[SOLVED] efficient way to evaluate circuit performance in spreadsheet if we have Spice-generated technology data

[melkord]

I have generated device performance data (gm, rout, fT, id/W, gm/Id).
Now I want to use Excel to predict the performance of my circuit.
What I am doing right now is selecting gm/Id that I want and inserting gm, rout, etc to the Excel manually.

I get the general idea that it is faster than doing trial and error in SPICE.
But I find it tedious and error prone to manually input the data to the Excel sheet.
How do you usually do this process? Is there any best practice?

I basically following this idea.

 

[sutapanaki]

Matlab is usually a much better and faster option for that. Prof.Murmann uses Matlab himself.

 

[melkord]

Matlab is usually a much better and faster option for that. Prof.Murmann uses Matlab himself.

Hi, could you inform me a little bit more detail about how Matlab and other recommended softwares are supposed to be used in this phase?
Is it like a script that allows us to choose gm/Id as input and the script look for device performance data used in the analytic formulas e.g., Gain, BW, Noise, etc in the csv file?

 

[sutapanaki]

Usually, you do simulations to extract a few things for devices as function of gm/Id. For example Id/W, gm/Cgg, gm/gds, Cds/Cgg, Cdb/Cgg - all as functions of gm/Id. These things are independent (to 1st order) on W, but depend on L. These functions are accessed in Matlab as tables. You can create matlab functions and use them to find for example Id/W for a given gm/Id. Or maybe find gm/Id for given Ft=gm/Cgg, or find Ft for given gm/Id. When you do design you try to work with gm/Id as your primary variable and then find the rest of the design parameters from there.

 

[melkord]

I get it now, at least for circuit with smaller number of MOSFETs. I can plot circuit performance with respect to gm or other device parameters. Thanks.