On Tue, Feb 25, 2025 at 05:37 AM, <j.bernabe1@...> wrote:
help topics
About Ltspice
show change Log..
"Help topics" is the same as pressing F1.
?
Once the Help window opens up, there SHOULD be a column of contents on the left, and the main body on the right.? If you don't see the part on the left, then you might have clicked the "Hide" button along the top.? If so, click the "Show" button along the top.
?
Now that the Contents sub-window is present, select the "Contents" tab.? Then either click items one by one, or use Right-click > Open All, which opens up the entire Table of Contents.? From there, you can navigate to any of the Help pages.? They contain a lot of essential information for every LTspice user.
?
Andy
?
|
Re: Issues running LTspice as a batch service
On Tue, Feb 25, 2025 at 08:32 PM, Jeff Kayzerman wrote:
.... Does anyone know why running in batch mode as a service account it would fail to look in the search paths?
This is only a guess here.? But I suspect you have defined User folders for your symbols and models, which reside in one of your account's folders, perhaps under its "Documents" folder or "AppData" or whatever.? The other account does not have the same folders.? Well, it has the same-named folders but they are distinct for each Windows user.? So it does not see the same library areas.
?
The settings in the .ini file are relative to each Windows account's folders, as managed by MS-Windows.? Windows does a good job of hiding their actual physical locations.
?
Andy
?
|
Issues running LTspice as a batch service
I have a python program kicking off LTspice as a background windows process with the following command.
?
process = await asyncio.create_subprocess_exec(LTSPICE_PATH, "-b", "-run", solverFlag, fpath, stdout=asyncio.subprocess.PIPE, stderr=asyncio.subprocess.PIPE)
?
I am running this process on a server and when I run it as my user account everything works as expected. Now I need to have it run as a windows service which requires a service account. I set this up and tried running but I get messages that symbols are missing from the simulation. I figured maybe the .ini file in the service account's folder was missing the symbol and library search paths so I made sure to match the .ini file in my service account's AppData/Roaming folder to match the one on my account. Unfortunately this did not help. Does anyone know why running in batch mode as a service account it would fail to look in the search paths?
|
Re: Ltspice alternative for AD797
From the ADI data sheet: 110 MHz gain
bandwidth (G = 1000)
8 MHz bandwidth (G = 10)
On 2025-02-25 17:22, Andy I via
groups.io wrote:
On Tue, Feb 25, 2025 at 03:06 AM, john23 wrote:
Hello
, is there an alternative I could use for AD797 , LTspice
model will be best.
Have you tried the product selection guides, which most IC
vendors and many distributors offer on their websites?
The
model I got is very bad and its oscilating for no good
reason.
There probably is a reason.? It might be worth investigating
why.? High-bandwidth amps are more susceptible.? Maybe your
circuit into which you plugged the AD797 was the cause.
Gain
bandwidth: 100 MHz
-3
dB bandwidth: 8 MHz
Certainly that figure requires qualifications.? I know of no
op-amps that would have an 8 MHz open-loop -3dB bandwidth.? So
it is likely a closed-loop measure - and? then, at what
closed-loop gain is that measured?? It makes all the difference
in the world.
?
Andy
?
?
--
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
|
On 25/02/2025 10:28, Tony Casey wrote:
This
is also true with .MEAS. The time points are printed in single
precision in the logfile, by default. You can get double precision
by adding .option measdgt=15. This is true whether or not you also
use .option numdgt=15.
Having just installed V24.1.4, I noticed that:
.option measdgt=x
..appears to have been disabled, and doesn't appear to do anything
now - the .MEAS time points are now always double precision.
The log file content has changed a bit from 24.0.12 to 24.1.4. The
listed circuit is now the .net file instead of the .asc, and the
active .options are now listed, which is useful.
.option numdgt=x
..seems to works the same as before, but the raw files contents
differ slightly.
--
Regards,
Tony
|
On Tue, Feb 25, 2025 at 04:31 AM, John Woodgate wrote:
I suppose the OP posted more than one version of the .ASC. I don't think it's worth pursuing any further, but the fact that it runs OK with the simple opamp suggests that there is a problem with the AD757 model.
