Re: creating gm/Id vs Id/W plot in ltspice
That depends.? Be more specific.
?
Andy
|
Re: when is a capacitor not (quite) a capacitor
Thanks Tony and Andy.? I'll try out GFARAD, sounds like it's the 'right' approach here.? I should study the wiki documentation for undocumented bits more closely.? Tomorrow I might try it on my home machine which is at 24.1.5.? What's interesting is the brittleness of the op point search in a circuit that contains only passives and amplifiers straight from ADI's install.? And Jerry I know schematics are helpful, but this is a bit proprietary and I would need to strip it down for upload here, which of course would probably eliminate the error... I did do a dump of problematic circuits to Mathias and Mike recently, I'm still bent out of shape by how easy it is to get LTspice to give an operating point with nodes that utterly fail KCL.? But one has to move on rather than b!tch about this generally fantastic free tool we all benefit from.? I just occasionally feel that the guys over there need a heads-up on things that aren't right.
|
Re: creating gm/Id vs Id/W plot in ltspice
Hello, nasickly I have Y1 vs X? and Y2 vs X ,Is there a wat to combine them so I'll get a Y1 vs Y2 plot?
Thanks.
|
Re: How to simulate the gate charge characteristic given in the datasheet of a mosfet?
On Fri, Apr 11, 2025 at 12:36 AM, Dennis wrote:
Ankit,
?
I will try to address your questions.
Dear Dennis,
?
A very good day to you.
Once again thank you so very much for answering all my questions. I am grateful to you, teacher. I apologise to you for replying with a delay.
?
I will make use of the knowledge that you have shared with me. Soon I will upload some simulation files for review.
With Regards,
Ankit
|
Re: when is a capacitor not (quite) a capacitor
On 15/04/2025 15:53, Andy I via
groups.io wrote:
On Tue, Apr 15, 2025 at 06:53 AM, Tony Casey wrote:
... Normally, unless otherwise specified, Rpar = ¡Þ
in capacitors. This shouldn't be affected by the Gmin setting,
because Gmin is only placed across junctions. The option for
placing a conductance between each node is Gshunt. The default
for Gshunt is zero.
I think that is not quite true, but it is one of those
"surprises" that you may not have realized was there.
?
The LTwiki ()
says that there is a non-zero conductance across capacitors,
with the name GFARAD and value GFARAD*C, and that the default
GFARAD is 1e-12 Siement/farad which happens to equal GMIN.? The
Help page for .OPTIONS says the same.? In any event, you do have
the ability to controls the parallel conductance across your
capacitors:
?
? ? .OPTIONS GFARAD=0? ?; disables it
?
The Help page says that the default GFARAD is 1e-12 and not
GMIN, but I wonder if its default value is really GMIN.? Note
that their units are not the same (one is Siemens and the other
is Siemens/farad) even though their values are the same.
?
In addition, depending on how you connected your series LC
circuits, it's possible that you created floating nodes with no
DC path to ground.? For example, a series LC trap with one C is
probably OK but if it had two C's in series, it has a floating
node where the capacitors join, and LTspice "fixes" that
user-mistake by adding a GFLOAT conductance there.? That might
be in addition to the GFARAD*C conductance, but LTspice might
combine the two instead of giving you GMIN + GFARAD*C.
?
GSHUNT is yet another thing, independent of all the above.
?
If the log file pops ups every time, it means you
have warnings or errors. At least that is the case with
24.0.12 or 24.1.4. I haven't moved on yet.
Yes, and I don't think that can be disabled, except by
eliminating all errors and warnings.
I actually checked this before I wrote the message. I had a
schematic with just a default 100pF capacitor grounded at one end. I
set the? initial conditions as IC=10.
I ran:
.TRAN 0 6000 0 10m
..and the capacitor voltage at the end was still exactly 10V.
I got the same result in 17.1.15, 24.0.12 and 24.1.4.
From that I concluded there was no shunt resistor, unless explicitly
specified.
