LTspice is a difficult tool to make useful

Am 28.05.2020 um 10:48 schrieb Ricketty C:

Hello Rick,

LTspice is used by most R&D engineers, because its free and you get free support. There is a large and very helpful LTspice user group. Please join it.

Especially the SPICE-models from Microchip make a lot of trouble due to convergence problems, because they contain a lot of functions like IF() and TABLE()) which have a discontinuous derivative. I simulate them often with the Universalopamp2 - SpiceModel level.3a or level3.b. Right-mouse-click on it and adjust GBW, slew rate and Rin.

Helmut

LT Spice is wonderful. I'd pay a lot for it if I had to.

I use Universalopamp2 by default too. If an opamp is sufficiently strange that Universalopamp2 won't be good enough, I use an available model and don't trust it.

Analog Devices is behind on providing models lately, and some are buggy.

Of course, that is true. But LT-Spice is a tool written by an enginer for other enginers and we all have a personality like Lucy.

If it was written by salesdruids it would look nice and cool without any good functionality. I this case you are the football on the ground.

It is an interesting question how it will change now with Analog. Perhaps it is a good idea to save a working copy for future use....

Olaf

There's a video interview with Mike, the inventor of LT Spice, where he says that the real value of a circuit simulator is to "cultivate your intuition."

I think he is a bit wrong about some other points, but the instinct training thing is very real.

I have got people started using LT Spice in literally 5 minutes.

I hope Analog doesn't wreck it.

uch of it is counter intuitive and even dysfunctional. I just when through an hour or two trying to change a model connected to a FET. Turns out tha t saving a matching symbol under the new name and editing the library file name in the symbol attributes doesn't cut it. The symbol in the schematic has attributes taken from that symbol that it won't show you or let you edi t from the schematic. So once you change the symbol you have to delete the component from your schematic and add it back in again from the library.

uld think it would show up in the attributes editing dialog with all the ot hers.

otball away from Charlie Brown. And just like Charlie, I keep coming back.

Yes, I am aware of that group and have been posting there about the many is sues I've had over the last few weeks. I do get help with the problems and that is greatly appreciated. However, I also get many comments that I nee d to read the documentation... which is often not there.

In one particular case I has having trouble learning how to use models from the CD4000 library. I was told to read the info files that come with the library. However those files simply tell you to unzip the files and have n o information at all on how to use the models! I suppose that particular p oster never actually read the help files for themselves.

Thanks, but this part is not an opamp, it's a comparator. I would need to construct my own model with open collector outputs, internal reference, etc . and part of the reason I want to simulate is to verify I am using the par t correctly. Constructing my own model then verifies that I make the same mistake in both model and usage.

I'm having the same problem with a similar model from Maxim. Do they also write poor models?

Looking at the various models I've downloaded I see multiple files that wer e written by a company MODPEX. I suppose they are a third party that compa nies use rather than writing their own models?

Am 28.05.2020 um 20:43 schrieb Ricketty C:

Hello Rick, Sorry about the not so good experience with the group in this case. I have written the models and symbols of the CD4000 and 74HC/74HCTxx devices starting 2003 in my free time. The models are a compromise between complexity and simulation speed. I didn't jump into this thread, because I thought many people already answered the questions.

Mike Engelhardt (author of LTspice) created a special model OTA with good behavior due to convergence. It has been used in many opamp models of LTspice. This OTA-model is not available in normal SPICE as an intrinic model. That's why models from other companies often use IF() and TABLE() models. MAXIM and Microchip provides a (limited) SPICE-software from SiMetrix for their ICs. I am not sure whether SiMetrix use the PSPICE-models from the web page or they use modified models for better convergence.

Best regards, Helmut

PS: I am one of the moderators of the LTspice group in groups.io. This group is the successor of the Yahoo group.

from the CD4000 library. I was told to read the info files that come with the library. However those files simply tell you to unzip the files and ha ve no information at all on how to use the models! I suppose that particul ar poster never actually read the help files for themselves.

Yes, several people responded and I thanked them for their help. I did eve ntually get the parts working. The point I raised in that thread is that t he documentation leaves a great deal to be desired. That point seemed to b e disputed which led to the comment about reading the LTspice help and the readme files with the models and symbols which had nothing regarding the us e, only unpacking the zip files.

to construct my own model with open collector outputs, internal reference, etc. and part of the reason I want to simulate is to verify I am using the part correctly. Constructing my own model then verifies that I make the s ame mistake in both model and usage.

lso write poor models?

Perhaps I am missing something. The part I am using is a comparator. Are you suggesting I should make my own comparator from the opamp model?

Yes, I am aware of both you and both groups. Thanks for your support. You have been amazing over the years.

Am 28.05.2020 um 22:15 schrieb Ricketty C: > Perhaps I am missing something. The part I am using is a comparator. > Are you suggesting I should make my own comparator from the opamp model?

Hello Rick,

I often use the comparator "diffschmitt" from [Digital]. Right-mouse-click on it in the schematic and enter some parameters in the attribute "SpiceLine".

SpiceLine Vhigh=5 Vt=0 Vh=0.1m

If you need an open collector output, then add a resistor between the output of the diffschmitt and the base of a default NPN-Transistor or a NMOS-transistor. If you use a NMOS-transistor, you have to add a model line with a useful parameter Kp, because the value of KP of the default-NMOS is too small in most cases.

Best regards, Helmut

Ok, I get that. I can always make my own models. But as I mentioned, they aren't as simple as falling off a log. The comparators I'm using have bui lt in references and hysteresis so that by the time I've built the model an d done the research to try to match the parameters I'm modeling and then te sted against those parameters, I've done a huge amount of work.

That's why manufacturers make models for customers to use. I wonder how th ey deal with the discontinuities with other simulators?

I often use a behavioral voltage source with a tanh() as a comparator.

The best and most effective thing they could do with LTSpice is to open-source it. So many crazy little quirks would get fixed very quickly.

I'm surprised no-one had mentioned SuperSpice, which Kevin open-sourced last year or so. Make it read LTSpice files and you'd have something really nice.

CH

Nah. When you think ADI I think they'd be more likely to take it in a Texas Instruments-direction.

i.e. make it a "Web 2.0"-enabled suite of "design tools" like TI's suite of bloated dumb shitty mostly-useless "tools" they want you to create an account with your life story to use.

Sometimes the FAE from TI looks a little bit sad when I told them it is hard to think about there parts because I can not test them in LT-Spice. :-D

Olaf

One of my guys can actually run Tina in an emergency. I've never seen WeBench work.

You don't have WURTH in the US with there gummi bears and free refill for gummi bears and part boxes?

Oh..and they made an interesting video for LT-Spice and EMC:

Olaf

Wurth is awful. A couple of their guys called on me and persuaded me to *buy* their horrible book about inductors. I read some and threw it away.

Wurth have published a pretty good book (740 pages) about using LTspice. The only complaint I have is that part of the language is franglais, probably due to the original French version.

You are not important enought. Normaly they offer there books for free.

Olaf

I think it is translated to German, too.

Olaf