Or it suggests that there is a problem with THAT circuit when combined with a very wide-bandwidth op-amp like the AD797 (not 757).
?
The simple circuits I tried for the AD797 ran OK.
?
The fact that your schematic changes things, is indeed disturbing.
?
?
|
Re: Ltspice alternative for AD797
On Tue, Feb 25, 2025 at 03:06 AM, john23 wrote:
Hello , is there an alternative I could use for AD797 , LTspice model will be best.
Have you tried the product selection guides, which most IC vendors and many distributors offer on their websites?
The model I got is very bad and its oscilating for no good reason.
There probably is a reason.? It might be worth investigating why.? High-bandwidth amps are more susceptible.? Maybe your circuit into which you plugged the AD797 was the cause.
Gain bandwidth: 100 MHz
-3 dB bandwidth: 8 MHz
Certainly that figure requires qualifications.? I know of no op-amps that would have an 8 MHz open-loop -3dB bandwidth.? So it is likely a closed-loop measure - and? then, at what closed-loop gain is that measured?? It makes all the difference in the world.
?
Andy
?
?
|
Re: Monitor simulation percent completion from python
On Tue, Feb 25, 2025 at 02:51 AM, @HermanVos wrote:
For the percentage completion you need to decode the actual begin and end times from the .tran statements (assuming you do transient analysis of circuits). Sometimes these are straightforward and sometimes I see very interesting and complex constructions with formulas and .params.
I don't remember with certainty if this is true - but I believe the .TRAN command has numbers when you look at the expanded netlist.? In other words, the formulas have been evaluated.? That is true of most things (e.g., element values) in the expanded netlist.
?
Of course this assumes that you still have access to the expanded netlist, and newer versions of LTspice eliminated that option.
?
Andy
?
|
Re: Estimating Base spreading resistance for a bipolar transistor via LTspice
When temperature increases, the current resulting from the
reverse-biased gate-drain junction increases exponentially, so
yes, noise increases with temperature.
Le 25/02/2025 à 16:36, Jack Walton via
groups.io a écrit?:
toggle quoted message
Show quoted text
I am wondering whether the noise at higher current values
won't be affected by the warming of the device.? I know this to
be a limiting affect for JFETs.
|
Re: Estimating Base spreading resistance for a bipolar transistor via LTspice
I am wondering whether the noise at higher current values won't be affected by the warming of the device.? I know this to be a limiting affect for JFETs.
|
On 25/02/2025 12:55, eewiz via
groups.io wrote:
Tony, I was not out to prove that 2.000000000 volts never
happens.
I was out to prove that 2.000000000 volts did not happen in
my case.
So that I could therefrom propose that the test ".MEAS =
something FALL=1" in my case does not test for equality to
something because that something does not exist.
Until someone can show me proof of interpolation to obtain
the .MEAS result in my case, I will assume that it is actually
doing the more efficient test for a value that was more than
something and now has become less than something.
?
When it comes to plotting, that's a different story.
If your pixels are a certain size and your two data points
span more than two pixels then interpolate to find the missing
pixel coordinates.
But would you interpolate 100 new coordinates if you know you
only have to locate 14 pixels between the given data points.
No, that would be inefficient to crank an interpolation a 100
times to fix the coordinates of 14 missing points.
?
The same logic applies to using interpolation at all, when
the task at hand is to analyse the data stream to find any data
value greater than something followed by any data value less
than something.
That simply requires if(X <= something && priorX
>= something) to be true [C syntax not precisely correct]
which is considerably simpler than interpolation.
How would you know how many times to turn the interpolator
crank to find if any interpolated point could be found that was
exactly 2.00000000 volts so that it could be compared for
equality to 2.
?
I predict that the .MEAS statement never tests for equality.
Even in this example from the help file, I posit that:
.MEAS
TRAN res2 FIND V(out)*I(Vout) WHEN V(x)=3*V(y)
Print
the value of the expression V(out)*I(Vout) the first time
the condition V(x)=3*V(y) is met. This will be labeled res2.
Does not ever test that V(x) becomes exactly equal to 3*V(y).
