Author |
Message |
minsc New member Username: minsc
Post Number: 1 Registered: 01-2010
| Posted on Friday, February 19, 2010 - 03:15 pm: | |
Dear Dr. Nelson: I am calculating electrical potential of the PN junction using this equation: EQUATIONS U: dxx(U)+ dyy(U)= -(qe/esi)*(nd-na- nd*exp((U-vl)/vth)+ na*exp((vr-U)/vth)) My code works fine in Version 4.2.16, but it always shows error message about exceeding the limit 790 of exponential in Version 5.1.4. Could you tell how to solve this? Thank you very much! |
rgnelson Moderator Username: rgnelson
Post Number: 1331 Registered: 06-2003
| Posted on Saturday, February 20, 2010 - 12:36 am: | |
Without seeing the rest of the script, I can only reiterate what the diagnostic says, that either (vr-U)/vth or (U-vl)/vth has become larger than 790, at which point the computer hardware is no longer able to represent the value of the exponential. You might be able to locate the trouble if you MONITOR all the quantities involved in the equation and SELECT BUSYMON. This will show monitors at each Newton iteration. It is possible that the Newton's method is diverging, in which case you will not even get the first monitor. You could put a MIN() control on the arguments of the exponential. But I expect you have started the calculation with wildly erroneous initial values, in which case Newton's method may not be able to converge. Try initializing your variables to a reasonable guess for the solution.
|
minsc New member Username: minsc
Post Number: 2 Registered: 01-2010
| Posted on Monday, February 22, 2010 - 12:41 am: | |
Dear Dr. Nelson: Here I attach my code. Please take a look. Thank you very much.
|
rgnelson Moderator Username: rgnelson
Post Number: 1332 Registered: 06-2003
| Posted on Monday, February 22, 2010 - 06:57 pm: | |
When I run this problem on Version 4.2.16, it reports the same error (Argument of EXP is too large: 1.3e+10). Look at the things you may have changed since version 4 worked correctly. Also, you have discontinuous initial conditions. U is zero everywhere and -5.8 on the boundary of box2. This does not qualify as a "reasonable" estimate of the solution.
|