lukeS
New Member
Offline
Posts: 5
Silicon Valley
|
Hi all,
I am running a dc sim on a bandgap in spectre where I ramp AVDD from 0 to 3.3 in 500uV increments. I have all the simulator options defaulted except pivrel=1e-3, and I am saving selected voltages and currents. I get a warning saying the default gmin of 1pS is too large and may affect my result. Then it prints 'dV(I21.MN16.QNPND:int_e)=-27.814mV' or something like that. What does this line mean, and what is it telling me?
I also get a line printed that says 'use the 'gmin_check' option to eliminate or expand this report' What is the gmin_check option and how does it work? When I look in the ADE L -> Simulator -> Options -> Analog -> Algorithm tab, I see the gmin_check line with 4 check box options: 'no', 'max_v_only', 'max_only', 'all' What do each of these mean/do/how do they work?
Finally, I get a 'Bad pivoting' warning and it suggests I turn on 'dc_pivot_check' to improve convergence. When I put 'pivrel' to default this warning goes away. What is 'bad_pivoting' and 'dc_pivot_check' in the context of dc analysis? I know dc analysis assumes the solution to be constant valued, and solves the nonlinear system by setting derivatives to zero and linearizing/solving for each dc operating point (at each supply voltage, for my specific simulation). So what is 'pivoting' and how does it fit into this context?
When I rerun the sim with the 'all' option checked for gmin_check and 'yes' checked for dc_pivot_check, I still see the gmin warning about 1pS being too large, but that's it. No warnings about dc error 'dV'. My speedup is from 1min31s to 1min37s for the sim (7% speedup) from not checking gmin_check and dc_pivot_check. Thank you for your help! :)
|