Re: Simulation runs very slowly: test.asc
Yes, I know that was Mike's argument, and of
course he is notoriously unbiddable. So I suggest it could be an
option in the symbol creating page, 'Do not add full path to the
model file'. Or the opposite.
On 2025-03-30 23:56, Andy I via
groups.io wrote:
On Sun, Mar 30, 2025 at 06:46 PM, John Woodgate wrote:
Yes, it can be fixed, but
wouldn't it be better if it didn't happen? Should we ask
for it to be changed to only show the model filename?
It is actually this way quite intentionally.? I had a discussion
with Mike Engelhardt about it several years ago.
?
Auto-generating symbols makes it as easy as possible for
anyone to make a symbol without trying.? As such, it needs to
encode where the model file lives, on the assumption that it
might not already be located in a place where LTspice looks to
find model files.? Thus, it saves the whole filespec to the
model file, in the symbol.
?
Shortening that to just the filename SHOULD (in my opinion)
be done, manually, by any LTspice user who wants to have a hand
in the simulation process.? But it would break it for other
LTspice users who don't care or don't want to understand
anything about "what's under the hood".? It would be bad (in my
opinion) to change LTspice in a way that makes the Automatic
Symbol Generation step fail for many users, especially the
newbies.
?
?
--
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
|
Re: PTC model with internal temperature rise
On Sat, Mar 29, 2025 at 03:37 PM, Andy I wrote:
It is currently in the Temp folder at the group's website.
Hello Andy,
it's ok now, I move it to the good folder.
|
Re: Simulation runs very slowly: test.asc
This is what happens if you don't start a new thread: Email
threads. Changing the subject doesn't affect how they are
listed in a threaded display. It just confuses things.
People use threads in (news)groups in order to see which message is
a direct reply to another, and which messages are not. Threads may
contain hundreds of messages, all laid out logically.
--
Regards,
Tony
On 30/03/2025 23:45, Christopher Paul
via groups.io wrote:
toggle quoted message
Show quoted text
I’m
sorry.
?
I
count one email with a definitely wrong subject: RE: [LTspice]
Model of BF970
?
The
others have different subject files which I have emailed, but
they are each my own:
test.asc
test.asc
and ADA4084-2.zip
test_2.zip.
?
I
see now that I should have kept one email title for all of
these, which were the results of problems that Andy requested
I fix.
?
I
use gmail. So:
?
???????????????
In the future, I should continue to post my schematic files in
the temp folder. All emails related to my topic should have a
unique one and only subject. Correct?
?
In
spite of my mistakes, is there something that I can do now to
request help?
?
Apologies,
|
I tried to simulate a negative resistance using btdeboi's netlist, but I can't complete the schematic.
Can you help me complete it ?
See Negative Conductance.asc
|
Re: Simulation runs very slowly: test.asc
Chris,
?
Analog Devices has another SPICE model for the ADA4084, downloadable from the part's webpage.? The model that comes with LTspice has elements that require LTspice.? The downloadable model is generic SPICE.? They were likely created by different people at ADI.
?
I tried it in your circuit.
?
It seems to "work" with either Solver, with or without a 10K pulldown resistor at the output.? However, it is rather slow using the Normal Solver.? It's not nearly as slow as your circuit was.? It makes slow but steady progress.? Also there are glitches, so something (an oscillation?) seems to be going on.
?
With the Alternate Solver, it takes a few seconds to find the initial operating point, but then it simulates the rest in the blink of an eye, and I see no glitches in the waveform.
?
These are interesting results, but puzzling.
?
Andy
?
?
|
Re: Simulation runs very slowly: test.asc
Chris,
?
Here is another thing that seems to help:? Add a pull-up or pull-down resistor between the output of the op-amp and its V+ or V- supply.? I used a 10K resistor and that seemed to do a nice job with LTspice's Normal Solver.? In fact, using the 10K pulldown to V- also makes LTspice find the initial operating point much faster.? (Source Stepping succeeds, but failed without it.)
?
I was noticing the fact that the ADA4084's output pin seems to be the junction of two collectors, so its open-loop Zout might be rather large, and I wondered what were the consequences of that when driving nothing more than a MOSFET gate.? So I added a resistor there to maybe dampen and lower the impedance, and for whatever reason, it seems to quench whatever goes wrong in the simulation.? No guarantees.
?
I don't know if this translates to the need for something similar in actual hardware.
?
The datasheet mentions that the ADA4084 open-loop voltage gain depends on the load connected to the output.
