Designer’s Guide Community

Pages: 1

Which is faster? (Read 3329 times)

sgalal New Member Offline Posts: 2	Which is faster? Jan 09^th, 2007, 4:01pm Hi All, A MOS transistor with m=100 is split into 100 instances each having m=1. Which structure is faster in terms of simulation time in spectre? Since they are all in parallel then there are no extra nodes added so I would expect the matrix size to be the same. However the number of current signals is now 100x more! Can anyone help? Thanks
Back to top	IP Logged

Ken Kundert

Global Moderator

Offline

Posts: 2384
Silicon Valley

Re: Which is faster?
Reply #1 - Jan 9^th, 2007, 4:18pm

Having one device with m=100 will of course be faster than 100 devices with m=1. It is true that there will be the same number of external nodes in both cases, but in the case with 100 devices the device code will be evaluated 99 more times. In addition, if there are parasitic resistors or current probes associated with the device, the result would be many more nodes in the circuit.

-Ken

IP Logged

sgalal New Member Offline Posts: 2	Re: Which is faster? Reply #2 - Jan 9^th, 2007, 4:40pm Thanks Ken, Is there any option in Spectre that merges the 100 identical devices in one or at least evaluates the device code only once. Thanks
Back to top	IP Logged

Ken Kundert Global Moderator Offline Posts: 2384 Silicon Valley	Re: Which is faster? Reply #3 - Jan 9^th, 2007, 11:35pm No.
Back to top	IP Logged

Pages: 1

‹ Previous topic | Next topic ›