Re: LTspice vs ngspice 12AU7 tube amplifier transient analysis
On Thu, Mar 6, 2025 at 03:24 AM, Carlo wrote:
can you advice a good (advanced) SPICE book ?
There are many.? So many.? Here are some that have stood the test of time and are sometimes recommended, but are not cheap:
?
Ken Kundert's "Designer's Guide to SPICE and Spectre".
?
Paul Tuinenga's "SPICE: A Guide to Circuit Simulation and Analysis Using PSpice".
?
LTspice's Help has a FAQ page, "What about a Paper Manual?"? Have you looked there?? It showed 10 book titles with "LTspice" in the name.? That was as of about 2015 so I am sure there are many more now.? One that might not be on that list is W¨¹rth Elektronik's "The LTspice XVII Simulator", published in both German and English.? I can't tell you which ones on the list are better or which should be avoided, and I am sure the slant (the approach used by the author) varies from book to book.? Some may be designed to get a newbie going, whereas others delve into the details better.
?
My personal experience with short tutorials about LTspice is that many of them are rubbish.? Some of their authors clearly do not know how to use LTspice well, and they cover up their mistakes by telling you to do things incorrectly in an attempt to fix what they did incorrectly.? Any Joe can write a tutorial and put it to video and claim that they know what they are doing.? But getting yourself published on paper takes a bit more skill.
?
I recommend doing some checking around to see what's out there and what the reviewers say.
?
Andy
?
|
Re: LTspice vs ngspice 12AU7 tube amplifier transient analysis
On Wed, Mar 5, 2025 at 08:31 AM, Andy I wrote:
I suspect most books about SPICE will tell you.
Hello Andy,
can you advice a good (advanced) SPICE book ?
?
Thanks.
?
|
Re: Restricting search in 'Messages' possible?
Marcel,
?
yes, that's exactly what I needed! Thanks a lot indeed!
?
Ryu
|
Re: Restricting search in 'Messages' possible?
On Wed, Mar 5, 2025 at 08:19 PM, Ryu wrote:
Is there a way to exclude 'four' and search only for .FOUR (period-four)?
(I assume you mean in a browser). Search for `.FOUR` (without back-ticks) and then click the `Relevance` button. -marcel
|
Restricting search in 'Messages' possible?
Hello all,
?
I wanted to look through messages concerning the .FOUR command, but there are lots of messages
containing the number 'four' which also get included, way too much so I gave up.
Is there a way to exclude 'four' and search only for .FOUR (period-four)?
?
Ryu
|
Re: Issues running LTspice as a batch service
FYI for anyone who comes next, I got it working with spicelib and passing a .net file instead of a .asc file. It might work with the normal cli passing a .net file as well but I'm too lazy to go back and try now that I have it working with spicelib
|
Re: LTspice vs ngspice 12AU7 tube amplifier transient analysis
On Wed, Mar 5, 2025 at 11:20 AM, Carlo wrote:
... Using default settings, LTspice will try Direct NR to calculate the ITS/DC operating point.
I believe that is the normal SPICE behavior, to try there first.
?
How many "Direct NR" iterations are allowed to run before giving up (no convergence achieved) and start using homotopies (i.e. gmin stepping, source stepping etc..) ?
It's been ages since I checked on that.? But it might be the option ITL1 = 100.
?
I suspect most books about SPICE will tell you.
?
Andy
?
?
|
Re: LTspice vs ngspice 12AU7 tube amplifier transient analysis
BTW, .option Noopiter in LTspice skips Direct Newton-Raphson (NR) iterations to calculate the ITS/DC operating point. Using default settings, LTspice will try Direct NR to calculate the ITS/DC operating point.
?
How many "Direct NR" iterations are allowed to run before giving up (no convergence achieved) and start using homotopies (i.e. gmin stepping, source stepping etc..) ?
?
?
|
Re: LTspice vs ngspice 12AU7 tube amplifier transient analysis
On Wed, Mar 5, 2025 at 01:22 AM, Andy I wrote:
NGSPICE has had a VDMOS model since release 32 (current is 44.2).
That is what I heard.? BUT -- Is it the same as LTspice's VDMOS?? ?That is,
did ngspice try to copy LTspice's VDMOS?? How successfully did they copy
LTspice's VDMOS?
We know exactly what ngspice is doing: (reading the source of the intrinsic model implementation, it is the standard DMOS and discrete SPICE diode model but has a thermal output node). According to the help, the LTspice model has a nonstandard Cgd(V) characteristic, and the anti-parallel diode is (of course) not like the substrate diode of a monolithic device. I don't think ngspice tries to copy LTspice, both try to approximate the same datasheet from the same parameters supplied by a .model line. Neither implementation is likely to produce nVs where Volts are expected. -marcel
|
Re: LTspice vs ngspice 12AU7 tube amplifier transient analysis
On Wed, Mar 5, 2025 at 04:27 AM, Andy I wrote:
I am uncertain what you suggested.? Why would either simulator use a "small signal linear equivalent circuit" for the transient simulation of something which is or might? be nonlinear?? Shouldn't it always solve the non-linear equations at each timestep?? That is how it is supposed to work, isn't it?
