testing of the same code For Watcom there was no mistaking the symptoms of trouble execution died when first tested October 26 Later that same day MAINOO was found to be missing the cancellation of MAXKNT which indicates a PCVP study in this case involving a supporting program Two Type 59 S M models are connected to TACS in large standard test case DC 47 and this simulation suddenly failed for Watcom during the final week of October Whether the cause was a change of compiler options that use of ox for optimization or the removal of unused COMMON blocks see story about Schultz s afterburner is not known In any case Wa
57. n GNU C The following day your Editor concluded Why did I do the work If you can satisfy the Salford library routines we do not need my work Too bad I did not see this earlier Yes if you can supply Salford like routines we can use Salford ATP FORTRAN We do not need the neutral interface that I wrote Yes we can do away with the extra layer of interface This is perfect I like your idea very much I think Schultz would be proud of you This is more than a convenient i e easier way to proceed this isa better way to handle the problem Somehow I was tied to Szymanski s procedure of a decade ago Of course that was before Salford Today Szymanski probably would do exactly what you have proposed Mr Kan worked quickly attaching his work to E mail dated September 13 I wrote Salford compatible C like functions in GNU C I also attach testclk f and makefile which are used to test these functions After verification using Mr Kan s small test program GNU installation dependent modules were modified by the addition of Salford like code to handle C like files Then came verification of test cases changing all usage from UNFORMATTED to C like where appropriate The job was completed without the discovery of a single error in Mr Kan s code as best your Editor can recall To conclude GNU ATP today allows C like PL4 files as demonstrated by any of the numerous standard test cases GTPPLOT is Orlando Hevia s inter
58. ncluded with a summary great writing The program is developed entirely by volunteer effort and the user assumes all risk of using it If you find it unsuitable for your use do not use it It is probably Y2K compliant since it does not use dates It does calculate run time statistics by comparing elapsed times since 1970 measured in hundredths or thousandths of seconds I am not personally aware of when that counter may become a problem or whether it might overflow into anything significant The program will undoubtedly be revised thousands of times before then anyhow so its current status is somewhat irrelevant If it breaks it gets fixed Excuse me if this sounds blunt But if your first question is What does the program do you are obviously not a licensed user of the program and may not be entitled to any more detailed response I am forwarding your note to the user group however Perhaps they wish to respond further The user group s answer no Laszlo Prikler of the Technical University of Budapest shared his own response to Claudio Hernandez in E mail dated November 12 About your Editor s mention of DBOS Prof Prikler advised If you are using Watcom ATP under NT the hypothetical problem with alphabetical sorting of DBOS mentioned in October issue of the Can Am EMTP News does not effect you But why Your Editor responded as follows on November 15 The DOS in a DOS window of NT is different Maybe it shows 1998 for
59. ng variable of substitution So if one intermediate variable is used by two or more variables of substitution it must be repeated before each Later if this is found to result in substantial repetition the rule might be modified But for now it is convenient As usual the capability evolves bottom up design as with most creative engineering An induction motor model prompted an increase in the number of intermediate variables from 15 the original limit to 25 This was a couple of days later It did not take long for limits of the preceding paragraph to be recognized as inadequate Writing to Orlando Hevia and Gabor Furst on October 20 your Editor finally could report success handling Mr Hevia s induction motor data I have been debugging my way through the pocket calculator I modified the code to preserve intermediate variables after each substitution variable Preceding intermediate variables remain defined Next I found lack of protection against overflow of temporary variables This morning I allowed only 15 it seemed like a lot at the time clearly I had not anticipated heavy duty usage such as yours So I added protection against overflow at the same time I expanded the storage to 25 Next I found that xxxxxx was an assembly language verb produced by my compiler But it was missing in the code to execute So I now add the missing execution code no problem once the failing is recognized Good de