However, I just checked again, and adding numdgt=15 revealed ~62nV
discharge at 6ks. Then adding GFARAD=0 eliminated this discharge.
62nV is less than the single precision resolution at 10V, so it
would not normally be seen. With GFARAD=0 and with a 1e10T shunt
resistor, the discharge again was ~62nV.
So, I concede LTwiki and the Help are correct.
--
Regards,
Tony
|
Re: when is a capacitor not (quite) a capacitor
Having posted a schematic (.asc) would probably have helped clearing the issue.
|
Re: when is a capacitor not (quite) a capacitor
On Tue, Apr 15, 2025 at 06:53 AM, Tony Casey wrote:
... Normally, unless otherwise specified, Rpar = ¡Þ in capacitors. This shouldn't be affected by the Gmin setting, because Gmin is only placed across junctions. The option for placing a conductance between each node is Gshunt. The default for Gshunt is zero.
I think that is not quite true, but it is one of those "surprises" that you may not have realized was there.
?
The LTwiki () says that there is a non-zero conductance across capacitors, with the name GFARAD and value GFARAD*C, and that the default GFARAD is 1e-12 Siement/farad which happens to equal GMIN.? The Help page for .OPTIONS says the same.? In any event, you do have the ability to controls the parallel conductance across your capacitors:
?
? ? .OPTIONS GFARAD=0? ?; disables it
?
The Help page says that the default GFARAD is 1e-12 and not GMIN, but I wonder if its default value is really GMIN.? Note that their units are not the same (one is Siemens and the other is Siemens/farad) even though their values are the same.
?
In addition, depending on how you connected your series LC circuits, it's possible that you created floating nodes with no DC path to ground.? For example, a series LC trap with one C is probably OK but if it had two C's in series, it has a floating node where the capacitors join, and LTspice "fixes" that user-mistake by adding a GFLOAT conductance there.? That might be in addition to the GFARAD*C conductance, but LTspice might combine the two instead of giving you GMIN + GFARAD*C.
?
GSHUNT is yet another thing, independent of all the above.
?
If the log file pops ups every time, it means you have warnings or errors. At least that is the case with 24.0.12 or 24.1.4. I haven't moved on yet.
Yes, and I don't think that can be disabled, except by eliminating all errors and warnings.
?
Andy
?
|
Re: when is a capacitor not (quite) a capacitor
On 14/04/2025 23:03, Kendall
Castor-Perry via groups.io wrote:
Hi all - anyone seen or been bitten by this?? I'm working on
a middling-sized circuit, nothing fancy, all passives and
library ADI opamps, inamps and diff drivers.? Implementing a
little gain and (as you'd expect from me) some filtering
functions, including some series LC traps close to the inputs.?
It's one of those circuits that's on the edge of not finding a
DC op point (24.0.12 on this work machine, 24.1.x deffo not
ready for salaried work IMHO).? Series traps are from signal
nodes to ground.? So there's a capacitor on the branch between
that node and ground.? So in principle it cannot participate in
any DC op point calculations.? But...
?
I get to the point where, slowly dropping down the inductor's
series resistor (most tests with internal but external seemed
similar), I would cross a threshold where it couldn't find an op
point.? Quite abruptly; for instance, Rs >=65 gave an op
point, Rs <=64 and lower did not. (alternate solver; didn't
test normal).? So, clearly, the NAM must have some DC path
across that cap.? Tried changing the Gmin setting from default
1E-12 to 1E-14.? Success!? Now all realistic configurations give
me an operating point.
?
This is something that could be fixed (hi @mstokowski and
@mborn) - by making sure that Gmin==0 across any capacitor
that's not otherwise set up to be different, _just_ during the
op point search.? Of course the DC solver needs to be robust to
hanging branches so there might need to be some topology
trim/prep, solving a slightly different matrix (the caps in
series case will probably be irritating).? So not trivial, just
simple... I have a sneaking feeling that some of my older op
point troubles could have been fixed up this way.
?
Another way of doing this would be to have a flag for a
capacitor to explicitly set Re (shunt admittance) to zero - but
that's much more work for the user.? Better for the code to do
it inherently (perhaps with a setting choice box).
