Two gm/Id questions - need explanation

[hyc]

Two gm/Id questions

1. The literature says the gm/id method works for all regions, does it mean it will also work for transistors in linear region or is it only work for transistors in satuations?
2. Following the method, if I know Id and I pick L and gm/Id, I should be able to come up with the transistor width W from the chart, will this set of (W/L) guarantee the transistor to always operate in saturation region, if not, how do I get the transistor back into the saturation region using this method?

Thanks in advance.

 

[EmbdASIC]

I also have question similar to 1.
Essentially, some of the the papers on gm/id methodology use the strong inversion and weak inversion regions ? e.g the output stage in a different region, why should that be ?
Any explanation would be appreciated!

 

[Tre]

Gm/Id is a design methodology that does not apply to only one mode of operation, it works for transistors biased in stong inversion, weak inversion, triode, etc..

If you choose L and Gm/Id, then from the characterization charts you will get Id/W. Obviously once you set your Gm, you can calculate Id and W. In my experience, if you do a good job characterizing your technology (generating the charts/plots), very little tweaking is necessary after the transistors have been sized. I design complete OTA's in Matlab/Matchad and do VERY little tweaking on them and they are exactly at the specs where I designed them.

If you can access them, the EE214 lectures from Stanford (now EE315a) are very good starting points to get familiarized with the Gm/Id methodology.

 

[EmbdASIC]

ok thanks for sharing your experience.
I have done my necessary homework now.. but do you think i need anymore than these two graphs?
1. Gm/iD vs Id/(W/L) ... > to get W/L
2. Gm/Id vs. Va (early voltage) ..> to get length

I didnt really start of with matlab, but trusted spectre instead !

 

[hhashemi]

I think you also need Gm/Id vs Vov curve to get Vov in order to sweeing adjacement,or vice versa,adjacement of Vov to determine Gm/Id and then w/l(from id/(w/l) vs Gm/Id curve).

And i also question..
can anyone attach a simple hspice code for extracting id/(w/l) vs Gm/Id curve?
thanks

 

[erikl]

can anyone attach a simple hspice code for extracting id/(w/l) vs Gm/Id curve?
thanks

See here!