From: Michael C. <mi...@fl...> - 2018-03-22 06:43:31
|
I've been able to use the optimization feature quite effectively. However, I've been having some trouble using the "E24 series" type when defining optimization variables. Here's how I've tried to set up such a variable: Name: Lseries Initial: 10e-9 Min: 1e-9 Max: 100e-9 However, when I do this, I get an optimization error: initialize.c - scaleto -- Initial value in *.cfg is larger than the maximum > value If I look in "asco_netlist.cfg" as suggested by the error, then I see: Parameter 1:#Lseries#:10e-9:1e-9:100e-9:E24:OPT It looks like the initial value here is just fine, so I assume I am somehow using the "E24 values" feature incorrectly. Can someone tell me what I'm doing wrong here? I am currently using "Version 0.0.19 (9aa60413f)", but have encountered the same problem with the stable "Version 0.0.19". Thanks for your help! |
From: Claudio G. <cla...@vi...> - 2018-03-22 07:22:44
|
Hello, I think I've seen something similar before and concluded it was due to some incorrect rounding in the external optimizer, ASCO. I always used the E-series optimization feature with values greater than 1, I assume it should work also for the small values you are using but maybe this is triggering the rounding issue I mentioned. As a workaround you may try to set the optimization limits from 1 to 100 and scale the optimization variables in a separate equation. If you can share the failing schematic I can try to take a closer look. Regards, Claudio > Il 22 marzo 2018 alle 7.43 Michael Cooper <mi...@fl...> ha scritto: > > > I've been able to use the optimization feature quite effectively. However, > I've been having some trouble using the "E24 series" type when defining > optimization variables. Here's how I've tried to set up such a variable: > > Name: Lseries > Initial: 10e-9 > Min: 1e-9 > Max: 100e-9 > > > However, when I do this, I get an optimization error: > > initialize.c - scaleto -- Initial value in *.cfg is larger than the maximum > > value > > > If I look in "asco_netlist.cfg" as suggested by the error, then I see: > > Parameter 1:#Lseries#:10e-9:1e-9:100e-9:E24:OPT > > > It looks like the initial value here is just fine, so I assume I am somehow > using the "E24 values" feature incorrectly. Can someone tell me what I'm > doing wrong here? > > I am currently using "Version 0.0.19 (9aa60413f)", but have encountered the > same problem with the stable "Version 0.0.19". > > Thanks for your help! > ------------------------------------------------------------------------------ > Check out the vibrant tech community on one of the world's most > engaging tech sites, Slashdot.org! http://sdm.link/slashdot > _______________________________________________ > Qucs-help mailing list > Quc...@li... > https://lists.sourceforge.net/lists/listinfo/qucs-help |
From: Michael C. <mi...@fl...> - 2018-03-22 15:33:32
|
Thanks! That worked perfectly. On Thu, Mar 22, 2018 at 1:22 AM, Claudio Girardi via Qucs-help < quc...@li...> wrote: > Hello, > I think I've seen something similar before and concluded it was due to > some incorrect rounding in the external optimizer, ASCO. I always used the > E-series optimization feature with values greater than 1, I assume it > should work also for the small values you are using but maybe this is > triggering the rounding issue I mentioned. As a workaround you may try to > set the optimization limits from 1 to 100 and scale the optimization > variables in a separate equation. > If you can share the failing schematic I can try to take a closer look. > > Regards, > Claudio > > > Il 22 marzo 2018 alle 7.43 Michael Cooper <mi...@fl...> ha > scritto: > > > > > > I've been able to use the optimization feature quite effectively. > However, > > I've been having some trouble using the "E24 series" type when defining > > optimization variables. Here's how I've tried to set up such a variable: > > > > Name: Lseries > > Initial: 10e-9 > > Min: 1e-9 > > Max: 100e-9 > > > > > > However, when I do this, I get an optimization error: > > > > initialize.c - scaleto -- Initial value in *.cfg is larger than the > maximum > > > value > > > > > > If I look in "asco_netlist.cfg" as suggested by the error, then I see: > > > > Parameter 1:#Lseries#:10e-9:1e-9:100e-9:E24:OPT > > > > > > It looks like the initial value here is just fine, so I assume I am > somehow > > using the "E24 values" feature incorrectly. Can someone tell me what I'm > > doing wrong here? > > > > I am currently using "Version 0.0.19 (9aa60413f)", but have encountered > the > > same problem with the stable "Version 0.0.19". > > > > Thanks for your help! > > ------------------------------------------------------------ > ------------------ > > Check out the vibrant tech community on one of the world's most > > engaging tech sites, Slashdot.org! http://sdm.link/slashdot > > _______________________________________________ > > Qucs-help mailing list > > Quc...@li... > > https://lists.sourceforge.net/lists/listinfo/qucs-help > > ------------------------------------------------------------ > ------------------ > Check out the vibrant tech community on one of the world's most > engaging tech sites, Slashdot.org! http://sdm.link/slashdot > _______________________________________________ > Qucs-help mailing list > Quc...@li... > https://lists.sourceforge.net/lists/listinfo/qucs-help > |
From: Michael C. <mi...@fl...> - 2018-03-23 05:54:53
|
A related question: For a couple of days, I've been working on building impedance matching networks using Qucs. To do this, I model my transmission lines and matching components then use optimization to find a set of component values which results in the desired impedance, filter characteristics, etc. At first, I did this optimization using ideal matching components. Once I had a good result with ideal components, I would swap them out for 2-port models of actual components with those values to get a more precise idea of the matching network's attributes. However, I noticed there was a fairly large change when I did this. To mitigate this problem, what I've done is to create a parametric model of the actual components, so that, e.g., I can say fairly accurately what the parasitic capacitance of a 4.7 nH inductor will be. This results in a much smaller change when I introduce the 2-port models, but there is still some change. Ideally, I'd like to skip the middle step and have the optimizer use the correct s2p file for each component value. This way, the optimizer would actually select a set of s2p files (one for each component) which results in the best matching network. Alas, I haven't been able to find a way to do this in Qucs. I am a programmer, so I've subscribed to the qucs-devel list as well with the notion of adding this functionality if it is not already present. However, before I get ahead of myself, I thought I would post here to see if this can already be accomplished somehow in Qucs. Can anyone think of a way? On Thu, Mar 22, 2018 at 9:33 AM, Michael Cooper <mi...@fl...