Re: solvers in Fipy

[Daniel Wheeler]

On Oct 13, 2006, at 10:55 AM, hzhatlboro wrote: Hi, thanks for this! I've updated but it seems that the program can not pass line 267-268 for phase 1 (phase 0 was ok).  Is this because of the convergence problem or possibly some other reasons? It is convergence problems. After the first sweep, the value of the phase field is large and negative, which is clearly wrong. This is because S0 + S1 + S2 + S3 is large and negative. You either have to make those sources implicit or reduce the time step by many orders of magnitude. I would also recommend starting from a much simpler system. Thanks and regards, hzhatlboro ----- Original Message ---- From: Daniel Wheeler <[EMAIL PROTECTED]> To: Multiple recipients of list <fipy@nist.gov> Sent: Friday, 13 October, 2006 2:59:30 PM Subject: Re: solvers in Fipy On Oct 12, 2006, at 5:45 PM, Daniel Wheeler wrote: You have "2 / p" in a number of your coefficients. Replace that with "2 / (p==0 * 1e-20 + p)". That will stop the divide by 0 error. Apologies. The above should be "2 / ((p == 0) * 1e-20 + p)". -- Daniel Wheeler Send instant messages to your online friends http://uk.messenger.yahoo.com <pf2dtest.py> -- Daniel Wheeler