I have been working with ngSpice circuit simulator for a couple of months now and have had to give up because of bad (noisy and glitchy) simulation results. I don't feel like I can trust any of the results I have gotten. I finally attributed this to floating point roundoff errors. My circuit has atto amp to milli amp currents. This wide range is probably what is causing the errors. The best any PC can do is 64 bit floats. In order to prove that my problem is floating point roundoff I am trying to find a site where I can run a Spice simulation that can handle 128 bit floats. Does anyone know of such a site?
I thought about how I might be able to scale things but decided that was not going to work in my case. I am deriving my atto amp to milli amp current from a BJT and an FET characteristic. I decided that I would not be able to scale because I need to directly use the BJT and FET to produce the current using the BJT and FET inherent parameters. This is not a normal simulation in the sense that I am operating the BJT and FET way out of their normal operating conditions. But you have encouraged me to look at the possibility of scaling again. Thank You.
Spice, just to avoid floating nodes, connect a resistance to each node. This value, (typical 10^12) can be changed in the simulator setting, as you can change the abstol and reltol of the DAE solver. Since you work with such a high range of current these parameters can affect your simulation.
I do not know ngSpice and how to change these parameters in it, but it is quite easy to do it in LTSpice. Do you have tried another simulator ?
Greetings
S.
My circuit has a resistor that is 10^16, which is difficult to simulate and may not even be practical to produce. So I am unable to put other Rs around the circuit without ruining the integrity of the circuit. I think I will ultimately need a different design. I am stuck with it for now. But Thank You.
About the 10^16 Ohm resistor, if you mount your circuit in a PCB you will have big fights with surface resistivity, either in FR4, PTFE or other material you will use.
Greetings
S.
https://madpcb.com/glossary/surface-resistivity/
Yes, I expected that big resistor to be a problem. A couple of Chip Manufacturers that I have contacted confirmed that this is probably going to be a problem. Thank You.
- Badges
- Science topic
- Similar topics
- Correction