It does this; "if((V(x) <= 3*V(y) && priorV(x)
>= 3*V(y) || (V(x) >= 3*V(y) && priorV(x) <=
3*V(y)))" == true" then it would be known that the data has
passed through the point 3*V(y) without ever having to resort to
a test of equality between floating point numbers.
Nobody ever said that LTspice tests for equality. It can't if the
data doesn't exist. LTspice uses interpolation to estimate it. I
thought that had been made clear. The closer the data points, the
better the estimation is likely to be. You can, of course, choose to
believe anything you wish to.
BTW, in case you hadn't realised, when using the waveform cursor,
pressing the → and ← keys advances or retards the cursor position to
the next actual data point. Any reading of the cursor position
between data points is also interpolated. Note that if you don't
have .option plotwinsize=0 enabled, not all actual analysis data
points are stored in the .raw file, and therefore cannot be
displayed or used for interpolation. The line display appears to be
a spline fit, but zooming-in shows this to be realised as a series
of linear approximations.
--
Regards,
Tony
|
I don't have a video card.
Mine is on the motherboard.
Intel UHD Graphics 750.
Sent:?Tuesday, February 25, 2025 at 5:21 AM
From:?"Tony Casey via groups.io" <tony@...>
To:[email protected]
Subject:?Re: [LTspice] .MEAS Failure
On 25/02/2025 10:53, eewiz via groups.io wrote:
I have 136GB free space on the drive where the .RAW file grows.
I did have Mark Data Points turned on which makes the plot window render many times slower.
?
Also, I can crash the plot window, taking LTspice 24.0.12 along with it, whenever I wish.
See
Close a window that overlays the plot window, as the plot window is rendering, and poof, there goes LTspice.
The plot window crashing is nothing new to me.
?
I agree also that increasing numdgts was not the cause of the crash.
Ah, OK. I seem to remember this issue might also be dependent on the graphics card/driver. There are also message threads from many years ago, before V24, that certain (Intel) CPUs were more susceptible to crashes. Helmut documented which types were most problematic, but I can't find that right now.
--
Regards,
Tony
|
Tony, I was not out to prove that 2.000000000 volts never happens.
I was out to prove that 2.000000000 volts did not happen in my case.
So that I could therefrom propose that the test ".MEAS = something FALL=1" in my case does not test for equality to something because that something does not exist.
Until someone can show me proof of interpolation to obtain the .MEAS result in my case, I will assume that it is actually doing the more efficient test for a value that was more than something and now has become less than something.
?
When it comes to plotting, that's a different story.
If your pixels are a certain size and your two data points span more than two pixels then interpolate to find the missing pixel coordinates.
But would you interpolate 100 new coordinates if you know you only have to locate 14 pixels between the given data points.
No, that would be inefficient to crank an interpolation a 100 times to fix the coordinates of 14 missing points.
?
The same logic applies to using interpolation at all, when the task at hand is to analyse the data stream to find any data value greater than something followed by any data value less than something.
That simply requires if(X <= something && priorX >= something) to be true [C syntax not precisely correct] which is considerably simpler than interpolation.
How would you know how many times to turn the interpolator crank to find if any interpolated point could be found that was exactly 2.00000000 volts so that it could be compared for equality to 2.
?
I predict that the .MEAS statement never tests for equality.
Even in this example from the help file, I posit that:
.MEAS TRAN res2 FIND V(out)*I(Vout) WHEN V(x)=3*V(y)
Print the value of the expression V(out)*I(Vout) the first time the condition V(x)=3*V(y) is met. This will be labeled res2.
Does not ever test that V(x) becomes exactly equal to 3*V(y).
It does this; "if((V(x) <= 3*V(y) && priorV(x) >= 3*V(y) || (V(x) >= 3*V(y) && priorV(x) <= 3*V(y)))" == true" then it would be known that the data has passed through the point 3*V(y) without ever having to resort to a test of equality between floating point numbers.
?
All for now
?
Sent:?Tuesday, February 25, 2025 at 4:28 AM
From:?"Tony Casey via groups.io" <tony@...>
To:[email protected]
Subject:?Re: [LTspice] .MEAS Failure
On 25/02/2025 08:53, eewiz via groups.io wrote:
Yes, I agree.