60. nyone who answers is sticking a neck out to be sued More great writing Unfortunately there can be no attribution this contributor too prefers to remain anonymous Data Sorting by Class ATP data sorting by class first was reported to be incompatible with ATPDRAW created data on September 28 In his E mail complaint Prof Juan A Martinez Velasco of the Polytechnic University of Catalunya in Barcelona Spain provided several examples illustrating trouble using INITIAL for the specification of initial conditions It is also obvious that we know how to solve this But the main problem is when we use ATPDRAW Then the file is created by the graphical preprocessor Of course a hand modification or some other tricks can always solve this problem Beginning September 30 SZBED is used as a flag to choose between old sorting logic and new simplified sorting logic If the user does nothing his data will be sorted by the old logic of course But if a minus sign is appended to normally positive SZBED new sorting logic will be used It really is a question of old and new rather than right or wrong Each has its own philosophy and different limitations Neither will handle all possibilities correctly The new is offered as an easy way to solve the problem posed by INITIAL and any other class of data that has no blank card associated with it This is what complicated the old logic data between the blank card ending sources and plot c
61. of Mustafa Kizilcay of FH Osnabrueck in Germany For background on LMFS in ATP see the January and April 1992 issues of this newsletter Two sorts of unintended and unprotected limits were exceeded by Prof Kizilcay s data which used 100 points per decade This was an industrial strength scan never before attempted it would seem The two limits exceeded for the first time were for 1 local storage of floating point numbers and 2 rows of the LMFS comparison table Changes were confined to module LMFS11 with local storage replaced by the List 28 floating point storage of MODELS which henceforth is prohibited to LMFS data It is unlikely that this large cache 126K words for LISTSIZE BPA ever will fill but if it does an error message and program halt should protect the user As for rows of the output table they now are unlimited The warning message about nonlinear elements being either linearized for inductors or ignored for resistors during a phasor solution should be familiar to all Designed to precede the output of a single phasor solution it was found to be highly disruptive and annoying as part of repeated output of FREQUENCY SCAN or HARMONIC FREQUENCY SCAN The problem first was reported by Prof Mustafa Kizilcay So for these cases FS and HFS the message was suppressed on October 30 The output of current in an unexcited branch was reported to be missing by Orlando Hevia of Universidad Tecnologica Nacional in S
62. on to use the new command Sometimes it is easy to forget limitations of a mouse which are severe mouse clicks are interactive so are a pain for repeated operations For example if one wanted to process a family of 100 PL4 files typically one would bury his plot commands within a batch file and then execute this using the command Whereas mouse clicks could not be placed in such a batch file the new AUTO SCALE command could be The EXTREMA display was tied to the SMOOTH tolerance in a curious way prior to correction on November 7 This was the result of work with power and energy following an inquiry from Dr Peter Naumovich of Cooper Power Systems in Olean located about 100 miles south of Buffalo New York Before Cooper the company name was McGraw Edison Energy is unusual in that it is monotone non decreasing Furthermore for typical real signals values in jules are much smaller than power in watts or voltage in volts or current in amperes As a result the energy signal might become lost among the larger signals Of course it is common for the maximum to occur at the final point and for the user to begin with unity scaling assuming 2 or more curves If this happened and if all previous points had been discarded as being visually redundant using the initial default scaling of unity the final point was not being tested for extrema So both the minimum and the maximum were zero and the curve plotted as identically zero
63. ong time and for all computers Watcom ATP died during the execution and this was traced to subscripting in SSTACT that was erroneous The connection to the Type 59 S M was involved and all computers previously seem to have been using a zero subscript which not surprisingly accessed a zero value But suddenly Watcom accessed garbage Whether this was due to a change of compiler options that use of ox for optimization or the removal of unused COMMON blocks see story about Schultz s afterburner is not known MODELS use with FREQUENCY SCAN is illustrated by the 3rd subcase of DC 52 and the loop heading should reflect this usage For each MODELS variable the six letters MODELS should be seen in the first name of the pair of identifying names Unfortunately this was not the case prior to correction on October 27 In E mail earlier that same day Prof Mustafa Kizilcay of FH Osnabrueck in Germany alerted developers in Portland to the problem He was using different data COPENH4 DAT but similarity to DC 52 was unmistakable The problem was not noticed in DC 52 output because branch flows are requested and as aresult the column heading was inconspicuous Not so for Prof Kizilcay s data For the record only 2 lines differ Incorrect lines of DC52 LIS Step F Hz GEN LOAD GENIC Corresponding correct lines Step F Hz GEN LOAD MODELS VRE1 The FREQUENCY SCAN data of DC 52 produces output of just variable magnitud