?
Of course 24.1.5+ might already have fixed this, I can't try
that until at least this evening.
?
Hope this is interesting and relevant -- K
?
PS (1) how do you stop the .log file from popping up at the
end of each sim now, annoying and (2) where did the little
display of the key for which .step plotted to which color trace
go?
Something here is not quite right. Gmin is not a conductance across
all nodes. Normally, unless otherwise specified, Rpar = ¡Þ in
capacitors. This shouldn't be affected by the Gmin setting, because
Gmin is only placed across junctions. The option for placing a
conductance between each node is Gshunt. The default for Gshunt is
zero.
My guess is that the Gmin setting in you case is a red herring. It
may be affecting any models that have active devices, but presumably
your circuit has no convergence issues until the resonant traps are
added?
If the log file pops ups every time, it means you have warnings or
errors. At least that is the case with 24.0.12 or 24.1.4. I haven't
moved on yet.
--
Regards,
Tony
|
Re: AC Analysis - show Mag in dBuV versus Freuqncy
Thank you, but that crazy filter is not my
design.
On 2025-04-15 03:10, Michael Fortner
via groups.io wrote:
Hello John,
The land of dB can be confusing at first.? When dB was
invented, it was a way to keep track of power, as in watts.?
Each +10dB was 10 times as much power.
Recall that power, P = V^2 / R.? Since power increases to the
square of the voltage, 10 x voltage is +20dB.
The display defaults to reading 0dB as 1 volt.? For .001V, you
go down -20 dB for each decade or -60dB.? Going to a
microvolt, 1E-6 V, you get -120dB.? By the way, using the 1EX
convention is easy as you don't have to recall any symbols.
As to what level you're using, it usually doesn't matter,
because AC analysis assumes everything is linear.? So, you can
inject 1 volt and pretend it's 1uV and that corresponds to
0dB.? The solver won't care :)
As to parasitic aspects, it is good to remember the DC
resistance of the inductors, and better to find the resonance
of the inductor and place an equivalent C across it.? Good
fences build great neighbors, but including your parasitic
values makes great models.
Most of the ceramic capacitors are on the order of 2nH, and it
increases with the physical size of the capacitor.? I
construct filters on the order of 10's of MHz to 1GHz, so the
series inductors on caps become very important.? For 1MHz and
under, you can probably ignore such tiny values.
? ? I wish you the best on your model and circuit,
? ? Mike
On Monday, April 14, 2025 at 06:22:02 AM CDT, John
Woodgate <jmw@...> wrote:
This is very OT, but dBm is very
widely applied to 50 ohm circuits in the RF field.
On
2025-04-14 12:10, Jerry Lee Marcel via groups.io
wrote:
No, it's not. dBm applies only to 600 ohm
circuits. If you want independant on load, you must
use dBu (or dBV).
--
Best wishes John Woodgate RAYLEIGH Essex 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
--
Best wishes John Woodgate RAYLEIGH Essex 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: when is a capacitor not (quite) a capacitor
The "annotate steps" is in the "notes and annotations" menu you get after right clicking in the results menu after a successful step'd simulation.?
Cheers
Ian
|
Re: AC Analysis - show Mag in dBuV versus Freuqncy
Hello John,
The land of dB can be confusing at first.? When dB was invented, it was a way to keep track of power, as in watts.? Each +10dB was 10 times as much power.
Recall that power, P = V^2 / R.? Since power increases to the square of the voltage, 10 x voltage is +20dB.
The display defaults to reading 0dB as 1 volt.? For .001V, you go down -20 dB for each decade or -60dB.? Going to a microvolt, 1E-6 V, you get -120dB.? By the way, using the 1EX convention is easy as you don't have to recall any symbols.
As to what level you're using, it usually doesn't matter, because AC analysis assumes everything is linear.? So, you can inject 1 volt and pretend it's 1uV and that corresponds to 0dB.? The solver won't care :)
As to parasitic aspects, it is good to remember the DC resistance of the inductors, and better to find the resonance of the inductor and place an equivalent C across it.? Good fences build great neighbors, but including your parasitic values makes great models.
