# LTspice simulation problem - energy harvesting rectifier circuit

SO the circuit below is an approximation to an electrostatic energy harvesting system and device. Basically I want to observe the charging voltage waveform across the cap C3. In real life the cap charges in a few hours but I do not think LTspice likes completing that many iterations without converging on a steady state condition. It just either stops working after a while or gives a 'singular matrix' error (after the first iteration)depending upon the value I use for the cap C3.

Is there anyone that is handy with Ltspice that could tell me how to make it like this sort of transient analysis?

Perhaps this system just has too high source impedance for LTspice?

• Try with a real diode model instead of ideal ones Oct 2 '16 at 9:55
• One theory I have now while trying out your circuit is that LTspice perhaps doesn't handle longer timespans. It might be problems with floats/doubles if it has to work with nanosecond delta-times when the actual time is in the thousands. Adding more precision didn't help though, so maybe it's a flawed theory.
– pipe
Oct 2 '16 at 10:41

Try with a real diode model instead of ideal ones – PlasmaHH

I tried your setup, and I got the same problem. After replacing the diodes with "real" diodes, everything works:

The diodes were somewhat arbitrarily chosen from the first model that could withstand the voltages involved.

• It works with the default diode, too. For example with this model: .model d d ron=1 roff=100meg vfwd=0.7 vrev=1k epsilon=100m revepsilon=50m it charges close to 190V. Oct 2 '16 at 13:06
• @aconcernedcitizen I'm not sure I follow. You say it works with "the default diode", then you supply a completely customized model.
– pipe
Oct 2 '16 at 13:07
• The default model is the ideal one, so modifying it is nothing more than using the same default model but with different parameters. I suspect this may be a bit of a linguistic problem, as well, as I am using the meaning that's close to me. Maybe I should've said "default model" instead of "default diode". Oct 2 '16 at 13:11
• Indeed, the default diode is the one you get when you simply add a diode, as OP did, and it is not really "ideal".
– pipe
Oct 2 '16 at 13:16
• True, it is an idealized approach, "a linear model", "computationally light weight" (as per the manual), as opposed to the much evolved Berkeley-SPICE model. Oct 2 '16 at 13:27