The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Analog Design >> Criteria for choosing a particular W/L to plot gm/id Vs id/(w/l) graph
https://designers-guide.org/forum/YaBB.pl?num=1331922015

Message started by vinaydreddys on Mar 16^th, 2012, 11:20am

Title: Criteria for choosing a particular W/L to plot gm/id Vs id/(w/l) graph
Post by vinaydreddys on Mar 16^th, 2012, 11:20am

Hi all,

I am trying to understand the sizing approach used in gm/id methodology.

Though I know how to plot gm/id vs id/(w/l) curves for a given transistor, I don't know how to choose a particular W/L for a given process.

ex: TSMC v0.35um W/L = 50um/2um
TSMC 90nm W/L = 12um/0.5um
and so on...

( The above values are just an example).

So... How to decide the initial W/L values to plot the gm/id Vs id/(w/l) graph.

Thanks,
Vinay

Title: Re: Criteria for choosing a particular W/L to plot gm/id Vs id/(w/l) graph
Post by AnalogDE on Mar 16^th, 2012, 2:25pm

Choose L's that you're most likely to use. I would start with minimum, 2x minimum, 4x minimum L and so on. Choice of W doesn't really matter, since it's all normalized (i.e. Id/W). I would just pick something like W=10um, which should be within the W bounds of the model.

Title: Re: Criteria for choosing a particular W/L to plot gm/id Vs id/(w/l) graph
Post by loose-electron on Mar 16^th, 2012, 6:35pm

A different approach -

Start with 2 devices:

1 - min channel length used for switches. (call this device "D")

1- channel length where matching data shows good matching (call this devices "A)

Set the channel W of device "A" to a width that shows good matching.

"A" is now set W/L
"D" is min channel length, and is set to a "W" equal to that of "A"

From here on, your transistor sizing
is totally done by:

NF = #
M = 1, or 2 (for common centroid diff pairs)

Why do this? Now all your transistors are
the same width and makes the layout much easier to do.

Title: Re: Criteria for choosing a particular W/L to plot gm/id Vs id/(w/l) graph
Post by vinaydreddys on Mar 19^th, 2012, 10:34am

Thanks for your valuable inputs.

When i plot with different L s (Lmin, 2x, 4x.. ), I will get different curves, each varying with respect to the other.
So, Which one should I choose, to get W/L s of transistors in the design...?

Say for example, in the design of a 2-stage op-amp, the L for the input transistors is different from that of the L for current mirrors.
(source : MILLER OTA DESIGN USING A DESIGN METHODOLOGY BASED ON THE GM/ID AND EARLY-VOLTAGE CHARACTERISTICS: DESIGN CONSIDERATIONS AND EXPERIMENTAL RESULTS )

So, should I consider two different gm/id Vs id/(w/l) curves... or should I take the average of the curves obtained for different Ls..?

I am confused about the final curve(s), that is taken as reference to proceed with this gm-id approach.

Thanks a lot,
Vinay.

Title: Re: Criteria for choosing a particular W/L to plot gm/id Vs id/(w/l) graph
Post by loose-electron on Mar 19^th, 2012, 10:49am

As a general rule you can use a bias curve set to select voltages
and then adjust the desired bias current using changes to W
(or change the NF or M numbers in the method I have outlined)