Rather than assume such was true, I sought to prove that in my case there actually was no 2.0000000000 volts.
In general, it is usually impossible to prove the non-existence of something.
Therefore the effect, however it may be accomplished, is to print the time at which the next sample found is LESS-THAN 2 volts.
?
Wouldn't it be a phenomnal waste of processor time to interpolate between points, following the resultant curve, to find a point on that curve that is exactly 2.0000000000 volts when simply looking for the first value of data less-than 2 volts provides the same result.
Hardly a phenomenal waste of processor power. Linear interpolation between two points is a trivially simple operation. The .MEAS already finds the last data point before, and the first data point after the target is reached. Interpolation means you always get a more accurate answer even when not using very small values of maximum timestep, with very little overhead.
?
Regardless of how it is accomplished, the end result of using FALL=1 is to print the time at which the first data value less-than 2 is found.
This is obviously not true. The end result of using FALL=1 is to print out the interpolated value.
The reason why you found that apparently two points at the same time had two different voltages is due the printed precision of the time points. The time points are printed in single precision in the logfile. You can get double precision by adding .option measdgt=15. This is true whether or not you also use .option numdgt=15.
For example, to find the first falling value of 0, I added:
.MEAS T1 when V(test)=0 fall=1
..when analysing a 0.999Hz sine wave with .TRAN 1.1. By default the logfile was:
?
LTspice 24.0.12 for Windows
Circuit: * C:\users\tony\Desktop\LTspice\_Temp\Eewiz\MEAS_example.asc
Start Time: Tue Feb 25 10:09:09 2025
solver = Normal
Maximum thread count: 32
tnom = 27
temp = 27
method = modified trap
Direct Newton iteration for .op point succeeded.
t1: v(test)=0 AT 0.500501
Total elapsed time: 0.054 seconds.
After adding .option measdgt=15:
?
LTspice 24.0.12 for Windows
Circuit: * C:\users\tony\Desktop\LTspice\_Temp\Eewiz\MEAS_example.asc
Start Time: Tue Feb 25 10:10:33 2025
solver = Normal
Maximum thread count: 32
tnom = 27
temp = 27
method = modified trap
Direct Newton iteration for .op point succeeded.
t1: v(test)=0 AT 0.500500582598
Total elapsed time: 0.044 seconds.
Note: this option is sticky. It remains active for the session (until you close and restart LTspice), even if it is removed. Whether this is intended or not is not known. AFAIK, it is not documented.
?
--
Regards,
Tony`
|
Ok, gracias, lo investigaré.
|
On 25/02/2025 12:26, Tony Casey wrote:
If
that doesn't work, there is a recent PDF version of the Help in
the Files section of the group's website: LTspice
Manual. You could download it onto your Desktop.
I mistakenly linked to an older version, but I don't think it had
changed much. Newer version: LTSPICE_Manual
--
Regards,
Tony
|
On 25/02/2025 12:09, j.bernabe1 via
groups.io wrote:
The F1 key does
work, a window opens, but nothing of what you're telling me
appears
Help > Schematic Capture > Creating New Symbols ?
For V17, the Help has an additional heading:
Help > LTspice > Schematic Capture > Creating New Symbols
If that doesn't work, there is a recent PDF version of the Help in
the Files section of the group's website: LTspice
Manual. You could download it onto your Desktop.
--
Regards,
Tony
|
The F1 key does work, a window opens, but nothing of what you're telling me appears
Help > Schematic Capture > Creating New Symbols ?
|
?
I'm using version 17.0.37.0
|
On 25/02/2025 11:44, j.bernabe1 via
groups.io wrote:
He abierto la ayuda
con F1, pero la ventana que se abre, no contiene nada.
What version of LTspice are you using?
The F1 key works on V17.1, V24.0 and V24.1. If you're using one of
those and it doesn't work, your installation may be missing the help
file.
--
Regards,
Tony
|
He abierto la ayuda con F1, pero la ventana que se abre, no contiene nada.
|
John Woodgate
Tony Casey
eewiz