Most of the ceramic capacitors are on the order of 2nH, and it increases with the physical size of the capacitor.? I construct filters on the order of 10's of MHz to 1GHz, so the series inductors on caps become very important.? For 1MHz and under, you can probably ignore such tiny values.
? ? I wish you the best on your model and circuit,
? ? Mike
On Monday, April 14, 2025 at 06:22:02 AM CDT, John Woodgate <jmw@...> wrote:
This is very OT, but dBm is very widely
applied to 50 ohm circuits in the RF field.
On 2025-04-14 12:10, Jerry Lee Marcel
via groups.io wrote:
No, it's not. dBm applies only to 600 ohm circuits. If you
want independant on load, you must use dBu (or dBV).
--
Best wishes John Woodgate RAYLEIGH Essex 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
|
when is a capacitor not (quite) a capacitor
Hi all - anyone seen or been bitten by this?? I'm working on a middling-sized circuit, nothing fancy, all passives and library ADI opamps, inamps and diff drivers.? Implementing a little gain and (as you'd expect from me) some filtering functions, including some series LC traps close to the inputs.? It's one of those circuits that's on the edge of not finding a DC op point (24.0.12 on this work machine, 24.1.x deffo not ready for salaried work IMHO).? Series traps are from signal nodes to ground.? So there's a capacitor on the branch between that node and ground.? So in principle it cannot participate in any DC op point calculations.? But...
?
I get to the point where, slowly dropping down the inductor's series resistor (most tests with internal but external seemed similar), I would cross a threshold where it couldn't find an op point.? Quite abruptly; for instance, Rs >=65 gave an op point, Rs <=64 and lower did not. (alternate solver; didn't test normal).? So, clearly, the NAM must have some DC path across that cap.? Tried changing the Gmin setting from default 1E-12 to 1E-14.? Success!? Now all realistic configurations give me an operating point.
?
This is something that could be fixed (hi @mstokowski and @mborn) - by making sure that Gmin==0 across any capacitor that's not otherwise set up to be different, _just_ during the op point search.? Of course the DC solver needs to be robust to hanging branches so there might need to be some topology trim/prep, solving a slightly different matrix (the caps in series case will probably be irritating).? So not trivial, just simple... I have a sneaking feeling that some of my older op point troubles could have been fixed up this way.
?
Another way of doing this would be to have a flag for a capacitor to explicitly set Re (shunt admittance) to zero - but that's much more work for the user.? Better for the code to do it inherently (perhaps with a setting choice box).
?
Of course 24.1.5+ might already have fixed this, I can't try that until at least this evening.
?
Hope this is interesting and relevant -- K
?
PS (1) how do you stop the .log file from popping up at the end of each sim now, annoying and (2) where did the little display of the key for which .step plotted to which color trace go?
|
Re: AC Analysis - show Mag in dBuV versus Freuqncy
This is very OT, but dBm is very widely
applied to 50 ohm circuits in the RF field.
On 2025-04-14 12:10, Jerry Lee Marcel
via groups.io wrote:
No, it's not. dBm applies only to 600 ohm circuits. If you
want independant on load, you must use dBu (or dBV).
--
Best wishes John Woodgate RAYLEIGH Essex 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: AC Analysis - show Mag in dBuV versus Freuqncy
No, it's not. dBm applies only to 600 ohm circuits. If you want independant on load, you must use dBu (or dBV).
|
Re: AC Analysis - show Mag in dBuV versus Freuqncy
To be very precise, 0.775 V is approximately equivalent
to 1mW in 600 ohms, which is sqrt(600x0.001) = 0.774596...
That is the explanation of the voltage value.
On 2025-04-14 11:39, Goran Finnberg via
groups.io wrote:
John Woodgate:
>dBu? is decibels referred to 0.775 V, which is 1 mW in
600 ohms.