Ok yes, for transient analysis it is supposed to work that way. Therefore, at the end of transient analysis, there will be a total number of iterations done (some associated with ITS calculation and others to solve non-linear equations at successive timesteps until the end).
?
Using a small-signal linear equivalent circuit makes sense only for .AC analysis.
Ok, the above is made explicit in ngspice manual's section for .AC analysis. However, in the case at hand, even .AC analysis for both simulators doesn't provide a reasonable/good result (since the calculated DC operating point isn't reasonable).
?
Carlo
?
|
Re: LTspice vs ngspice 12AU7 tube amplifier transient analysis
On Wed, Mar 5, 2025 at 05:48 AM, Carlo wrote:
Btw, as far as I understood, transient simulation doesn't use the same "small signal linear equivalent circuit" for non-linear devices (i.e. diodes, bjt, mos etc..). Instead at each timestep it leverages on the solver (possibly iterating to solve non-linear equations) to work out circuit's voltages and currents.
I am uncertain what you suggested.? Why would either simulator use a "small signal linear equivalent circuit" for the transient simulation of something which is or might? be nonlinear?? Shouldn't it always solve the non-linear equations at each timestep?? That is how it is supposed to work, isn't it?
?
Using a small-signal linear equivalent circuit makes sense only for .AC analysis.
?
Andy
?
|
Re: LTspice vs ngspice 12AU7 tube amplifier transient analysis
On Wed, Mar 5, 2025 at 03:23 AM, <mhx@...> wrote:
NGSPICE has had a VDMOS model since release 32 (current is 44.2).
That is what I heard.? BUT -- Is it the same as LTspice's VDMOS?? ?That is, did ngspice try to copy LTspice's VDMOS?? How successfully did they copy LTspice's VDMOS?
?
Or does ngspice do a completely different VDMOS model?? Does this fall under the "VDMOS means different things to different people" which I suggested?
?
Andy
?
|
Re: LTspice vs ngspice 12AU7 tube amplifier transient analysis
On Tue, Mar 4, 2025 at 05:06 PM, Andy I wrote:
It looks like you used LTspice's schematic and transistor models (standard.???) in both simulations.? That MIGHT be a problem.? LTspice's discrete MOS models in standard.mos all use LTspice's proprietary "VDMOS" (Vertical Double-Diffused power MOSfet) model, which is a little different than SPICE's ordinary MOS models.? I don't know what ngspice does with the VDMOS model definition.
?
There is a chance that ngspice handles the IRF510 model incorrectly.
?
Thanks Andy, that's the point indeed ! I downloaded the following IRF510 model from
?
.MODEL IRF510 NMOS LEVEL=3 VTO=3.699 KP=20.82U RD=21.08M
+ RS=450.8M IS=202.7F CBD=366.6P
+ CGSO=604.9P CGDO=62.62P
?
LTspice and ngspice return exactly the same answer for ITS. However it isn't good enough since the amplifier's output signal isn't there (only some nV oscillating around 0V).
?
I let the simulation continue for 20 seconds (.tran 0 20 19) with no luck (with or without UIC to skip the ITS step). The output signal V(out1) isn't good at all both for LTspice and ngspice.
?
Btw, as far as I understood, transient simulation doesn't use the same "small signal linear equivalent circuit" for non-linear devices (i.e. diodes, bjt, mos etc..). Instead at each timestep it leverages on the solver (possibly iterating to solve non-linear equations) to work out circuit's voltages and currents.
?
Carlo.
?
?
?
?
|
Re: LTspice vs ngspice 12AU7 tube amplifier transient analysis
On Tue, Mar 4, 2025 at 02:06 PM, Andy I wrote:
I don't know what ngspice does with the VDMOS model definition.
NGSPICE has had a VDMOS model since release 32 (current is 44.2). -marcel
|
Re: LTspice vs ngspice 12AU7 tube amplifier transient analysis
Carlo, did you have succcess finding the ngspice user group?? (Not the brand new "AllSpice" group.)
?
Andy
?
|
Re: LTspice vs ngspice 12AU7 tube amplifier transient analysis
Carlo,
?
I'm sorry that I am not currently able to give you direct help with the ngspice part of your question.? But here is another thing to consider.
?
It looks like you used LTspice's schematic and transistor models (standard.???) in both simulations.? That MIGHT be a problem.? LTspice's discrete MOS models in standard.mos all use LTspice's proprietary "VDMOS" (Vertical Double-Diffused power MOSfet) model, which is a little different than SPICE's ordinary MOS models.? I don't know what ngspice does with the VDMOS model definition.