> wrote: > Thanks! That worked perfectly. > > On Thu, Mar 22, 2018 at 1:22 AM, Claudio Girardi via Qucs-help < > quc...@li...> wrote: > >> Hello, >> I think I've seen something similar before and concluded it was due to >> some incorrect rounding in the external optimizer, ASCO. I always used the >> E-series optimization feature with values greater than 1, I assume it >> should work also for the small values you are using but maybe this is >> triggering the rounding issue I mentioned. As a workaround you may try to >> set the optimization limits from 1 to 100 and scale the optimization >> variables in a separate equation. >> If you can share the failing schematic I can try to take a closer look. >> >> Regards, >> Claudio >> >> > Il 22 marzo 2018 alle 7.43 Michael Cooper <mi...@fl...> ha >> scritto: >> > >> > >> > I've been able to use the optimization feature quite effectively. >> However, >> > I've been having some trouble using the "E24 series" type when defining >> > optimization variables. Here's how I've tried to set up such a variable: >> > >> > Name: Lseries >> > Initial: 10e-9 >> > Min: 1e-9 >> > Max: 100e-9 >> > >> > >> > However, when I do this, I get an optimization error: >> > >> > initialize.c - scaleto -- Initial value in *.cfg is larger than the >> maximum >> > > value >> > >> > >> > If I look in "asco_netlist.cfg" as suggested by the error, then I see: >> > >> > Parameter 1:#Lseries#:10e-9:1e-9:100e-9:E24:OPT >> > >> > >> > It looks like the initial value here is just fine, so I assume I am >> somehow >> > using the "E24 values" feature incorrectly. Can someone tell me what I'm >> > doing wrong here? >> > >> > I am currently using "Version 0.0.19 (9aa60413f)", but have encountered >> the >> > same problem with the stable "Version 0.0.19". >> > >> > Thanks for your help! >> > ------------------------------------------------------------ >> ------------------ >> > Check out the vibrant tech community on one of the world's most >> > engaging tech sites, Slashdot.org! http://sdm.link/slashdot >> > _______________________________________________ >> > Qucs-help mailing list >> > Quc...@li... >> > https://lists.sourceforge.net/lists/listinfo/qucs-help >> >> ------------------------------------------------------------ >> ------------------ >> Check out the vibrant tech community on one of the world's most >> engaging tech sites, Slashdot.org! http://sdm.link/slashdot >> _______________________________________________ >> Qucs-help mailing list >> Quc...@li... >> https://lists.sourceforge.net/lists/listinfo/qucs-help >> > > |
From: Claudio G. <cla...@vi...> - 2018-03-27 19:58:21
|
Hello, interesting problem, not sure there is an easy solution currently. One, convoluted, possibility could be to substitute qucsator, the Qucs simulator called by the ASCO optimizer, with a script that actually modifies the netlist to use the right S-parameters file before calling the real qucsator. You could have in the schematic a dummy component, whose value ASCO tries to optimize, the script then will read the component value assigned during the current simulation run and based on this rewrite the netlist with the corresponding S-parameters file assigned to the component of interest and the run qucsator. ASCO will then read the simulation output and continue as usual. Hope the explanation makes sense - I did not actually try this myself. In the long term the best way would of course be to ask the ASCO http://asco.sourceforge.net/ author to support this general discrete optimization feature. Regards, Claudio > Il 23 marzo 2018 alle 6.54 Michael Cooper <mi...@fl...> ha scritto: > > A related question: For a couple of days, I've been working on building impedance matching networks using Qucs. To do this, I model my transmission lines and matching components then use optimization to find a set of component values which results in the desired impedance, filter characteristics, etc. > > At first, I did this optimization using ideal matching components. Once I had a good result with ideal components, I would swap them out for 2-port models of actual components with those values to get a more precise idea of the matching network's attributes. However, I noticed there was a fairly large change when I did this. > > To mitigate this problem, what I've done is to create a parametric model of the actual components, so that, e.g., I can say fairly accurately what the parasitic capacitance of a 4.7 nH inductor will be. This results in a much smaller change when I introduce the 2-port models, but there is still some change. > > Ideally, I'd like to skip the middle step and have the optimizer use the correct s2p file for each component value. This way, the optimizer would actually select a set of s2p files (one for each component) which results in the best matching network. Alas, I haven't been able to find a way to do this in Qucs. > > I am a programmer, so I've subscribed to the qucs-devel list as well with the notion of adding this functionality if it is not already present. However, before I get ahead of myself, I thought I would post here to see if this can already be accomplished somehow in Qucs. Can anyone think of a way? > > On Thu, Mar 22, 2018 at 9:33 AM, Michael Cooper <mi...@fl... mailto:mi...@fl... > wrote: > > > > Thanks! That worked perfectly. > > > > On Thu, Mar 22, 2018 at 1:22 AM, Claudio Girardi via Qucs-help <quc...@li... mailto:quc...@li... > wrote: > > > > > > > Hello, > > > I think I've seen something similar before and concluded it was due to some incorrect rounding in the external optimizer, ASCO. I always used the E-series optimization feature with values greater than 1, I assume it should work also for the small values you are using but maybe this is triggering the rounding issue I mentioned. As a workaround you may try to set the optimization limits from 1 to 100 and scale the optimization variables in a separate equation. > > > If you can share the failing schematic I can try to take a closer look. > > > > > > Regards, > > > Claudio > > > > > > > Il 22 marzo 2018 alle 7.43 Michael Cooper < mi...@fl... mailto:mi...@fl... > ha scritto: > > > > > > > > > > > > I've been able to use the optimization feature quite effectively. However, > > > > I've been having some trouble using the "E24 series" type when defining > > > > optimization variables. Here's how I've tried to set up such a variable: > > > > > > > > Name: Lseries > > > > Initial: 10e-9 > > > > Min: 1e-9 > > > > Max: 100e-9 > > > > > > > > > > > > However, when I do this, I get an optimization error: > > > > > > > > initialize.c - scaleto -- Initial value in *.cfg is larger than the maximum > > > > > value > > > > > > > > > > > > If I look in "asco_netlist.cfg" as suggested by the error, then I see: > > > > > > > > Parameter 1:#Lseries#:10e-9:1e-9:100e-9: E24:OPT > > > > > > > > > > > > It looks like the initial value here is just fine, so I assume I am somehow > > > > using the "E24 values" feature incorrectly. Can someone tell me what I'm > > > > doing wrong here? > > > > > > > > I am currently using "Version 0.0.19 (9aa60413f)", but have encountered the > > > > same problem with the stable "Version 0.0.19". > > > > > > > > Thanks for your help! > > > > ------------------------------ ------------------------------ ------------------ > > > > Check out the vibrant tech community on one of the world's most > > > > engaging tech sites, Slashdot.org! http://sdm.link/slashdot > > > > ______________________________ _________________ > > > > Qucs-help mailing list > > > > Quc...@li... mailto:Quc...@li... > > > > https://lists.sourceforge.net/lists/listinfo/qucs-help https://lists.sourceforge.net/lists/listinfo/qucs-help > > > > > > ------------------------------ ------------------------------ ------------------ > > > Check out the vibrant tech community on one of the world's most > > > engaging tech sites, Slashdot.org! http://sdm.link/slashdot > > > ______________________________ _________________ > > > Qucs-help mailing list > > > Quc...@li... mailto:Quc...@li... > > > https://lists.sourceforge.net/lists/listinfo/qucs-help https://lists.sourceforge.net/lists/listinfo/qucs-help > > > > > > > > > > > > > > |
