Maker Pro
Maker Pro

LTSpice: Slow on noise sims with foreign opamps

J

Joerg

Folks,

Yahoo has made the sign-in procedure so onerous (can't get in anymore)
that I'll ask here instead of the LTSpice group:

I want to do some noise simulations, for example with a TI opamp in
there. Loading the complete model it spends forever in the stepping
phase, then pseudo-damped analysis and whatnot.

Is there any way this can be sped up? All I need to see is ballpark
where the 1/f knees will be with various circuit constellations, input
current noise, input voltage noise, stuff like that.
 
J

Joerg

Jim said:
What? You refused to identify yourself with a Username and Password
?>:-}

I did ID myself. But now they slapped on a stupid "captcha". Which
promptly resulted in "Can't identify the machine bla-bla-bla, followed
by a challenge question to which they cannot have the answer.

IOW, they are on the path of oversophisticating the Yahoo groups to death.

And there is an E-mailed Digest... quite useful.

But can you probably can't ask questions via that.

Sounds like you don't know how to do a dotAC plus Noise analysis.

I don't do noise analyses much, that's usually done on the bench here
but in this case it should be sim'd.

Which OpAmp?

OPA4140.

[...]
 
J

Joerg

Jim said:
Do we live in the same universe? I've never experienced that.

Tried numerous times, always the captcha. Those things are nasty but
after four tries it let me past the virtual bouncer. Wot a pain.

Sure you can. I participate several times a week. They actually
discuss engineering there >:-}

I know they do. But so far I haven't figured out a way inside without
that dreaded Yahoo GUI. So where's the back door and how many knocks?
I want to do some noise simulations, for example with a TI opamp in
there. Loading the complete model it spends forever in the stepping
phase, then pseudo-damped analysis and whatnot.
Sounds like you don't know how to do a dotAC plus Noise analysis.
I don't do noise analyses much, that's usually done on the bench here
but in this case it should be sim'd.

Which OpAmp?
OPA4140.

[...]

Set-up how? Feedback, bias, supplies, etc.

Essentially I've plopped the whole Texas model in there. Right now
supplied +/-5V, for the test stripped back to a follower with just 10k
OUT to IN-, IN+ to 0V (ground).

With any LTC models it zips right through even with complicated stuff
around them, as usual.
 
J

Joerg

Joe said:
You can try to email to [email protected] but the email has to
come from your yahoo email id or from the id where you get the email
digest or individual messages. Or, just create a new account on
yahoo and sign up for the ltspice group again.

If all else fails I'll post this for you but you know the first
thing they are going to want is an example circuit.

I've got the circuit working, it's just that it doesn't seem to like
Texas models. No, no, not those :)
 
J

Joerg

Jim said:
Bwahahahahahaha! The OPA4140 Spice model provided by TI is what gives
model-makers a bad name... the sort of model you get when you take a
PhD with no hands-on-lab experience and sit them down in front of a
simulator.

I don't think I've ever seen such a badly constructed model before. No
wonder it takes forever to converge.

Kind of my hunch, but I hadn't thought it was this bad :-(

Burr-Brown used to be a pretty good bunch of engineers, real hands-on guys.

Looking at the bright side it did manage to heat my office by one degree
Fahrenheit over lunch.
 
J

Joerg

Jim said:
Hmmmmmm? Maybe cockpit error? In PSpice, Q-point found as fast as
you can blink:

**** 01/04/13 17:12:37 ******* PSpice 15.7.0 (July 2006) ****** ID#
xxxxxxxx *

* C:\Projects\Expments\OpAmps\OPA4140.sch


**** JOB STATISTICS SUMMARY


******************************************************************************



Total job time (using Solver 1) = .27 <<<<<<<<<<<

See...

http://www.analog-innovations.com/SED/OPA4140_Q-Point.pdf

What set-up do you want?

Noise info for what frequency range?

I'll have to include it in a measurement set-up and play around with
component values to find a good compromise. Possibly they have included
something that LTSpice does not like. I doubt cockpit error because
various other opamps run blink-eye speed here as well. Plus I did not
have a sip of the Fin du Monde ale yet, that's for later tonight.

From a noise point of view I think we can do a tad better than the 4140
anyhow. It's mostly up to a coupole hundred keelohoitzes in this case,
no RF stuff.