0
dBU is not 1mW in 600 ohms.
0 dBU is 0,775 V irrespective of impedance.
0 dBm is 1 mW in 600 ohms.
--------------------
Best regards,
Goran Finnberg
The Mastering Room AB
Goteborg
Sweden
E-mail: mastering@...
Learn from the mistakes of others, you can never live long
enough to
make them all yourself.??? -?? John Luther
(\__/)
(='.'=)
(")_(") Pyret, Ranglet, Aron, VovVov, Nero, Smurfen &
Pussin:RIP
--
Best wishes John Woodgate RAYLEIGH Essex 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: AC Analysis - show Mag in dBuV versus Freuqncy
Bruce:
>0 dBm is 1mW independent of the load resistance. https://en.wikipedia.org/wiki/DBm
In audio and telephony, dBm is typically referenced relative to the 600-ohm??commonly? used in telephone voice networks, while in radio-frequency work dBm is typically referenced relative? to a 50-ohm impedance.
-------
Best regards,
Goran Finnberg
The Mastering Room AB
Goteborg
Sweden
E-mail: mastering@...
Learn from the mistakes of others, you can never live long enough to
make them all yourself.??? -?? John Luther
(\__/)
(='.'=)
(")_(") Pyret, Ranglet, Aron, VovVov, Nero, Smurfen & Pussin:RIP
|
Re: AC Analysis - show Mag in dBuV versus Freuqncy
0 dBm is 1mW independent of the load resistance.
?
Bruce
On 14/04/2025 22:39 NZST Goran Finnberg via groups.io <mastering@...> wrote:
?
?
John Woodgate:
?
>dBu? is decibels referred to 0.775 V, which is 1 mW in 600 ohms.
? 0 dBU is not 1mW in 600 ohms.
?
0 dBU is 0,775 V irrespective of impedance.
?
0 dBm is 1 mW in 600 ohms.
?
?
--------------------
Best regards,
Goran Finnberg
The Mastering Room AB
Goteborg
Sweden
E-mail: mastering@...
Learn from the mistakes of others, you can never live long enough to
make them all yourself.??? -?? John Luther
(\__/)
(='.'=)
(")_(") Pyret, Ranglet, Aron, VovVov, Nero, Smurfen & Pussin:RIP
|
Re: AC Analysis - show Mag in dBuV versus Freuqncy
Le 14/04/2025 ¨¤ 01:02, Zivko Petru via groups.io a ¨¦crit?:
In LT spice AC analysis You are instructed to use 1Vac as input.
Actually, 1Vac is meaningless. Could be 1V rms, 1V peak or 1V peak-to-peak. The dB value is always referred to the input voltage, whatever the unit.
In order to have the results expressed in dBuV, the amplitude must be 1u.
?
?
dbu on the other hand is referenced to 1 uV.
dBu is an audio measurement reference, corresponding to an rms value of 0.775V
dB referred to 1uV is dBuV
|
Re: AC Analysis - show Mag in dBuV versus Freuqncy
John Woodgate:
>dBu? is decibels referred to 0.775 V, which is 1 mW in 600 ohms. 0 dBU is not 1mW in 600 ohms.
0 dBU is 0,775 V irrespective of impedance.
0 dBm is 1 mW in 600 ohms.
--------------------
Best regards,
Goran Finnberg
The Mastering Room AB
Goteborg
Sweden
E-mail: mastering@...
Learn from the mistakes of others, you can never live long enough to
make them all yourself.??? -?? John Luther
(\__/)
(='.'=)
(")_(") Pyret, Ranglet, Aron, VovVov, Nero, Smurfen & Pussin:RIP
|
Re: Creating a dirt-simple subcircuit component
FYI - not a bad idea to use a custom,?small symbol for a 0 V voltage source, to use either as an ammeter or a net isolator.? Other group members did that already.? Easy to make your own.? Of course it adds the overhead of a custom symbol, which you need to remember to include when sending someone your schematic.
?
Andy
|
Tony Casey
John Woodgate
Ian McCrum MI5AFL