From: Michael C. <mi...@fl...> - 2018-03-30 05:38:33
|
Hi Claudio-- Looking at the source code in a little more detail, I believe the flow of execution for optimizations goes something like this: 1. Qucs prepares a netlist with optimization variables identified with hash marks, e.g., L="#Length#". 2. Qucs calls ASCO, passing it the name of the netlist file as well as a file for simulation output. 3. To evaluate the fitness of a given model, ASCO builds a new netlist using the original as a template but replacing the marked values. Then it calls qucsator to simulate that netlist and parses. Finally, it parses the output file to determine the error value. 4. Once ASCO is done, the best values are stored in an output file and the program ends. Since ASCO is called only once, and doesn't return until optimization is complete, I'm inclined to agree that the best way to tackle this may be through ASCO, either by replacing qucsator as you suggest or, in the longer term, making the required changes in ASCO. ASCO currently has an "RF module" which allows the user to specify an equivalent circuit for resistors, capacitors and inductors, with the values of the individual components defined as a function of the parent's value. That might be a good place to introduce something like this. I'll report back if I'm able to replace qucsator with a script to accomplish this job. On Tue, Mar 27, 2018 at 1:44 PM, Claudio Girardi via Qucs-help < quc...@li...> wrote: > Hello, > > interesting problem, not sure there is an easy solution currently. > > One, convoluted, possibility could be to substitute qucsator, the Qucs > simulator called by the ASCO optimizer, with a script that actually > modifies the netlist to use the right S-parameters file before calling the > real qucsator. You could have in the schematic a dummy component, whose > value ASCO tries to optimize, the script then will read the component value > assigned during the current simulation run and based on this rewrite the > netlist with the corresponding S-parameters file assigned to the component > of interest and the run qucsator. ASCO will then read the simulation output > and continue as usual. > > Hope the explanation makes sense - I did not actually try this myself. > > > In the long term the best way would of course be to ask the ASCO > http://asco.sourceforge.net/ author to support this general discrete > optimization feature. > > > Regards, > > Claudio > > > Il 23 marzo 2018 alle 6.54 Michael Cooper <mi...@fl...> ha > scritto: > > > > A related question: For a couple of days, I've been working on > building impedance matching networks using Qucs. To do this, I model my > transmission lines and matching components then use optimization to find a > set of component values which results in the desired impedance, filter > characteristics, etc. > > > > At first, I did this optimization using ideal matching components. > Once I had a good result with ideal components, I would swap them out for > 2-port models of actual components with those values to get a more precise > idea of the matching network's attributes. However, I noticed there was a > fairly large change when I did this. > > > > To mitigate this problem, what I've done is to create a parametric > model of the actual components, so that, e.g., I can say fairly accurately > what the parasitic capacitance of a 4.7 nH inductor will be. This results > in a much smaller change when I introduce the 2-port models, but there is > still some change. > > > > Ideally, I'd like to skip the middle step and have the optimizer use > the correct s2p file for each component value. This way, the optimizer > would actually select a set of s2p files (one for each component) which > results in the best matching network. Alas, I haven't been able to find a > way to do this in Qucs. > > > > I am a programmer, so I've subscribed to the qucs-devel list as well > with the notion of adding this functionality if it is not already present. > However, before I get ahead of myself, I thought I would post here to see > if this can already be accomplished somehow in Qucs. Can anyone think of a > way? > > > > On Thu, Mar 22, 2018 at 9:33 AM, Michael Cooper <mi...@fl... > mailto:mi...@fl... > wrote: > > > > > > Thanks! That worked perfectly. > > > > > > On Thu, Mar 22, 2018 at 1:22 AM, Claudio Girardi via Qucs-help > <quc...@li... mailto:quc...@li... > > wrote: > > > > > > > > > Hello, > > > > I think I've seen something similar before and concluded > it was due to some incorrect rounding in the external optimizer, ASCO. I > always used the E-series optimization feature with values greater than 1, I > assume it should work also for the small values you are using but maybe > this is triggering the rounding issue I mentioned. As a workaround you may > try to set the optimization limits from 1 to 100 and scale the optimization > variables in a separate equation. > > > > If you can share the failing schematic I can try to take > a closer look. > > > > > > > > Regards, > > > > Claudio > > > > > > > > > Il 22 marzo 2018 alle 7.43 Michael Cooper < > mi...@fl... mailto:mi...@fl... > ha scritto: > > > > > > > > > > > > > > > I've been able to use the optimization feature quite > effectively. However, > > > > > I've been having some trouble using the "E24 series" > type when defining > > > > > optimization variables. Here's how I've tried to set > up such a variable: > > > > > > > > > > Name: Lseries > > > > > Initial: 10e-9 > > > > > Min: 1e-9 > > > > > Max: 100e-9 > > > > > > > > > > > > > > > However, when I do this, I get an optimization error: > > > > > > > > > > initialize.c - scaleto -- Initial value in *.cfg is > larger than the maximum > > > > > > value > > > > > > > > > > > > > > > If I look in "asco_netlist.cfg" as suggested by the > error, then I see: > > > > > > > > > > Parameter 1:#Lseries#:10e-9:1e-9:100e-9: E24:OPT > > > > > > > > > > > > > > > It looks like the initial value here is just fine, so > I assume I am somehow > > > > > using the "E24 values" feature incorrectly. Can > someone tell me what I'm > > > > > doing wrong here? > > > > > > > > > > I am currently using "Version 0.0.19 (9aa60413f)", but > have encountered the > > > > > same problem with the stable "Version 0.0.19". > > > > > > > > > > Thanks for your help! > > > > > ------------------------------ > ------------------------------ ------------------ > > > > > Check out the vibrant tech community on one of the > world's most > > > > > engaging tech sites, Slashdot.org! > http://sdm.link/slashdot > > > > > ______________________________ _________________ > > > > > Qucs-help mailing list > > > > > Quc...