64. onic nor equal to power freq Has anyone else had a problem with this It does seem that some sort of frequency scan was a necessary ingredient for such trouble As discovered later that same day standard DC 51 itself was not troubled Lack of BNDC at the end was a prerequisite for trouble Since all standard test cases are properly terminated testing in Portland did not reveal the weakness Well October 10 protection was added one line initializing MAXCRD immediately before DO 1758 in OVERI This modification permitted ATP to execute correctly using Mr Durbak s data Complexity of data gathering and validation were illustrated by a public E mail exchange on October 25 First Kenji Takahashi of Toshiba Corporation in Japan posed a question about the induction motor constant H as used in Gabor Furst s supporting program INDMOT see April 1997 issue There seemed to be ambiguity two equations were given and the confusion was confirmed by author Furst later that same day He explained In dynamic calculations the value of H and the value of WR 2 are unique they are calculated the same way by all manufacturers in the world Some European manufacturers calculate GD 2_ differently with one value being twice the other I too have come across this inconsistency which is the reason why I did not put GD 2 as an option into INDMOT Your best bet is to go back to the manufacturer and ask them for the H value If this is not poss
65. or 29 5 Mbytes RAM 14 inch monitor 24X CD ROM 2 1 Gbyte hard drive sound card and speakers 33 6 modem and MS style keyboard Amazing The 500 Internet computer has arrived and it is not a specialized limited appliance at all Rather it is a full power Pentium based PC About other prices a 17 inch 28 mm Philips Magnavox color monitor costs only 289 if purchased with a computer that has none 309 otherwise A 19 inch version is 529 549 otherwise How about a refurbished Epson Stylus 400 color printer for 100 Or a 4 3 Gbyte Seagate hard disk for 147 Computers and their parts are not worth much these days For more information the catalog prominently recommends that a reader Visit our new Web site at www mei microcenter com Miscellaneous Small Items Massimo Ceraolo of the University of Pisa in Italy continues to provide exemplary support for PL42MAT the utility that will convert PL4 files to MATLAB format The subject arose September 29 when Adeoti Adediran of Texas A amp M University in College Station reported trouble during heavy duty use The size of the files is about 20 Meg and the number of variables is above 90 Mr Ceraolo s advice the following day included the following Do you use the latest version of the program 19 3 la You can find it at the ATP ftp sites For large files it is highly advisable to use wpl42mat exe instead of pl42mat exe I would be very much surprised knowing that t
66. ough calling subroutines if the user presses the down arrow key In a word Salford DBOS ver 3 5 performance is exemplary Even the performance of ver 2 66 dating to April of 1992 was exemplary Time and exposure to newer competing products only makes the old Salford compiler look better lt Loaded from command line DISK gt illustrates the first of three new lines of output to the screen that confirm parameters of batch mode RUNTP like usage The idea comes from Robert Schultz and Robert Meredith of the New York City area Quoting from Meredith s E mail dated September 20 this is done in order that users do not have to wonder whether to respond or not to the last of the prompts that sit on the screen while the case runs Send desired disk file name for LUNIT6 lt CR gt R The preceding two prompts typically pass so quickly the user has no time to think of course It is after the third prompt for the output file name that nearly all of the time of a typical simulation is spent The new output has been a Watcom ATP feature for some time Schultz s initial s and the date 4 12 96 are seen on the line of code It became a Salford feature on September 23 in order that Watcom could share module CIMAG4 with Salford This eliminated the necessity of maintaining a second copy of what has grown to hundreds of lines The DIR command was a victim of the reform that allowed longer file names see story in preceding issue
67. pigs will tell To be continued next time A lot more already has been done and written but publication space is in short supply Interactive Plotting Programs Availability of GITPPLOT was announced by Dr Tsuyoshi Funaki of Osaka University in Japan New Gtpplot is uploaded was the Subject of his public E mail dated November 3 The message was noteworthy because of the inclusion of free Unix The new GTPPLOT is released from Mr Hevia This time the Linux version is added Also a 32 bit version for WinNT is released from Mr Kan All of them are placed at the password protected server http atp pwr eng osaka u ac jp support common gtpplot Shaft fatigue estimation with GITPPLOT was the Subject of public E mail dated October 3 In this author Orlando Hevia of Universidad Tecnologica Nacional in Santa Fe Argentina explained that The procedure with GTPPLOT is very similar to TPPLOT But the method is different The method is described in 1 an IPST97 paper by Argentine authors J A Nizovoy J L Alonso A C Alvarez and L M Bouyssede The authors of this paper developed an auxiliary program named PERVIDSM to estimate the fatigue reading the torques from the lis file the ATP output of each case The program was compiled with constant parameters corresponding to the machine studied and was verified with a commercial program of a European consulting firm GTPPLOT reads the torque values froma
