Singular Matrix Error Sometimes in LTSpice

[SparkyTheEngineer]

Hi,

I've been looking all over the internet for an answer, but can't seem to find one for my problem.

I have a rather large circuit (switch mode power supply) that I'm simulating in LTSpice, but in order to get the simulation started, I must hit stop and start multiple times to get past the "Singular Matrix" Error. I've read into shunt resistors and all that, but I've tried placing shunt resistors down on my floating sections, and still get the same error. If I stop and restart the simulation, I can get past that error and simulate fine. I've tried global shunts, but it's my understanding that LTSpice does it for you by default. I could be wrong on that, though.

Any Suggestions?

 

[Inquisitive]

Upload the .asc file and schematic so members can investigate the issue in simulation

 

[SparkyTheEngineer]

I was trying not to... but alright. And are there any general steps in trying to address this issue?

 

[SparkyTheEngineer]

Anyone? It's starting to irritate me how it chooses randomly when and when not to simulate. Especially at a rate of 5 percent of the time

 

[MikeMl]

Every part of a LTSpice schematic has to have a path to Node Zero (GND) or the Ground triangle symbol, even if you are trying to simulate a transformer-isolated power supply. In other words, both the primary side has to have a path to GND, and the secondary has to have a path to GND, even if you put a 1GΩ resistance in series with the GND symbol...

The "singular matrix" error message is because you have a "floating" part of some circuit.

Try this simple circuit. Then delete R2 and try it again.

 

[SparkyTheEngineer]

I've tried placing 10 Mega Ohm resistors to all possible floating nodes to no avail

 

[MikeMl]

Then one of your sub-circuits may have a floating node

 

[eTech]

When I simulate it, It takes forever, but no singular matrix errors are produced. Plenty of defcon messages though...

 

[SparkyTheEngineer]

What are defcon messages?

 

[eTech]

What are defcon messages?

It's LTspice nebulous way of indicating that it's having trouble simulating the circuit. The message usually doesn't point to anything specific except a point in time.

Check the circuit structure. Check that values are reasonable, etc. Check node voltages and currents.

You can try looking at various node voltages/currents at the point in time the defcon message reports. Are there spikes with no rise/fall time?, etc..

Good luck..

 

[alec_t]

Can you define 'defcon'? Google is puzzled by it.

 

[SparkyTheEngineer]

Wish I had the 'defcon' messages, but all I get are the singular matrix errors.

I ended up taking out the LT1431 chip and started using the LT4430 to drive the optocoupler, which seems to work way better for me. I don't get the singular matrix errors as much, but I do still get them. I have a feeling that leaving the "COMP" pin floating on LT1431 was adding to my problem. Even though LT says to leave it floating...

So any one who is looking for a solution to this: try identifying and eliminating any possible floating nodes.

Thanks Guys!

 

[atferrari]

Wish I had the 'defcon' messages, but all I get are the singular matrix errors.

I ended up taking out the LT1431 chip and started using the LT4430 to drive the optocoupler, which seems to work way better for me. I don't get the singular matrix errors as much, but I do still get them. I have a feeling that leaving the "COMP" pin floating on LT1431 was adding to my problem. Even though LT says to leave it floating...

So any one who is looking for a solution to this: try identifying and eliminating any possible floating nodes.

Thanks Guys!

Ask your question in the Yahoo forum specific for LTSpice. I am sure you could get a concrete suggestion. Saved my time several times.

 

[simonbramble]

I see you have designed an isolated supply I don't have time to look, but do all isolated outputs return to the same ground? If not, they should do. Obviously in the spice world, isolation does not matter, so you need to have a ground referenced circuit on both (all) sides of the transformer, or otherwise the floating side has no 0V reference and LTspice gets confused. Also try Simulate -> COntrol panel - SPICE tab then select Alternate in the Solver box. This can unblock things

 

[simonbramble]

My simulation seems to work. The output has ramped up to 63V with no errors

 

[alec_t]

The post #3 sim works fine here. No error messages.