@li... mailto: > Quc...@li... > > > > > https://lists.sourceforge.net/lists/listinfo/qucs-help > https://lists.sourceforge.net/lists/listinfo/qucs-help > > > > > > > > ------------------------------ > ------------------------------ ------------------ > > > > Check out the vibrant tech community on one of the > world's most > > > > engaging tech sites, Slashdot.org! > http://sdm.link/slashdot > > > > ______________________________ _________________ > > > > Qucs-help mailing list > > > > Quc...@li... mailto:Qucs-help@lists. > sourceforge.net > > > > https://lists.sourceforge.net/lists/listinfo/qucs-help > https://lists.sourceforge.net/lists/listinfo/qucs-help > > > > > > > > > > > > > > > > > > > > ------------------------------------------------------------ > ------------------ > Check out the vibrant tech community on one of the world's most > engaging tech sites, Slashdot.org! http://sdm.link/slashdot > _______________________________________________ > Qucs-help mailing list > Quc...@li... > https://lists.sourceforge.net/lists/listinfo/qucs-help > |
From: Michael C. <mi...@fl...> - 2018-03-30 05:41:47
|
Hi guys-- I think I've just found a better way to do this without creating a script. What I noticed is that ASCO uses the hash marks to identify optimization variables. So, I wondered, what would happen if I used the same hash marks in my file name? Here is an example schematic I created: <Qucs Schematic 0.0.19> <Components> <.SP SP1 1 80 80 0 63 0 0 "lin" 1 "3 MHz" 1 "6 GHz" 1 "201" 1 "no" 0 "1" 0 "2" 0 "no" 0 "no" 0> <.Opt Opt1 1 300 80 0 39 0 0 "Sim=SP1" 0 "DE=3|50|2|20|0.85|1|3|1e-6|10|100" 0 "Var=Integer|yes|8.000000E+000|0|10|LIN_INT" 0> <Pac P1 1 80 350 18 -26 0 1 "1" 1 "50 Ohm" 1 "0 dBm" 0 "1 GHz" 0 "26.85" 0> <GND * 1 80 380 0 0 0 0> <GND * 1 190 380 0 0 0 0> <C C1 1 300 350 17 -26 0 1 "1 pF" 1 "" 0 "neutral" 0> <GND * 1 300 380 0 0 0 0> <SPfile X1 1 190 320 -26 -59 0 0 "#Integer#.s2p" 1 "rectangular" 0 "linear" 0 "open" 0 "2" 0> </Components> <Wires> <190 350 190 380 "" 0 0 0 ""> <80 320 160 320 "" 0 0 0 ""> <220 320 300 320 "" 0 0 0 ""> </Wires> <Diagrams> </Diagrams> <Paintings> </Paintings> What I've done here is to create a 2-port S-parameter component whose File is set to "#Integer#.s2p". Then I added an optimization variable called "Integer" with linear integer type and range 0 to 10. When I run the simulation, I get an error, since I haven't actually created any S-parameter files with the appropriate names. However, the error text reads: error loading `D:/QUCS/ASCO Test_prj/8.000000E+000.s2p': No such file or directory This makes me think the ASCO optimizer has replaced #Integer# with the integer value, represented in scientific notation, in the file name as well. All that would be needed is to copy/rename the manufacturer's S-parameter files to this naming convention. On Thu, Mar 29, 2018 at 11:14 PM, Michael Cooper <mi...@fl...> wrote: > Hi Claudio-- > > Looking at the source code in a little more detail, I believe the flow of > execution for optimizations goes something like this: > > 1. Qucs prepares a netlist with optimization variables identified with > hash marks, e.g., L="#Length#". > 2. Qucs calls ASCO, passing it the name of the netlist file as well as > a file for simulation output. > 3. To evaluate the fitness of a given model, ASCO builds a new netlist > using the original as a template but replacing the marked values. Then it > calls qucsator to simulate that netlist and parses. Finally, it parses the > output file to determine the error value. > 4. Once ASCO is done, the best values are stored in an output file and > the program ends. > > Since ASCO is called only once, and doesn't return until optimization is > complete, I'm inclined to agree that the best way to tackle this may be > through ASCO, either by replacing qucsator as you suggest or, in the longer > term, making the required changes in ASCO. ASCO currently has an "RF > module" which allows the user to specify an equivalent circuit for > resistors, capacitors and inductors, with the values of the individual > components defined as a function of the parent's value. That might be a > good place to introduce something like this. > > I'll report back if I'm able to replace qucsator with a script to > accomplish this job. > > On Tue, Mar 27, 2018 at 1:44 PM, Claudio Girardi via Qucs-help < > quc...@li...> wrote: > >> Hello, >> >> interesting problem, not sure there is an easy solution currently. >> >> One, convoluted, possibility could be to substitute qucsator, the Qucs >> simulator called by the ASCO optimizer, with a script that actually >> modifies the netlist to use the right S-parameters file before calling the >> real qucsator. You could have in the schematic a dummy component, whose >> value ASCO tries to optimize, the script then will read the component value >> assigned during the current simulation run and based on this rewrite the >> netlist with the corresponding S-parameters file assigned to the component >> of interest and the run qucsator. ASCO will then read the simulation output >> and continue as usual. >> >> Hope the explanation makes sense - I did not actually try this myself. >> >> >> In the long term the best way would of course be to ask the ASCO >> http://asco.sourceforge.net/ author to support this general discrete >> optimization feature. >> >> >> Regards, >> >> Claudio >> >> > Il 23 marzo 2018 alle 6.54 Michael Cooper <mi...@fl...> ha >> scritto: >> > >> > A related question: For a couple of days, I've been working on >> building impedance matching networks using Qucs. To do this, I model my >> transmission lines and matching components then use optimization to find a >> set of component values which results in the desired impedance, filter >> characteristics, etc. >> > >> > At first, I did this optimization using ideal matching components. >> Once I had a good result with ideal components, I would swap them out for >> 2-port models of actual components with those values to get a more precise >> idea of the matching network's attributes. However, I noticed there was a >> fairly large change when I did this. >> > >> > To mitigate this problem, what I've done is to create a parametric >> model of the actual components, so that, e.g., I can say fairly accurately >> what the parasitic capacitance of a 4.7 nH inductor will be. This results >> in a much smaller change when I introduce the 2-port models, but there is >> still some change. >> > >> > Ideally, I'd like to skip the middle step and have the optimizer >> use the correct s2p file for each component value. This way, the optimizer >> would actually select a set of s2p files (one for each component) which >> results in the best matching network. Alas, I haven't been able to find a >> way to do this in Qucs. >> > >> > I am a programmer, so I've subscribed to the qucs-devel list as >> well with the notion of adding this functionality if it is not already >> present. However, before I get ahead of myself, I thought I would post here >> to see if this can already be accomplished somehow in Qucs. Can anyone >> think of a way? >> > >> > On Thu, Mar 22, 2018 at 9:33 AM, Michael Cooper < >> mi...