68. plied for checking of other more popular data A sawtooth waveform on the electrical side i e not in TACS or MODELS was the contribution of Orlando Hevia His E mail dated November 10 began as follows I send a file with a sawtooth source For your Editor this was a complete surprise Mr Hevia used the pocket calculator to define a Type 10 source as first mentioned in the October issue That is rather than use sawtooth logic that is built into the program Mr Hevia supplies his own analytically in data To be sure that this creative idea is not forgotten Mr Hevia s centered waveform has been added to the Ist subcase of DC 22 Might there be more such contributions Maybe Mr Hevia wrote I think that with some work a rectangular pulse train as TACS 23 source could be simulated The numerical interface of the pocket calculator was changed November 11 in order to remove a conflict with other sources This was found to be necessary for the just mentioned modification of DC 22 Surprisingly two signals of one other standard test case changed in the process node voltage COS and associated current COS TERRA of the Ist subcase of DCNEW 19 Clearly the signals have been improved previously they were wrong after the first time step Continuation lines of FORTRAN are distinguished by blank columns 1 through 5 and nonblank column 6 This capability was added to your Editors compiler on November 14 in order to solve a p
69. r Furst observed in E mail dated September 30 exponent value 2 was handled correctly whereas any other value seemed to result in a temporary error termination Apparently the general case never had been tested Once existence of the trouble was known correction was found to be easy enough the following morning Exponentiation A B was not properly handled for negative A prior to October 12 Working with Gabor Furst who had interest in the subject Orlando Hevia of Universidad Tecnologica Nacional in Santa Fe Argentina wrote that your Editor s assurances were theoretical rather than actual Yes but this is in FORTRAN not in PCVP I tried 1 KNT and ATP aborted with some error message like real exponent for negative base from DBOS So your Editor modified logic to check for non negative A Should A be negative ATP will continue with special logic provided B is an integer If B is not an integer there will be an error message PARAMETER definitions were limited to a single line for each variable that was being replaced This was prior to October 17 when intermediate variables first were allowed in response to a request from Mr Furst Now an arbitrary number of intermediate variables is allowed permitting much more complicated formulas For an illustration involving one intermediate variable see the 5 subcase of DCNEW 25 The user is warned however that all intermediate variables are erased upon use of the followi
70. re the first to be ignored for any data case that involved no MODELS usage This was quickly followed by the addition of logic for TACS the U M etc In colloquial terms partial table dumping is being referred to as Schultz s afterburner Schultz handled all variably dimensioned COMMON blocks and therein lay the problem as program dimensions increased without obvious bound Recall the writing of Robert Meredith in the preceding issue LABCOM for me is 3639004 which covers more than 14 Mbytes Meredith also advocated use of maximum dimensions the removal of LISTSIZE DAT which means that the entire 14 Mbytes would be handled by Schultz This would be very slow for any computer having 16 Mbytes or less of RAM Storage that for smaller dimensions could be confined to RAM would spill over onto disk there would be massive paging This also might happen on a computer with much more memory e g 32 or 64 Mbytes if Bill G s top of the line MS Windows NT is being used Recall the summary of Daniel Durbak s temporary problems in the preceding issue In an attempt to speed execution the memory of his PC was doubled to 64 Mbytes but even this was less than satisfying As your Editor wrote in public E mail dated October 8 The issue is not how much RAM the computer has but rather how much RAM Bill G is allowing the process of interest here Watcom or GNU ATP to use Maybe Meredith knows enough to control his resources under WinNT It
71. rently he forgot about the need for more work The output of DCNEW 20 now is substantially larger readers will find SMOOTH SATURATION USING TANH was less than properly illustrated by the final subcase of DC 13 prior to correction on November 14 As reported the preceding day by Orlando Hevia of Universidad Tecnologica Nacional in Santa Fe Argentina labeling of the FLTINF value is truncated on the right This is due to the truncation request of the preceding subcase So full normal 132 column output was requested at the start of Mr Hevia s data in order to cancel the truncation The computation of power and energy requested by a 4 punch in column 80 always has been noted in the heading of printed output of the time step loop the LIS file There were two extra lines of output if applicable But missing prior to October 25 was any indication of the number of variables involved This was a trivial but informative recent addition For example see DC 16 A permanent magnet synchronous motor driven by an inverter was of interest to Andrea Moratto who submitted public E mail from itric mail abb com Does anyone have suggestions for modelling of this kind of motor he asked This was October 23 Later that same day Prof Chee Mun Ong of Purdue University responded with a reference to his book J have something on that in Chaps 7 and 10 of Dynamic Simulation of Electric Machinery Prentice Hall 1998 ISBN 0 13 723785 5 t