Thanks for testing it. In case you want to try: When you hang a 10k in
the FB and feed a current into IN-, then runs .NOISE, does it run on
your sim? If yes then it's highly likely that the model doesn't fit LTSpice.
 
J

Joerg

Jim said:
I think I know the problem...

I did this...

.SUBCKT OPA4140 +IN -IN Vout V+ V- ; -IN +IN V- V+ Vout (Rearranged
Order to Match PSpice Symbol)

You need to make sure the "pin" order matches the LTspice OpAmp
symbol... otherwise you will experience pain and agony.

(The TI pin order is just a wee bit unusual. I do this rearrangement
exercise almost daily, so it's second nature to me.)

Then I get this:

Circuit: * E:\Simulations\PGL\PosSense_AC_1.asc

Expanding empty subcircuit: opa4140
WARNING: Less than two connections to node V+. This node is used by V1.
WARNING: Less than two connections to node V-. This node is used by V2.
Direct Newton iteration for .op point succeeded.

Date: Fri Jan 04 17:14:43 2013
Total elapsed time: 0.047 seconds.

tnom = 27
temp = 27
method = trap
totiter = 6
traniter = 0
tranpoints = 0
accept = 0
rejected = 0
matrix size = 17
fillins = 0
solver = Normal
Matrix Compiler1: 32 opcodes
Matrix Compiler2: 967 bytes object code size
 
J

Joerg

Rich said:
No captcha here, either. I'm not a "member" of the LTSpice Yahoo!
(can't forget the '!') group but am of a couple of others in the
tech.groups.yahoo.com tree. Maybe something with your browser (is
javascript on? needed?) or cookie settings?

If not, it may be that somebody (or some algorithm) has decided that
your login is being used by a 'bot, since 'bot blocking is the usual
reason for putting a captha in-line with the login.


Maybe because I leave myself logged in until it throws me off. Usually
lasts a few weeks. I really don't like log-in hassles. Why can't they
just set a cookie and be done with it?
 
J

Joerg

Jim said:
They can. You probably have it blocked. Mine works.... Firefox
v17.0.1

I always turn cookie on when I look onto the, it's sitchable.
 
J

Jasen Betts

Folks,

Yahoo has made the sign-in procedure so onerous (can't get in anymore)
that I'll ask here instead of the LTSpice group:

I want to do some noise simulations, for example with a TI opamp in
there. Loading the complete model it spends forever in the stepping
phase, then pseudo-damped analysis and whatnot.

Is there any way this can be sped up? All I need to see is ballpark
where the 1/f knees will be with various circuit constellations, input
current noise, input voltage noise, stuff like that.

If you want to play with TI psrts perhaps try the free TI spice program
"TINA"
 
M

miso

Usually if spice is choking, there is something wrong with the circuit.
Now you can't float single nodes in spice, but you can have nets that
should be tied to something that might be floating.

I would look carefully at the connectivity of the subcircuit.
 
J

Joerg

Jim said:
Huh? What does...

"Direct Newton iteration for .op point succeeded." <<<<<<

... mean?

Nothing in this case because LTSpice stopped. Plus it normally never
opens such message windows unless it is unhappy with something. The
usual waveform window never opened and that's a clear sign that it ain't
happy.
 
J

Joerg

Jasen said:
If you want to play with TI psrts perhaps try the free TI spice program
"TINA"

Well, I don't want to learn a new CAD everytime a vendor thinks they
absolutely have to boil their own proprietary broth. The stuff I do with
it often gets to be complicated and that would take a long learning
curve every time.
 
J

Joerg

Jim said:
Nope. It means that it DID find the OPerating point.

But no workie.

You probably don't have the .AC set up properly. I've just posted
another pass.


So how come other opamp model work fine with exactly the same setup?
 
S

SoothSayer

That's what I get with the other opamps (in my case nV/rtHz because
that's what I want). I guess the model works in PSpice but for some
reason it doesn't in LTSpice


Parasitics not properly included in the faulty engine?
 
F

Fred Bartoli

Joerg a écrit :
But no workie.




So how come other opamp model work fine with exactly the same setup?

Hey Joerg, are you the Bear of spice ?-)
 
Top