@fl... mailto:mi...@fl... > wrote: >> > >> > > > Thanks! That worked perfectly. >> > > >> > > On Thu, Mar 22, 2018 at 1:22 AM, Claudio Girardi via >> Qucs-help <quc...@li... mailto:quc...@li...urcef >> orge.net > wrote: >> > > >> > > > > > Hello, >> > > > I think I've seen something similar before and >> concluded it was due to some incorrect rounding in the external optimizer, >> ASCO. I always used the E-series optimization feature with values greater >> than 1, I assume it should work also for the small values you are using but >> maybe this is triggering the rounding issue I mentioned. As a workaround >> you may try to set the optimization limits from 1 to 100 and scale the >> optimization variables in a separate equation. >> > > > If you can share the failing schematic I can try to >> take a closer look. >> > > > >> > > > Regards, >> > > > Claudio >> > > > >> > > > > Il 22 marzo 2018 alle 7.43 Michael Cooper < >> mi...@fl... mailto:mi...@fl... > ha scritto: >> > > > > >> > > > > >> > > > > I've been able to use the optimization feature quite >> effectively. However, >> > > > > I've been having some trouble using the "E24 series" >> type when defining >> > > > > optimization variables. Here's how I've tried to set >> up such a variable: >> > > > > >> > > > > Name: Lseries >> > > > > Initial: 10e-9 >> > > > > Min: 1e-9 >> > > > > Max: 100e-9 >> > > > > >> > > > > >> > > > > However, when I do this, I get an optimization error: >> > > > > >> > > > > initialize.c - scaleto -- Initial value in *.cfg is >> larger than the maximum >> > > > > > value >> > > > > >> > > > > >> > > > > If I look in "asco_netlist.cfg" as suggested by the >> error, then I see: >> > > > > >> > > > > Parameter 1:#Lseries#:10e-9:1e-9:100e-9: E24:OPT >> > > > > >> > > > > >> > > > > It looks like the initial value here is just fine, so >> I assume I am somehow >> > > > > using the "E24 values" feature incorrectly. Can >> someone tell me what I'm >> > > > > doing wrong here? >> > > > > >> > > > > I am currently using "Version 0.0.19 (9aa60413f)", >> but have encountered the >> > > > > same problem with the stable "Version 0.0.19". >> > > > > >> > > > > Thanks for your help! >> > > > > ------------------------------ >> ------------------------------ ------------------ >> > > > > Check out the vibrant tech community on one of the >> world's most >> > > > > engaging tech sites, Slashdot.org! >> http://sdm.link/slashdot >> > > > > ______________________________ _________________ >> > > > > Qucs-help mailing list >> > > > > Quc...@li... mailto: >> Quc...@li... >> > > > > https://lists.sourceforge.net/ >> lists/listinfo/qucs-help https://lists.sourceforge.net/ >> lists/listinfo/qucs-help >> > > > >> > > > ------------------------------ >> ------------------------------ ------------------ >> > > > Check out the vibrant tech community on one of the >> world's most >> > > > engaging tech sites, Slashdot.org! >> http://sdm.link/slashdot >> > > > ______________________________ _________________ >> > > > Qucs-help mailing list >> > > > Quc...@li... mailto: >> Quc...@li... >> > > > https://lists.sourceforge.net/lists/listinfo/qucs-help >> https://lists.sourceforge.net/lists/listinfo/qucs-help >> > > > >> > > > > > >> > > >> > > > >> > >> ------------------------------------------------------------ >> ------------------ >> Check out the vibrant tech community on one of the world's most >> engaging tech sites, Slashdot.org! http://sdm.link/slashdot >> _______________________________________________ >> Qucs-help mailing list >> Quc...@li... >> https://lists.sourceforge.net/lists/listinfo/qucs-help >> > > |
From: Claudio G. <cla...@vi...> - 2018-03-30 07:34:38
|
ah, you may be onto something here. Seems you were able to turn a (sort of) bug into a feature :) . Let us know if you are able to make it work this way. Regards, Claudio > Il 30 marzo 2018 alle 7.41 Michael Cooper <mi...@fl...> ha scritto: > > Hi guys-- > > I think I've just found a better way to do this without creating a script. What I noticed is that ASCO uses the hash marks to identify optimization variables. So, I wondered, what would happen if I used the same hash marks in my file name? Here is an example schematic I created: > > > > > <Qucs Schematic 0.0.19> > > <Components> > > <.SP SP1 1 80 80 0 63 0 0 "lin" 1 "3 MHz" 1 "6 GHz" 1 "201" 1 "no" 0 "1" 0 "2" 0 "no" 0 "no" 0> > > <.Opt Opt1 1 300 80 0 39 0 0 "Sim=SP1" 0 "DE=3|50|2|20|0.85|1|3|1e-6|10|100" 0 "Var=Integer|yes|8.000000E+000|0|10|LIN_INT" 0> > > <Pac P1 1 80 350 18 -26 0 1 "1" 1 "50 Ohm" 1 "0 dBm" 0 "1 GHz" 0 "26.85" 0> > > <GND * 1 80 380 0 0 0 0> > > <GND * 1 190 380 0 0 0 0> > > <C C1 1 300 350 17 -26 0 1 "1 pF" 1 "" 0 "neutral" 0> > > <GND * 1 300 380 0 0 0 0> > > <SPfile X1 1 190 320 -26 -59 0 0 "#Integer#.s2p" 1 "rectangular" 0 "linear" 0 "open" 0 "2" 0> > > </Components> > > <Wires> > > <190 350 190 380 "" 0 0 0 ""> > > <80 320 160 320 "" 0 0 0 ""> > > <220 320 300 320 "" 0 0 0 ""> > > </Wires> > > <Diagrams> > > </Diagrams> > > <Paintings> > > </Paintings> > > > > > > What I've done here is to create a 2-port S-parameter component whose File is set to "#Integer#.s2p". Then I added an optimization variable called "Integer" with linear integer type and range 0 to 10. When I run the simulation, I get an error, since I haven't actually created any S-parameter files with the appropriate names. However, the error text reads: > > > > > error loading `D:/QUCS/ASCO Test_prj/8.000000E+000.s2p': No such file or directory > > > > > > This makes me think the ASCO optimizer has replaced #Integer# with the integer value, represented in scientific notation, in the file name as well. All that would be needed is to copy/rename the manufacturer's S-parameter files to this naming convention. > > On Thu, Mar 29, 2018 at 11:14 PM, Michael Cooper <mi...@fl... mailto:mi...