?
There is a chance that ngspice handles the IRF510 model incorrectly.
?
(I'm also aware that "VDMOS" can mean different things to different people and their tools.)
?
To put both simulators on equal footing, try downloading an IRF510 model from the internet and plug it into both simulators.
?
Andy
?
|
Re: LTspice vs ngspice 12AU7 tube amplifier transient analysis
On Tue, Mar 4, 2025 at 01:36 PM, Carlo wrote:
As you can check, .tran 100n 10m analysis gives quite different answers for ITS/DC operating point. Btw, as convergence aid, option noopiter gminsteps=0 is used in LTspice to skip direct NR and gmin stepping to get a "reasonable" ITS solution.
FYI, LTspice ignores the first parameter of the .TRAN command.? Therefore, ".tran 100n 10m" is identical to ".tran 0 10m" or ".tran 10m 10m".? In the old days of SPICE, the first parameter was the print step interval for SPICE's line-printer output.? ".tran 100n 10m" would have given you about 100,001 lines of waveform data, which might have spanned about two thousand pages (four reams) of paper in your line-printer.? LTspice never prints waveforms like that, so the first parameter is always ignored in LTspice.? Maybe ngspice still uses it for something; I don't know.? But that's off-topic anyway.
?
I have not used ngspice in years, so I can't try that simulation, but someone else here might.? Since ngspice's input is a SPICE Netlist, and since LTspice can simulate SPICE Netlists, did you also try simulating the ngspice netlist in LTspice?? It is a "sanity check", if nothing else.
?
Because your simulation uses the troublesome 12AU7 model which we already know has issues with multiple convergent operating points, it would not surprise me if differences were seen.? We saw that already, even from people using just one simulator (LTspice).? It's a flip of the coin whether it finds one initial operating point, or another, especially when they are both valid, mathematically.
For instance 12AU7's plate voltage is 12.8V for LTspice vs 22.6V for ngspice.
DC currents are also quite different and, in the end, output signal for ngspice isn't good at all even for a 1Khz amplifier's input signal.
Did you let the ngspice simulation continue for several more seconds, to see if it drifted towards a more appropriate operating point?? This might be just another case like the ones we had earlier with LTspice.? If the 12AU7 model, or any of the other models, do not handle out-of-bound conditions properly, then any simulator might find a mathematically valid but physically incorrect initial operating point.? That is really a problem with the model, not the simulator which is just doing its job.
?
I don't want to say that this IS the difference you saw, but it might be.
?
Andy
?
|
Re: LTspice vs ngspice 12AU7 tube amplifier transient analysis
Carlo,
?
Remember that uploaded files should be uploaded to the "Temp" directory only.? As it says, "Do not upload your files here" where 'here' is the top-level directory -- and "Click 'Temp' first, before uploading."
?
Please re-read the group's guidelines on the group's main webpage.? It reminds you about this:
To upload:? First navigate to the?Temp?folder, then click the "New/Upload" button.
?
Andy
?
|
Re: opamp BW slew rate search engine sorting
On Tue, Mar 4, 2025 at 03:31 PM, john23 wrote:
Hello , The analog devices website is not aloowing to filter OPAMPs exactly by the BW slew rate etc parmater.
For example as you can see here.Its not allowing me to sort OPAMP models.
Why do you say that it does not allow you to sort bandwidth or slew rate columns?? It works for me.? Did you click on the up- and down-arrows at the top of the columns?? Was there something wrong about the sort orders you saw after doing that?
?
If it did not work the way you desired, did you send feedback to Analog Devices?? Most of their webpages have a link in the lower right corner for "Site Feedback".? And there is another one near the lower right corner for "How would you describe your experience with this page?"? I can't say if web features such as that are available in all parts of the world, but there PROBABLY is a way in your country for telling ADI how their webpage did not meet your expectations.
?
The group you are in is about LTspice.? It's not a forum to complain about someone's website, even if that 'someone' is the same company that releases LTspice.? Their part selector pages have nothing to do with LTspice.? The group you are in is not run by Analog Devices, and complaining here has no assurance of reaching anyone at Analog Devices..
?
Andy
?
|
Re: Model of TDA2822, Dual Low-Voltage Audio Amplifier
I have tested it, and it seems to work quite
well. But read the data sheet with care.
On 2025-03-04 20:36, Kerim via
groups.io wrote:
On Tue, Mar 4, 2025 at 10:56 PM, John Woodgate wrote:
Have you asked ST Micro if they have a SPICE model
(not an 'LTspice model')?
During the search, I heard someone in a forum who did and didn't
get a reply.
I think I will likely have to test it in real if no one here
heard of it.
Thank you.
?
--
OOO - Own Opinions only
If something is true:
* as far as we know - it's science
*for certain - it's mathematics
*unquestionably - it's religion
|
Carlo
John Woodgate