?
Andy
?
|
Re: Simulation runs very slowly: test.asc
On Sun, Mar 30, 2025 at 08:35 PM, Christopher Paul wrote:
... Only one of the _IN, V_Supply and _OUT supplies is active at a time. When active, they test the circuits’ transfer, supply noise suppression, and output impedances over frequency at DC currents of 10, 100 and 1000mA.
What are the _IN and _OUT supplies?? I'm guessing they are on schematics you have but did not upload?
?
?
|
Re: Simulation runs very slowly: test.asc
Chris,
?
These kinds of internal "oscillations" are more mathematical, not real.? Has nothing to do with being unity-gain stable.? I just want to suggest the possibility of some sort of numerical instability, perhaps in the sub-femtosecond range.
?
It's hard to say if their model alone is at fault.? Might be interaction between devices.
?
Andy
|
Re: Simulation runs very slowly: test.asc
Hi Andy, ? ??????????????? Thank you for all your help. ? I had wondered if oscillations were the cause the problem, and were associated with the ADA4084-2 model since replacing it with opamp.sub made things run normally. ? ??????????????? Would you suggest contacting Analog Devices about this? The datasheet says unity gain operation is acceptable. ? Thanks and Regards, ? Chris ?
toggle quoted message
Show quoted text
From: [email protected] <[email protected]> On Behalf Of Andy I via groups.io
Sent: Sunday, March 30, 2025 8:14 PM
To: [email protected]
Subject: Re: [LTspice] Simulation runs very slowly: test.asc? I changed the Subject line of this topic from "test.asc" to "Simulation runs very slowly: test.asc", because "test.asc" by itself was meaningless. It won't affect most of you, and it does not change the problem that it hijacked another topic.? Sorry to you Thunderbird email users, but we are stuck with that.
|
Simulation runs very slowly: test.asc
Hi Andy and John, ? ??????????????? To answer John’s question: ? ??????????????? Both the LM317 circuit and the ?“M1” circuits are voltage-controlled current sources whose loads are the 0V DC voltage sources whose names terminate in _OUT. The LM317 is a standard linear voltage regulator ? The point is to compare the operations of the two circuits. Only one of the _IN, V_Supply and _OUT supplies is active at a time. When active, they test the circuits’ transfer, supply noise suppression, and output impedances over frequency at DC currents of 10, 100 and 1000mA. ? ??????????????? Thanks for the 6 diodes trick, but the schematic is also meant to describe the circuits to non-spice users, as you surmised. ? - Chris. I don't understand you circuits. What are they supposed to do? Where is the output? I can see it might be the top ends of the diode stacks, but that doesn't work for the circuit that doesn't have a diode stack.
I am not sure but I figured the LM317 might be some sort of current limiter or regulator?? I'm guessing the "output" of those sections is the current through the diodes and current-sensing voltage source at the bottom of each stack. The one section without the diode stack doesn't have the LM317 either.? I think it just tests that the current sink (MOSFET + op-amp) below the LM317 does the right thing.? That part of the circuit appears in each of the other three sections. I can't run your sims because your LM317 symbol has the full path to its model which is on your computer only.
That is "easily" fixed and by now I have gotten accustomed to doing that after group members upload their schematics.? It is an unfortunate consequence of letting LTspice auto-generate symbols.? I can't blame the LTspice user for having that.? (Although I think it is quite unnecessary to auto-generate symbols.) Not that it matters, but SPICE gives you a short-cut for having multiple diodes in series.? It is the N=<value> instance parameter.? By adding "n=6" to a diode element, it turns it into effectively six individual diodes in series.? With LTspice, you can do that using either of these: - Right-click on the text "1N4007" and change it to "1N4007 N=6"; or
- Ctrl-right-click on the diode symbol and add "N=6" (without quotes) in the Value2 line, and optionally add an X in the Vis. column.
Doing so does not actually change anything (to any significant degree), but it makes the schematic more compact, as well as the circuit matrix.? You might or might not want to "simplify" your schematic this way because seeing six diodes in series conveys meaning.
|
Re: Simulation runs very slowly: test.asc
I changed the Subject line of this topic from "test.asc" to "Simulation runs very slowly: test.asc", because "test.asc" by itself was meaningless.
?
It won't affect most of you, and it does not change the problem that it hijacked another topic.? Sorry to you Thunderbird email users, but we are stuck with that.
?
Andy
?
|
Re: Simulation runs very slowly: test.asc
Chris,
?
