simulation interrupts, error:timestep too small, trouble with node. Solution?

[Stepan]

Evaluating Multisim, a simple half bridge inverter, using function generator to control IGBTs, scopes and current probes. Simulation starts and interrupts shortly. The error message is: Timestep too small, time=xxxx, timestep=yyyy, trouble with node zzzz.

Can anyone give me some idea please, what is going on here,  am I doung something wrong or this is software/model bug?

Simulation is unstable, even when it works, if I change the point where I sense the current or voltage, or component value, it fails.

Any ideas would be appreciated.

Thanks.

[J POWER]

Hi Stepan,

Thank you for posting your issue, without seeing the circuit I can only suggest checking all the wires are connected properly (no wires shorting out or components in the wrong place or polarity).

but would it be possible for you to post a screen shot or the code, along with any full error messages that you are given?

Regards

[Stepan]

Hi Jordan,

Thanks for replying to my message.

Please see attached the schematics, the scope trace (interrupted by simulator) and the error message, which reads:

Timestep too short, time=0.000198866, timestep=1.25e-017, trouble with node vfgen_src_negative:xfg2#branch.

If I change the load resistor value to say 4 Ohm the simulation does not interrupt and runs infinitely, but if the resistor value is at 100% it will fail. Same happens if I change the current probe to another point.

Regards,

Stepan.

[J POWER]

Hi Stepan,

Just looking at your screen shot and i can see that you haven't connected the negative terminal from your voltage gererater to ground, this may solve that error.

and a seggestion would be to ground all terminals that are not being used, but try the first step, run the simulation so we can cross that issue off the error list.

Regards

[Stepan]

Hi Jordan,

Thanks for advise, but connecting the terminal to ground does not solve the problem. The problem is not in the circuit but in the software. You can see that the simulatiion worked for several periods, so the circuit is OK. Please tell me, what is the exact meaning of the error message, "timestep too small"? Why is the software working for one value of the component and not for another value? My interpretation is that when software encounters some values during calculations that it does not like for some reason, it gives this message without properly defining the problem so that the user can take practic action to rectify it. Are there any settings which I must define before starting the simulatioin, like sampling rate or anything that would eliminate "timestep too small" message?

I am evaluating the simulator software to make a decision whether it is practically useful for the company to purchase. If 8 out of 10 times the result of simulation is an incomprehensible error message instead of a waveform, then I have to assume that the software is not mature.

Please advise.

Best regards,

Stepan.

[J POWER]

Hi Stepan,

Here is a link that shows three steps to try to get your code to run:

 http://digital.ni.com/public.nsf/allkb/4B99B2CD6C0C3B6A86257205005D58E0

hope these sort out your problem,

Regards

[lacy]

O.K. You may not think it is the circuit, but 95% of the time it is. Here is my take on this as I have run into simular problems like this before,. Look at the way the potentiometer is wired in the circuit. Once the pot reaches the lowest point it is actually reading zero ohms across it.  Because the pot say 100% doesn't necesarrly mean 100% full resistance. It depends on how it ias wired into the circuit Therefore, with a short circuit across the diodes, what do you think is going to happen? The simulator is going to flag this error and "timestep too small" is it's way of doing that.

Look in the simulation audit/error log and see if it doesn't say something about a "singular matrix" somewhere in the circuit. This may be what is causing this, but without the MS10 file this is just speculation.

Also check T1 as the secondary is ungrounded. The transformers in MS do night like to be ungrounded. Both sides haver to have a ground of some sort.

Message Edited by lacy on 11-01-2008 07:29 PM

[Stepan]

Hi Lacy,

What you have explained is all correct. There is no singular matrix and the pot is not shorting, trust me, I am doing circuits since 1970. The simulator works for some value combinations, which means that the connections and values are OK, then I change the input frequency by 10%, the simulator fails. I do not think this is a useful behaviour, I would like to see the circuit behaviour at this specific frequency. A good software is one that positively helps the user to progress with the work. Great deal of effort has been put in this software but the way it fails makes it useless.

Thanks for trying to help anyway.

Regards,

Stepan. 

[lacy]

The pots in Multisim are weird (at least to me). I do not beileive they should read zero ohms at the low end. I am going to post an example circuit of what I am seeing with this situation. Like you said, it may not have any bearing on your circuit, but I think it needs NI examine it and determine if they think this may be a problem or not.

I am not sure what frequency you are running, but MS does seem to have problems with extremely high frequencies. This may also cause the timestep error also. You can adjust this to a cetain extent. Like say if tyou are running it at 1E-005 you can mchange that to 1E-007. This will make the simulator run slower, but it may be enough to eliminate the timestep error. I refer to this "the frequency is to fast for the math" meaning that the number crunching that the simulator is too slow for the frequency used. This may help and it may not.

O,K. I have ehausted what I know on this situation. I will post my potentiometer circuit for everyone to look at and if they want give an opinion.

[Stepan]

Hi Lacy,

Your first guess was right, connected the secondary to earth and it worked, I did not realise that earth is

that important to MS. The frequency is not high only 100Khz.

The pot example did not show anything on my machine, in my library the multimetre

does not measure resistance, only voltage and current, don't know why, another feature of MS?.

Thanks again for being helpful.

Best regards,

Stepan.