@fl... > wrote: > > > > Hi Claudio-- > > > > Looking at the source code in a little more detail, I believe the flow of execution for optimizations goes something like this: > > 1. Qucs prepares a netlist with optimization variables identified with hash marks, e.g., L="#Length#". > > 2. Qucs calls ASCO, passing it the name of the netlist file as well as a file for simulation output. > > 3. To evaluate the fitness of a given model, ASCO builds a new netlist using the original as a template but replacing the marked values. Then it calls qucsator to simulate that netlist and parses. Finally, it parses the output file to determine the error value. > > 4. Once ASCO is done, the best values are stored in an output file and the program ends. > > Since ASCO is called only once, and doesn't return until optimization is complete, I'm inclined to agree that the best way to tackle this may be through ASCO, either by replacing qucsator as you suggest or, in the longer term, making the required changes in ASCO. ASCO currently has an "RF module" which allows the user to specify an equivalent circuit for resistors, capacitors and inductors, with the values of the individual components defined as a function of the parent's value. That might be a good place to introduce something like this. > > > > I'll report back if I'm able to replace qucsator with a script to accomplish this job. > > > > On Tue, Mar 27, 2018 at 1:44 PM, Claudio Girardi via Qucs-help <quc...@li... mailto:quc...@li... > wrote: > > > > > > > Hello, > > > > > > interesting problem, not sure there is an easy solution currently. > > > > > > One, convoluted, possibility could be to substitute qucsator, the Qucs simulator called by the ASCO optimizer, with a script that actually modifies the netlist to use the right S-parameters file before calling the real qucsator. You could have in the schematic a dummy component, whose value ASCO tries to optimize, the script then will read the component value assigned during the current simulation run and based on this rewrite the netlist with the corresponding S-parameters file assigned to the component of interest and the run qucsator. ASCO will then read the simulation output and continue as usual. > > > > > > Hope the explanation makes sense - I did not actually try this myself. > > > > > > > > > In the long term the best way would of course be to ask the ASCO http://asco.sourceforge.net/ author to support this general discrete optimization feature. > > > > > > > > > Regards, > > > > > > Claudio > > > > > > > Il 23 marzo 2018 alle 6.54 Michael Cooper <mi...@fl... mailto:mi...@fl... > ha scritto: > > > > > > > > A related question: For a couple of days, I've been working on building impedance matching networks using Qucs. To do this, I model my transmission lines and matching components then use optimization to find a set of component values which results in the desired impedance, filter characteristics, etc. > > > > > > > > At first, I did this optimization using ideal matching components. Once I had a good result with ideal components, I would swap them out for 2-port models of actual components with those values to get a more precise idea of the matching network's attributes. However, I noticed there was a fairly large change when I did this. > > > > > > > > To mitigate this problem, what I've done is to create a parametric model of the actual components, so that, e.g., I can say fairly accurately what the parasitic capacitance of a 4.7 nH inductor will be. This results in a much smaller change when I introduce the 2-port models, but there is still some change. > > > > > > > > Ideally, I'd like to skip the middle step and have the optimizer use the correct s2p file for each component value. This way, the optimizer would actually select a set of s2p files (one for each component) which results in the best matching network. Alas, I haven't been able to find a way to do this in Qucs. > > > > > > > > I am a programmer, so I've subscribed to the qucs-devel list as well with the notion of adding this functionality if it is not already present. However, before I get ahead of myself, I thought I would post here to see if this can already be accomplished somehow in Qucs. Can anyone think of a way? > > > > > > > > On Thu, Mar 22, 2018 at 9:33 AM, Michael Cooper <mi...@fl... mailto:mi...@fl... mailto:mi...@fl... mailto:mi...@fl... > wrote: > > > > > > > > > > Thanks! That worked perfectly. > > > > > > > > > > On Thu, Mar 22, 2018 at 1:22 AM, Claudio Girardi via Qucs-help <quc...@li... mailto:quc...@li... mailto:quc...@li... mailto:quc...@li... > wrote: > > > > > > > > > > > > > Hello, > > > > > > I think I've seen something similar before and concluded it was due to some incorrect rounding in the external optimizer, ASCO. I always used the E-series optimization feature with values greater than 1, I assume it should work also for the small values you are using but maybe this is triggering the rounding issue I mentioned. As a workaround you may try to set the optimization limits from 1 to 100 and scale the optimization variables in a separate equation. > > > > > > If you can share the failing schematic I can try to take a closer look. > > > > > > > > > > > > Regards, > > > > > > Claudio > > > > > > > > > > > > > Il 22 marzo 2018 alle 7.43 Michael Cooper < mi...@fl... mailto:mi...@fl... mailto:mi...@fl... mailto:mi...@fl... > ha scritto: > > > > > > > > > > > > > > > > > > > > > I've been able to use the optimization feature quite effectively. However, > > > > > > > I've been having some trouble using the "E24 series" type when defining > > > > > > > optimization variables. Here's how I've tried to set up such a variable: > > > > > > > > > > > > > > Name: Lseries > > > > > > > Initial: 10e-9 > > > > > > > Min: 1e-9 > > > > > > > Max: 100e-9 > > > > > > > > > > > > > > > > > > > > > However, when I do this, I get an optimization error: > > > > > > > > > > > > > > initialize.c - scaleto -- Initial value in *.cfg is larger than the maximum > > > > > > > > value > > > > > > > > > > > > > > > > > > > > > If I look in "asco_netlist.cfg" as suggested by the error, then I see: > > > > > > > > > > > > > > Parameter 1:#Lseries#:10e-9:1e-9:100e-9: E24:OPT > > > > > > > > > > > > > > > > > > > > > It looks like the initial value here is just fine, so I assume I am somehow > > > > > > > using the "E24 values" feature incorrectly. Can someone tell me what I'm > > > > > > > doing wrong here? > > > > > > > > > > > > > > I am currently using "Version 0.0.19 (9aa60413f)", but have encountered the > > > > > > > same problem with the stable "Version 0.0.19". > > > > > > > > > > > > > > Thanks for your help! > > > > > > > ------------------------------ ------------------------------ ------------------ > > > > > > > Check out the vibrant tech community on one of the world's most > > > > > > > engaging tech sites, Slashdot.org! http://sdm.link/slashdot > > > > > > > ______________________________ _________________ > > > > > > > Qucs-help mailing list > > > > > > > Quc...@li... mailto:Quc...@li... mailto: Quc...@li... mailto:Quc...@li... > > > > > > > https://lists.sourceforge.net/lists/listinfo/qucs-help https://lists.sourceforge.net/lists/listinfo/qucs-help https://lists.sourceforge.net/lists/listinfo/qucs-help https://lists.sourceforge.net/lists/listinfo/qucs-help > > > > > > > > > > > > ----------------------------- - ------------------------------ ------------------ > > > > > > Check out the vibrant tech community on one of the world's most > > > > > > engaging tech sites, Slashdot.org! http://sdm.link/slashdot > > > > > > ______________________________ _________________ > > > > > > Qucs-help mailing list > > > > > > Quc...@li... mailto:Quc...@li... mailto: Quc...@li... mailto:Quc...@li... > > > > > > https://lists.sourceforge.net/lists/listinfo/qucs-help https://lists.sourceforge.net/lists/listinfo/qucs-help https://lists.sourceforge.net/lists/listinfo/qucs-help https://lists.sourceforge.net/lists/listinfo/qucs-help > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > ------------------------------ ------------------------------ ------------------ > > > Check out the vibrant tech community on one of the world's most > > > engaging tech sites, Slashdot.org! http://sdm.link/slashdot > > > ______________________________ _________________ > > > Qucs-help mailing list > > > Quc...@li... mailto:Quc...@li... > > > https://lists.sourceforge.net/lists/listinfo/qucs-help https://lists.sourceforge.net/lists/listinfo/qucs-help > > > > > > > > > > > > > > |
