The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> RF Design >> How do you choose Vgs for a gm?
https://designers-guide.org/forum/YaBB.pl?num=1085050442

Message started by naren on May 20^th, 2004, 3:54am

Title: How do you choose Vgs for a gm?
Post by naren on May 20^th, 2004, 3:54am

Hi members,

I am using a 0.18um process and I simply don’t understand how to bias a MOSFET to obtain a specific gm. I know the short channel effect and all!!!!!!

In the good old long channel case we could just use the nice square law equation to calculate the Vgs when we know gm!!!!!

How do you do such a thing for short channel devices?

Not one person seems to address this issue!!!!!

And no prof mails back too!!!!!

....for otherwise there is no more design....its just playing with your simulator until u get the gm u want...then what will u write in a paper.???

The bsim3 is useless for design!!!!

Does someone have any tricks...i mean does some know of any hand calculation equations!!!!!!

For Id=f(Vgs,Vds)????

Can someone answer such a simple question instead of talking about designing a LNA or a super dooper mixer!!!!

Thanks

???

Title: Re: How do you choose Vgs for a gm?
Post by Andrew Beckett on May 20^th, 2004, 9:18pm

The trouble is that at small geometries, it is no longer simple!

That's why complex models like bsim3v3 and bsim4 have come about, in order to describe the behaviour properly. It's not really the model's fault that it is hard to hand calculate at small geometry; it's the physics.

Really those text book equations have only been useful for understanding relative relationships between parameters (for example, if you double this current, that voltage will change by this factor). Trying to use them to solve your circuit by hand has long been difficult (even if you were using a 3um process, say).

Probably not the answer you were looking for, but thought I should say it.

Regards,

Andrew.

Title: Re: How do you choose Vgs for a gm?
Post by City on May 24^th, 2004, 1:02am

Hi,

you can obtain a good approximation of short channel gm when using the square-law together with velocity saturation and mobility reduction effects which reduces the gm by bias dependant factors. You can obtain the necessary value parsing your bsim3v3 parameters. (vsat, K1, ...). You can plot curves of the coefficient vs vds for velocity saturation and and vgs for vertical mobility effect. Together with your square-law gm, that will give you a good guess of the your effective transconductance. Do not forget spacing in for the effective length ... that is 0.18um min length is not the effective length of your channel, which in reality is lower.

City.

Title: Re: How do you choose Vgs for a gm?
Post by Rae YE on Aug 9^th, 2004, 12:43am

You can get an approximate by Square law. Then try to chance some parameter then you can find rules for the change of gm. I did it when I am setting the Vgs for a Gm.

Title: Re: How do you choose Vgs for a gm?
Post by Paul on Aug 25^th, 2004, 12:40pm

Hi naren,

If you design with minimum channel length, you can also try to simulate Id=f(Vg) if you have a good estimate of Vd (for example if you have a cascode topology), for a given W and minimum length. Then plot gm/W=f(Vg) by derivating and dividing by W.

In the design phase, from this graph you can trade off Vg versus W for the desired gm. For typical channel width, you can neglect narrow channel effects and with this assumption, gm should remain proportional to W at constant Vg (and of course for minimum channel length).

Paul

Title: Re: How do you choose Vgs for a gm?
Post by naren on Aug 25^th, 2004, 9:57pm

thanks Paul... :)
thats how I have been doing things these days.
Seems like the more advanced tech we use the more we turn to using techniques similar to using data sheets of transsistors and designing bias points or something.

thanks again

naren

Title: Re: How do you choose Vgs for a gm?
Post by Paul on Aug 26^th, 2004, 5:42am

Hi Naren,

I believe technology guys do their best to provide good devices and we have to work with what they can give us. Unfortunately, we have to trust more and more the simulators, although I think that good understanding of long-channel operation is still required to do a good design.

It is interesting to notice (that's at least my perception) that the device models are most accurate in short-channel region and become less accurate in long-channel operation. Has anybody else observed something similar?

Paul

Title: Re: How do you choose Vgs for a gm?
Post by Frank Wiedmann on Aug 27^th, 2004, 1:27am

When MOS models are fitted to measured data, most importance is usually attached to a good fit at minimum gate length, because this is the value that is most often used and where performance is most critical.

Title: Re: How do you choose Vgs for a gm?
Post by Paul on Aug 29^th, 2004, 12:09pm

Frank,

I only partially agree with this point of view. If you use a 0.18u or 0.13u technology to do RFCMOS design, you are completely right. But if you design a mixed-signal block using that technology just because it goes with a big digital core, then you may be wrong. In the second case, you may well use longer device sizes and accuracy of the models may still matter a lot for you...

Well, that's my point of view and (fortunately) I am currently in the first situation.

Paul

Title: Re: How do you choose Vgs for a gm?
Post by Frank Wiedmann on Aug 30^th, 2004, 12:03am

There are of course situations where one would like to have very accurate models also for longer gate lengths. However, since they are relatively rare, fewer effort is usually spent to accomodate those needs. In many cases, designers of such "exotic" circuits will have to live with these limitations and find some way to work around them.

Title: Re: How do you choose Vgs for a gm?
Post by mannysingh on Dec 17^th, 2004, 3:30pm

Naren,

looks like you want gm to be a function of overdrive (Vod = Vgs - Vth) so use gm = (2*Id)/Vod. Thanks.