72. roblem posed by Orlando Hevia His E mail the preceding day explained J tried the pulse source but I can t reduce the number of lines to two First the Type 10 analytical source allowed just columns 11 60 Then when more space was required CONT in columns 61 65 was provided to indicate the desire for a single continuation line But even this was not enough to satisfy Mr Hevia s creative mind it would seem So expansion to an indefinite number of lines now is being provided This is independent of CONT which will continue to be honored for historical reasons Yet it is not required FORTRAN like continuation is being allowed and is illustrated in the 3 subcase of DCNEW 19 a double dollar sign was selected to mark an intermediate variable within PARAMETER data By definition an intermediate variable is one that is used to define later data symbols but is not itself a data symbol Such a symbol will not be searched for and replaced in data cards that follow the PARAMETER data structure then Illustration is provided in the 5 subcase of DCNEW 25 although this use is trivial and artificial The need came from Orlando Hevia during mid October as he was constructing a frequency dependent model of an induction motor for Gabor Furst s HFS use In a disk file named GABMOT DAT this data uses 20 intermediate variables followed by 5 data symbols Sample lines are scheduled for illustration in the associated Rule Book section
73. s not even know if there will be a problem So he is easily frightened into purchasing a Y2K compliant upgrade Question 4 Advice on how to test if needed for the year 2000 problem Your Editor s reaction These people do not know how to set the calendar ahead on their computers Meredith should be in the Y2K consulting business Sellers are not the only ones who exploit Y2K hysteria however Some probably many buyers are profiting from the opportunity as well Y2K readiness seems to be a 16 nearly perfect justification just one step below national security Consider the following explanation which was received November 19 from a source who probably would prefer anonymity Sometimes there is a few minutes delay before the processing is done Sun6 is only a sparcl I m getting a replacement server for Y2K readiness that will be much faster and have more memory and disk space I m also getting version 5 5 of Framemaker so it is Y2K ready I also bought X designer version 5 to put on Verdi for Y2K readiness That way we can say that the Y2K ready package is in place If XXXX doesn t use it that s his Y2K problem Stupid hardly begins to cover it Company policy is to ask all vendors whether their stuff is Y2K compliant file the response for future lawsuits then ignore the answer and try to test it This will be the ambulance chasing lawyers full employment act for the next ten years A
74. st output variable is a line voltage and it now has maximum 713185 9 compare with 713186 0 yesterday A change to SUBR1 provided correction Stu Cook Uses Apple Macintosh STATUS OLD was added to the OPEN used by INSERT as a result of Absoft testing Previously there was no protection against a missing file In effect without an explicit STATUS declaration UNKNOWN was assumed and this resulted in successful opening of an empty new file rather than an error as intended Other compilers may well behave comparably thus far the statement remains useful So the change was made to CIMAGE on December 11 as changes from Mr Cook were processed in Portland This prefaced retranslation Absoft is the first compiler not to return exactly unity as the cosine of zero Apparently all other compilers of current usage treat zero as a known special case thereby avoiding possible numerical error Absoft does not as explained in E mail from Absoft Tech Support dated December 8 Mr Cook had inquired about the problem as follows the DCOS function doesn t return a true value of unity Rather the value returned is one bit less than unity The following program demonstrates this The Absoft response was both helpful and reasonable The PowerPC doesnt have built in hardware transcendental functions DCOS is calculated by a software routine supplied in the runtime library The runtime library SIN COS routines are implemented as r
