Forums » CircuitLab Support Forum

I have left a circuit "Trouble2" as a public circuit.

Its a FET-BJT amplifier I have used before. SPICE gives me some really weird DC simulations. Anybody know why?

CircuiLab is having trouble finding the DC initial conditions.

Here are two ways to fix that problem for a transient analysis.

Both also shows that your circuit oscillates.

Note the use of a behavioural source to start the circuit up from Vcc = 0V

And:

which sets an initial condition on C1.

I haven't managed to get either version to run an AC simulation to show the peaking in the frequency response.

Hmm, yes, something is really wacky, it shows 2 volts emitter-base, which is impossible. I've always wanted realistic simulations, with black smoke and a bad burned phenolic smell.

The capacitor shouldn't have anything to do with a DC simulation anyway.

It also shows 65 watts being dissipated in R3, but under a watt being drawn from the power supply. Congrats, you've invented an infinite energy source.

If you do:

Simulate > Time Domain > Run Time-Domain SImulation

on either of

or

then the values of Vbe of Q1 and the current through R3 are reasonable.

"The capacitor shouldn't have anything to do with a DC simulation anyway."

This is only true if the solver isn't having a bad day.

I've given it a good talking to in my versions of the circuit and it has decided, at least for a transient simulation, that it will sort itself out.

Clearly it is still having a bad day in AC analysis space and I haven't tried asking it for a DC operating point. Then again, if it can't find the correct DC operating point then it won't be able to do a sensible AC analysis.

When I first tried a transient sim on the original circuit, it was showing voltages of in some parts of the circuit. around 180V

:)

@signality,

Thanks... I guess I am an old die-hard when it comes to my tools...

Maybe the guys here at CircuitLab can see if the model for the FET is right. I get the strong idea its some little hiccup in the model definition that's causing this.

In the real world, many IC's have startup circuitry to get things rolling. Soe IC's that didn't do this depended on the dv/dt of the power supplies to get things perking.

In simulation you sometimes have to somehow jolt the circuit into operation. My favorite is a 22pF capacitor hooked up to a 10 volt square wave generator. If the circuit simulates as dead until the first or second edge of the square wave, but then perks up, you know you have an an initial condition situation.

I've added some tags to to help search for several examples of gotchas and ways to fix initial conditions, startup and convergence problems:

startup

convergence

intitial-conditions

Hi @anubi, just wanted to let you know that we do recognize this DC convergence problem as a bug in CircuitLab and will be working to resolve it. I will post again in this thread when I have an update.

@mrobbins Thanks! and thanks for access to a great site!

Its not a trivial thing to model circuits - and electronic devices are not ideal. Good luck tracing the convergence bug. I have several in my account that fail to converge - I'll flag them with the tags signality provided - feel free to test with them.

My mistake ...

the tags should be:

startup

convergence

initial-conditions

@anubi: have you put the tags on and made your circuits public?

Ok... I put the startup, convergence, and initial-conditions tags on "Trouble1" and "Trouble2" and made them public. There are almost identical circuits I am playing with still "private". Nothing new there.

I was trying to recreate an old circuit I used back in the late 60's in the oilfield to boost geophone signals. When I noted Spice having problems with the big one, I simplified the core feedback loop as "Trouble2" to see if I could understand more where the problems were coming from.

I have a few more circuits in there I am playing around with. Most are intermediate steps to the designs I put as public, as I did not want to put out half-baked stuff.

I guess I am hoping your site will become an online equivalent of Electronic Design magazines "Ideas for Design", where someone looking for example circuits can get a few ideas on how to get started building something.

Kinda like Markus "Sourcebook of Electronic Circuits" was for me in the 60's.

We just deployed new changes to the JFET model and overall simulator convergence code, which hopefully should help out here.

@anubi's "Trouble2" now converges and amplifies:

The larger "Trouble1" converges, but doesn't amplify: It now appears that Q1 and Q2 are saturated.

Thanks!

Its now working. I have adjusted the bias to bring Q1 and Q2 into the linear region. I have renamed the circuit to what it is supposed to be... Geophone Amplifier.

It runs great and acts just like the ones I made back in the early 70's. Its an old circuit. Its seen its day. It may still be useful as an example of differential circuit design.