Also, FYI -? In many cases like this, one remedy (e.g., Alternate Solver) might not be the only one that helps.? There might be other settings that work even with the Normal Solver.
?
Andy
?
?
|
Re: Simulation runs very slowly: test.asc
Chris,
?
FYI - Your simulation runs OK after changing to the Alternate Solver.
?
Control Panel (Settings) > SPICE tab > Solver: change to Alternate.
?
When done using the Alternate solver, you might want to change it back to the Normal solver.? Newer versions (24.1.*) of LTspice now have a way to specify which solver to use, but older versions need to be manually changed back to Normal if that is what you normally want.
?
Your circuit has much trouble simulating.
?
The error log reports three instances of "Questionable use of curly braces" and three (apparently false) Errors, which are probably caused by T.I.'s LM317 SPICE model.? In my experience, those warnings and errors seem to be harmless and do not seem to affect the outcome.? They can be "fixed", if desired, so that those messages do not show up - but it's probably not necessary.
?
Then the simulation has great difficulty finding the initial (DC) operating point.? Direct Newton Iteration fails.? GMIN Stepping fails.? Source Stepping fails.? Pseudo-Tran might have succeeded, but I can't be sure.
?
Once past that, the transient simulation has considerable difficulty when using the Normal Solver, but only with the Normal Solver.? The .log file is full of Heightened DefCon warnings, and some of the voltages are "funny" and probably wrong.
?
My guess is that something in the circuit is oscillating at very high frequency, causing considerable slowness.? Lower frequency oscillations are visible at some nodes, maybe a consequence of a >UHF oscillation.
?
I can't make conclusions, but one possibility is that something in this simulation needs the extra few digits of precision of the Alternate Solver.? And maybe there's a bad model.? Since it is the circuit with the ADA8084-2 that has trouble, that one is suspect, but that's all I can say so far.
?
Andy
?
|
Re: Simulation runs very slowly: test.asc
On Sun, Mar 30, 2025 at 06:46 PM, John Woodgate wrote:
Yes, it can be fixed, but wouldn't it be better if it didn't happen? Should we ask for it to be changed to only show the model filename?
It is actually this way quite intentionally.? I had a discussion with Mike Engelhardt about it several years ago.
?
Auto-generating symbols makes it as easy as possible for anyone to make a symbol without trying.? As such, it needs to encode where the model file lives, on the assumption that it might not already be located in a place where LTspice looks to find model files.? Thus, it saves the whole filespec to the model file, in the symbol.
?
Shortening that to just the filename SHOULD (in my opinion) be done, manually, by any LTspice user who wants to have a hand in the simulation process.? But it would break it for other LTspice users who don't care or don't want to understand anything about "what's under the hood".? It would be bad (in my opinion) to change LTspice in a way that makes the Automatic Symbol Generation step fail for many users, especially the newbies.
?
?
|
Re: Simulation runs very slowly: test.asc
Yes, it can be fixed, but wouldn't it be
better if it didn't happen? Should we ask for it to be changed
to only show the model filename?
On 2025-03-30 23:39, Andy I via
groups.io wrote:
That is "easily"
fixed and by now I have gotten accustomed to doing that after
group members upload their schematics.? It is an unfortunate
consequence of letting LTspice auto-generate symbols.?
--
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
|
Re: Simulation runs very slowly: test.asc
On Sun, Mar 30, 2025 at 06:17 PM, John Woodgate wrote:
I don't understand you circuits. What are they supposed to do? Where is the output? I can see it might be the top ends of the diode stacks, but that doesn't work for the circuit that doesn't have a diode stack.
I am not sure but I figured the LM317 might be some sort of current limiter or regulator?? I'm guessing the "output" of those sections is the current through the diodes and current-sensing voltage source at the bottom of each stack.
?
The one section without the diode stack doesn't have the LM317 either.? I think it just tests that the current sink (MOSFET + op-amp) below the LM317 does the right thing.? That part of the circuit appears in each of the other three sections.
?
I can't run your sims because your LM317 symbol has the full path to its model which is on your computer only.
That is "easily" fixed and by now I have gotten accustomed to doing that after group members upload their schematics.? It is an unfortunate consequence of letting LTspice auto-generate symbols.? I can't blame the LTspice user for having that.? (Although I think it is quite unnecessary to auto-generate symbols.)
?