75. tcom ATP died in SSTACT and this was traced to subscripting of data that was not properly initialized So simple avoidance was applied on October 30 The patch is used by all computers apparently to good effect Watcom ATP execution has been restored to normal within roundoff differences and the Salford and GNU solutions are unchanged UNFORMATTED PL4 files might never be created by the average Watcom ATP user because C like are available and C like offer the advantage of portability As distributed from Portland Watcom STARTUP requests C like But the UNFORMATTED alternative is available and was the choice of Dan Durbak of Power Technologies PTI in Schenectady New York He uses his own plotting program and the logic of this assumed UNFORMATTED PL4 files He also uses a different compiler Surprisingly Mr Durbak found compatibility for small files But for large files e g DC1 PL4 he found that Watcom ATP was adding unexpected end of record marks Presumably Watcom ATP would read these correctly e g if used with REPLOT but his non Watcom compiler did not prior to mandatory expansion of the record size to 32 Kbytes on November 12 This is progress in that the previously unknown and unpredictable has become predictable News About TACS and MODELS TACS and Type 59 S M modeling form a dominant part of large DC 47 It was during the final week of October that the solution was found to have been incorrect fora l
76. tors open during the phasor solution this should not be a problem except for leakage paths as are provided by snubber circuits Yes snubbers do create a complication but this could be avoided in one of two ways First one might be able to ignore snubbers only during the phasor solution by using small enough TOLMAT Yet this is a little risky one would be disabling his natural protection A better approach might be for us to ignore certain branches during the phasor solution For an illustration see the new 4 and 5 subcases of DC 10 One more cell of List 7 storage is required following the modification to recognize SSOMIT BEGIN NEW DATA CASE or BNDC in abbreviated form is supposed to appear at the bottom of a data case as a bound This is recommended practice Among other virtues BNDC at the end allows recovery from KILL code errors providing a location for the execution of a following data subset to begin But many program users do not bother with such niceties and one finally was burned badly due to a recent program addition the HFS code Daniel Durbak of PTI Power Technologies Inc in Schenectady New York made public his wound in mail of the Fargo list server dated October 9 None of the three versions are working at all for me when I attempt to run a good old FREQUENCY SCAN like DC51I DAT My simulation ends early with the following message Halt in OVER4 Minimum Type 14 source frequency is neither 1 harm
77. utions were correct Using manual specification of initial conditions fora JMARTI line Dr Tsu huei Liu showed a decrease of List 8 from 82 cells to 41 cells Frequency in Hz may be specified on the right of the first 17 node voltage card of initial conditions In the absence of a phasor solution the program had no other way of knowing the frequency of the solution That dates back to year one before ATP But ATP does have another way variable STATFR of STARTUP So if frequency as read from the first node voltage is not positive the value of STATFR will be assumed Interpretation of the first Type 2 node voltage card will show this value if and only if the frequency field of columns 39 53 is non blank not non zero That is the LIS file of test cases will be unchanged Several involve initial condition cards but only for lumped elements which do not require frequency Output of these remains unchanged DCNEW 27 is a new data case that is being devoted to the verification of initial condition cards for different distributed parameter transmission line models in their most basic form without the complication of card sorting for example This began November 22 following success using a single phase distributed model Comments of the data describe significance as complexity is increased from single to multi phase and from constant parameter to Jmarti and Semlyen models In all cases just a few typically 20 time steps
78. utput and data variables and suggest a support file on the correct format This new module of ATPDraw is activated if the user selects a file with extension MOD under MODELS in the Component selection menu Version 1 2 of ATPDraw also supports RECORD of model variables Direct execution of ATP was removed from ATPDraw when introducing the Windows version 1 0 This was due to possible conflicts with DBOS Now this feature has been reintroduced in a new and powerful way The user is allowed to specify several programs to execute and what to send as parameter when calling these programs All the older circuit objects of version 1 0 are supported in the new version but some of them have been removed from the Component selection menu and replaced by other more general components The general 3 phase transformer object supports both 2 and 3 windings three types of coupling for each winding Y D lead and D lag and the option to switch between high or low homopolar reluctance TRANSFORMER or TRANSFORMER THREE PHASE by clicking on a 3 leg button The icon of the transformer object adapts the user settings The TACS transfer function object supports both fixed and named limits For the new _ Statistic systematic switches the concept of independent master slave has been introduced The Type94 MODELS objects are handled in a very special way Harmonic Frequency Scan is now supported by ATPDraw Some new load models have been

Can / Am EMTP News - the European EMTP

Contents

Download Pdf Manuals

Related Search

Related Contents