From: Michael C. <mi...@fl...> - 2018-03-30 20:27:22
|
It worked beautifully! Let's say I want to optimize the value of a single inductor. I would follow these steps: 1. Batch download the S2P files from your manufacturer's website. In my case, I've downloaded all 20 files for the Murata LQG15HS family. 2. Copy these files, in order of component value, to files named, e.g., "0.000000E+000.s2p". 3. Create an optimization variable, say "L_Index", of type "linear integer" with a range covering the values above. In my case, "min" is 0 and "max" is 19. 4. In your schematic, add a 2-port S-parameter part. Change the file name, e.g., to "D:\Parts\LQG15HS\#L_Index#.s2p". ASCO will replace the part between hash marks with an integer value represented in scientific notation as noted above. 5. Set up your optimization goals and run the optimization. That should do it. This depends on a couple of "bugs", as Claudio pointed out, which hopefully will remain in the Qucs and ASCO software: - Technically you can't use a hash mark in a file name, at least in Windows. Luckily Qucs doesn't filter this value too carefully. - ASCO seems to do a global find/replace for the value's "tag" even if it's part of a file name. As I said before, possibly this kind of functionality could be added to ASCO's RF module, but that would require new support within Qucs (or running the optimization manually outside of Qucs). I'm not sure we could do much better than the method above, provided everyone leaves these "bugs" open. On Fri, Mar 30, 2018 at 1:34 AM, Claudio Girardi via Qucs-help < quc...@li...> wrote: > ah, you may be onto something here. Seems you were able to turn a (sort > of) bug into a feature :) . Let us know if you are able to make it work > this way. > > > Regards, > > Claudio > > > > Il 30 marzo 2018 alle 7.41 Michael Cooper <mi...@fl...> ha > scritto: > > > > Hi guys-- > > > > I think I've just found a better way to do this without creating a > script. What I noticed is that ASCO uses the hash marks to identify > optimization variables. So, I wondered, what would happen if I used the > same hash marks in my file name? Here is an example schematic I created: > > > > > > > > <Qucs Schematic 0.0.19> > > > <Components> > > > <.SP SP1 1 80 80 0 63 0 0 "lin" 1 "3 MHz" 1 "6 GHz" 1 "201" 1 > "no" 0 "1" 0 "2" 0 "no" 0 "no" 0> > > > <.Opt Opt1 1 300 80 0 39 0 0 "Sim=SP1" 0 > "DE=3|50|2|20|0.85|1|3|1e-6|10|100" 0 "Var=Integer|yes|8.000000E+000|0|10|LIN_INT" > 0> > > > <Pac P1 1 80 350 18 -26 0 1 "1" 1 "50 Ohm" 1 "0 dBm" 0 "1 GHz" > 0 "26.85" 0> > > > <GND * 1 80 380 0 0 0 0> > > > <GND * 1 190 380 0 0 0 0> > > > <C C1 1 300 350 17 -26 0 1 "1 pF" 1 "" 0 "neutral" 0> > > > <GND * 1 300 380 0 0 0 0> > > > <SPfile X1 1 190 320 -26 -59 0 0 "#Integer#.s2p" 1 > "rectangular" 0 "linear" 0 "open" 0 "2" 0> > > > </Components> > > > <Wires> > > > <190 350 190 380 "" 0 0 0 ""> > > > <80 320 160 320 "" 0 0 0 ""> > > > <220 320 300 320 "" 0 0 0 ""> > > > </Wires> > > > <Diagrams> > > > </Diagrams> > > > <Paintings> > > > </Paintings> > > > > > > > > > What I've done here is to create a 2-port S-parameter component > whose File is set to "#Integer#.s2p". Then I added an optimization variable > called "Integer" with linear integer type and range 0 to 10. When I run the > simulation, I get an error, since I haven't actually created any > S-parameter files with the appropriate names. However, the error text reads: > > > > > > > > error loading `D:/QUCS/ASCO > Test_prj/8.000000E+000.s2p': No such file or directory > > > > > > > > > This makes me think the ASCO optimizer has replaced #Integer# with > the integer value, represented in scientific notation, in the file name as > well. All that would be needed is to copy/rename the manufacturer's > S-parameter files to this naming convention. > > > > On Thu, Mar 29, 2018 at 11:14 PM, Michael Cooper < > mi...@fl... mailto:mi...@fl... > wrote: > > > > > > Hi Claudio-- > > > > > > Looking at the source code in a little more detail, I believe > the flow of execution for optimizations goes something like this: > > > 1. Qucs prepares a netlist with optimization variables > identified with hash marks, e.g., L="#Length#". > > > 2. Qucs calls ASCO, passing it the name of the netlist file > as well as a file for simulation output. > > > 3. To evaluate the fitness of a given model, ASCO builds a > new netlist using the original as a template but replacing the marked > values. Then it calls qucsator to simulate that netlist and parses. > Finally, it parses the output file to determine the error value. > > > 4. Once ASCO is done, the best values are stored in an > output file and the program ends. > > > Since ASCO is called only once, and doesn't return until > optimization is complete, I'm inclined to agree that the best way to tackle > this may be through ASCO, either by replacing qucsator as you suggest or, > in the longer term, making the required changes in ASCO. ASCO currently has > an "RF module" which allows the user to specify an equivalent circuit for > resistors, capacitors and inductors, with the values of the individual > components defined as a function of the parent's value. That might be a > good place to introduce something like this. > > > > > > I'll report back if I'm able to replace qucsator with a script > to accomplish this job. > > > > > > On Tue, Mar 27, 2018 at 1:44 PM, Claudio Girardi via Qucs-help > <quc...@li... mailto:quc...@li... > > wrote: > > > > > > > > > Hello, > > > > > > > > interesting problem, not sure there is an easy solution > currently. > > > > > > > > One, convoluted, possibility could be to substitute > qucsator, the Qucs simulator called by the ASCO optimizer, with a script > that actually modifies the netlist to use the right S-parameters file > before calling the real qucsator. You could have in the schematic a dummy > component, whose value ASCO tries to optimize, the script then will read > the component value assigned during the current simulation run and based on > this rewrite the netlist with the corresponding S-parameters file assigned > to the component of interest and the run qucsator. ASCO will then read the > simulation output and continue as usual. > > > > > > > > Hope the explanation makes sense - I did not actually > try this myself. > > > > > > > > > > > > In the long term the best way would of course be to ask > the ASCO http://asco.sourceforge.net/ author to support this general > discrete optimization feature. > > > > > > > > > > > > Regards, > > > > > > > > Claudio > > > > > > > > > Il 23 marzo 2018 alle 6.54 Michael Cooper < > mi...