Not that it matters, but SPICE gives you a short-cut for having multiple diodes in series.? It is the N=<value> instance parameter.? By adding "n=6" to a diode element, it turns it into effectively six individual diodes in series.? With LTspice, you can do that using either of these:
- Right-click on the text "1N4007" and change it to "1N4007 N=6"; or
- Ctrl-right-click on the diode symbol and add "N=6" (without quotes) in the Value2 line, and optionally add an X in the Vis. column.
Doing so does not actually change anything (to any significant degree), but it makes the schematic more compact, as well as the circuit matrix.? You might or might not want to "simplify" your schematic this way because seeing six diodes in series conveys meaning.
?
?
|
Re: Simulation runs very slowly: test.asc
On Sun, Mar 30, 2025 at 05:45 PM, Christopher Paul wrote:
In spite of my mistakes, is there something that I can do now to request help?
I have mostly needed to be away from my computer all day today, so I did not yet actually try simulating your circuit.? I only saw that the pieces were there.
?
Andy
?
|
Re: Simulation runs very slowly: test.asc
Yes, correct. You have not been banned, so you
can still ask for help. But I give you another tip - in future
use more explicit subjects than 'test'. Everyone and his cat has
a file called 'test'.
I don't understand you circuits. What are they
supposed to do? Where is the output? I can see it might be the
top ends of the diode stacks, but that doesn't work for the
circuit that doesn't have a diode stack.
I can't run your sims because your LM317
symbol has the full path to its model which is on your computer
only.
On 2025-03-30 22:45, Christopher Paul
via groups.io wrote:
? In the
future, I should continue to post my schematic files in the temp
folder. All emails related to my topic should have a unique one
and only subject. Correct?
--
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
|
Re: Simulation runs very slowly: test.asc
I’m sorry. ? I count one email with a definitely wrong subject: RE: [LTspice] Model of BF970 ? The others have different subject files which I have emailed, but they are each my own: test.asc test.asc and ADA4084-2.zip test_2.zip. ? I see now that I should have kept one email title for all of these, which were the results of problems that Andy requested I fix. ? I use gmail. So: ? ??????????????? In the future, I should continue to post my schematic files in the temp folder. All emails related to my topic should have a unique one and only subject. Correct? ? In spite of my mistakes, is there something that I can do now to request help? ? Apologies, ? Chris ?
toggle quoted message
Show quoted text
From: [email protected] <[email protected]> On Behalf Of Tony Casey via groups.io
Sent: Sunday, March 30, 2025 5:29 PM
To: [email protected]
Subject: Re: [LTspice] test.asc? No, you didn't correct it. You continued the same thread, but with a different subject.
By "start a new topic", I mean you start a new email thread by either: - Clicking "New Topic", if you're using the group website, or
- Clicking "New Message", if you are using Thunderbird, or "Compose", if you're using Gmail.
If you don't do this, the (normally) hidden email header signifies that you have answered an existing message, even if you've changed the subject. ? On 30/03/2025 19:26, Christopher Paul via groups.io wrote: Yes, I noticed this and apologized for and corrected it in subsequent posts. ? Regards, ? Chris ? ? Can you please not hijack a thread. Start a new topic. What you did was reply to a message in the "Model of BF970 ?" thread, and simply change the subject.
This is very irritating for anyone that uses threads in their email. client.
--
Regards,
Tony
? On 30/03/2025 15:03, Christopher Paul via groups.io wrote: Apologize for the wrong email heading you just received. This is the corrected one.
?
|
Re: Simulation runs very slowly: test.asc
No, you didn't correct it. You continued the same thread, but with a
different subject.
By "start a new topic", I mean you start a new email thread by
either:
- Clicking "New Topic", if you're using the group website, or
- Clicking "New Message", if you are using Thunderbird, or "Compose",
if you're using Gmail.
If you don't do this, the (normally) hidden email header
signifies that you have answered an existing message, even if
you've changed the subject.
--
Regards,
Tony
On 30/03/2025 19:26, Christopher Paul
via groups.io wrote:
toggle quoted message
Show quoted text
Yes,
I noticed this and apologized for and corrected it in
subsequent posts.
?
Regards,
?
Chris
?
?
Can you please not hijack a thread. Start a
new topic. What you did was reply to a message in the "Model of
BF970 ?" thread, and simply change the subject.
This is very irritating for anyone that uses threads in their
email. client.
--
Regards,
Tony
?
On 30/03/2025 15:03, Christopher Paul via
groups.io wrote:
Apologize for the wrong email heading you
just received. This is the corrected one.
|
John Woodgate
Tony Casey
jacfev