toye@CSV.RPI.EDU (Edward J. Toy) (04/14/87)
In article <8704090953.a008691@Huey.UDEL.EDU> Foster@UDEL.EDU writes: > > I have been attempting to run some simulations of some dynamic > CMOS circuits using SPICE2G6 and encountered some problems with > the numerical integration time step. > .... > The error I get in the output file is "INTERNAL TIME STEP TO SMALL". > In the documentation that I have I can find no way to control this > parameter. I have tried changing the method of integration, and the > other parameters associated with it and had no luck. > .... > I am using a LEVEL 2 MOS model, with parameters obtained from MOSIS. > There have been quite a few responses to the above message, but I think one more can't hurt. Please note that I am not a spice guru and I have little more then the basic knowledge on how spice works. When I encounter problems such as the above, I usually resort to the ad-hoc method of printing out the spice options and then trying to alter those options until I get spice to run. The solution below has worked for me but remember that I am not sure why. In reguard to the above message posted to INFO-VLSI, I have had the same problem and have managed to overcome it as follows. All that needs to be done is to set the parameters ITL3, ITL4 and ITL5 with the .OPTIONS command as follows: .OPTIONS ITL4=500 ITL3=100 ITL5=0 The defaults are ITL4=50, ITL3=100 and ITL5=5000. Setting ITL5=0 removes some sort of internal test spice performs. If the error persists, increase ITL3 & ITL4 but keep ITL5=0. I don't know any neat formula to select values for ITL3 & ITL4 (the above values seem to work most of the time and I now use them as my defaults in all my spice runs), I just start with these values and increase them if spice bombs out with the above error message. As a rule of thumb, I double the values if the spice run fails and then rerun spice. This may seem somewhat silly, but I have not encountered a case where I did not get spice to run in the end. By the way, I am running the same version of spice with the mosis parameters. Ed Toy toye@csv.rpi.edu