@fl... mailto:mi...@fl... > ha scritto: > > > > > > > > > > A related question: For a couple of days, I've > been working on building impedance matching networks using Qucs. To do > this, I model my transmission lines and matching components then use > optimization to find a set of component values which results in the desired > impedance, filter characteristics, etc. > > > > > > > > > > At first, I did this optimization using ideal > matching components. Once I had a good result with ideal components, I > would swap them out for 2-port models of actual components with those > values to get a more precise idea of the matching network's attributes. > However, I noticed there was a fairly large change when I did this. > > > > > > > > > > To mitigate this problem, what I've done is to > create a parametric model of the actual components, so that, e.g., I can > say fairly accurately what the parasitic capacitance of a 4.7 nH inductor > will be. This results in a much smaller change when I introduce the 2-port > models, but there is still some change. > > > > > > > > > > Ideally, I'd like to skip the middle step and have > the optimizer use the correct s2p file for each component value. This way, > the optimizer would actually select a set of s2p files (one for each > component) which results in the best matching network. Alas, I haven't been > able to find a way to do this in Qucs. > > > > > > > > > > I am a programmer, so I've subscribed to the > qucs-devel list as well with the notion of adding this functionality if it > is not already present. However, before I get ahead of myself, I thought I > would post here to see if this can already be accomplished somehow in Qucs. > Can anyone think of a way? > > > > > > > > > > On Thu, Mar 22, 2018 at 9:33 AM, Michael Cooper < > mi...@fl... mailto:mi...@fl... mailto:mi...@fl... > mailto:mi...@fl... > wrote: > > > > > > > > > > > > Thanks! That worked perfectly. > > > > > > > > > > > > On Thu, Mar 22, 2018 at 1:22 AM, Claudio > Girardi via Qucs-help <quc...@li... mailto: > quc...@li... mailto:quc...@li... > mailto:quc...@li... > wrote: > > > > > > > > > > > > > > > Hello, > > > > > > > I think I've seen something similar > before and concluded it was due to some incorrect rounding in the external > optimizer, ASCO. I always used the E-series optimization feature with > values greater than 1, I assume it should work also for the small values > you are using but maybe this is triggering the rounding issue I mentioned. > As a workaround you may try to set the optimization limits from 1 to 100 > and scale the optimization variables in a separate equation. > > > > > > > If you can share the failing schematic > I can try to take a closer look. > > > > > > > > > > > > > > Regards, > > > > > > > Claudio > > > > > > > > > > > > > > > Il 22 marzo 2018 alle 7.43 Michael > Cooper < mi...@fl... mailto:mi...@fl... mailto: > mi...@fl... mailto:mi...@fl... > ha scritto: > > > > > > > > > > > > > > > > > > > > > > > > I've been able to use the > optimization feature quite effectively. However, > > > > > > > > I've been having some trouble using > the "E24 series" type when defining > > > > > > > > optimization variables. Here's how > I've tried to set up such a variable: > > > > > > > > > > > > > > > > Name: Lseries > > > > > > > > Initial: 10e-9 > > > > > > > > Min: 1e-9 > > > > > > > > Max: 100e-9 > > > > > > > > > > > > > > > > > > > > > > > > However, when I do this, I get an > optimization error: > > > > > > > > > > > > > > > > initialize.c - scaleto -- Initial > value in *.cfg is larger than the maximum > > > > > > > > > value > > > > > > > > > > > > > > > > > > > > > > > > If I look in "asco_netlist.cfg" as > suggested by the error, then I see: > > > > > > > > > > > > > > > > Parameter > 1:#Lseries#:10e-9:1e-9:100e-9: E24:OPT > > > > > > > > > > > > > > > > > > > > > > > > It looks like the initial value here > is just fine, so I assume I am somehow > > > > > > > > using the "E24 values" feature > incorrectly. Can someone tell me what I'm > > > > > > > > doing wrong here? > > > > > > > > > > > > > > > > I am currently using "Version 0.0.19 > (9aa60413f)", but have encountered the > > > > > > > > same problem with the stable > "Version 0.0.19". > > > > > > > > > > > > > > > > Thanks for your help! > > > > > > > > ------------------------------ > ------------------------------ ------------------ > > > > > > > > Check out the vibrant tech community > on one of the world's most > > > > > > > > engaging tech sites, Slashdot.org! > http://sdm.link/slashdot > > > > > > > > ______________________________ > _________________ > > > > > > > > Qucs-help mailing list > > > > > > > > Quc...@li... > mailto:Quc...@li... mailto: > Quc...@li... mailto:Quc...@li... > > > > > > > > https://lists.sourceforge.net/ > lists/listinfo/qucs-help https://lists.sourceforge.net/ > lists/listinfo/qucs-help https://lists.sourceforge.net/ > lists/listinfo/qucs-help https://lists.sourceforge.net/ > lists/listinfo/qucs-help > > > > > > > > > > > > > > ----------------------------- - > ------------------------------ ------------------ > > > > > > > Check out the vibrant tech community > on one of the world's most > > > > > > > engaging tech sites, Slashdot.org! > http://sdm.link/slashdot > > > > > > > ______________________________ > _________________ > > > > > > > Qucs-help mailing list > > > > > > > Quc...@li... > mailto:Quc...@li... mailto: > Quc...@li... mailto:Quc...@li... > > > > > > > https://lists.sourceforge.net/ > lists/listinfo/qucs-help https://lists.sourceforge.net/ > lists/listinfo/qucs-help https://lists.sourceforge.net/ > lists/listinfo/qucs-help https://lists.sourceforge.net/ > lists/listinfo/qucs-help > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > ------------------------------ > ------------------------------ ------------------ > > > > Check out the vibrant tech community on one of the > world's most > > > > engaging tech sites, Slashdot.org! > http://sdm.link/slashdot > > > > ______________________________ _________________ > > > > Qucs-help mailing list > > > > Quc...@li... mailto:Qucs-help@lists. > sourceforge.net > > > > https://lists.sourceforge.net/lists/listinfo/qucs-help > https://lists.sourceforge.net/lists/listinfo/qucs-help > > > > > > > > > > > > > > > > > > > > ------------------------------------------------------------ > ------------------ > Check out the vibrant tech community on one of the world's most > engaging tech sites, Slashdot.org! http://sdm.link/slashdot > _______________________________________________ > Qucs-help mailing list > Quc...@li... > https://lists.sourceforge.net/lists/listinfo/qucs-help > |