The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Analog Design >> Different simulation result for gm and d(Id)/d(vg)
https://designers-guide.org/forum/YaBB.pl?num=1193191358

Message started by Julian18 on Oct 23^rd, 2007, 7:02pm

Title: Different simulation result for gm and d(Id)/d(vg)
Post by Julian18 on Oct 23^rd, 2007, 7:02pm

use cadence to simulate a single CS MOS, plot gm vs. vg and d(Id)/d(vg) vs. vg, get different results (similar but different plots) , why?

Thanks

Title: Re: Different simulation result for gm and d(Id)/d
Post by Julian18 on Oct 23^rd, 2007, 8:20pm

here is my simulation result

Title: Re: Different simulation result for gm and d(Id)/d
Post by rajeee1000 on Oct 23^rd, 2007, 9:12pm

The difference is probably due to the error in the computation of the derivative. Finite difference tends to the actual derivative as the sweep step approaches zero. May be, you will see the curves closer and closer as you progressively decrease the VGS sweep step.

Rajesh

Title: Re: Different simulation result for gm and d(Id)/d
Post by Julian18 on Oct 23^rd, 2007, 11:20pm

rajeee1000 wrote on Oct 23^rd, 2007, 9:12pm:

Thanks Rajesh
I use very small increment to simulate this but I still get the result as before. difference still be there. below is the simulation result

Title: Re: Different simulation result for gm and d(Id)/d
Post by buddypoor on Oct 24^th, 2007, 12:26am

Hello Julian,

since I am not familiar with Cadence: where is the source of "gm" ? Does Cadence have an output parameter called "gm" ?
In Addition, gm is the "mutual conductance" and comprises more than purely the slope of the Id-vg-curve (e.g. the influence of the backward conductance). May be this is the difference.
Regards
Lutz (Germany)

Title: Re: Different simulation result for gm and d(Id)/d
Post by Julian18 on Oct 24^th, 2007, 2:50am

buddypoor wrote on Oct 24^th, 2007, 12:26am:

Hi Lutz:
In cadence you can get gm of a MOS using save statement, spectre will give a variable to you called /M0/gm, and you can use it to plot. As to the mutual conductance, Is the definition of gm d(Id)/d(Vg)?? what else does it have to include to become mutual conductance?

Thanks

Title: Re: Different simulation result for gm and d(Id)/d
Post by buddypoor on Oct 24^th, 2007, 4:25am

Hi, Julian !
In fact, there are three small-signal parameters for the transconductance g:
1.) The mutual transconductance measures
the amount of drain current increase caused by the increment in the gate bias:
2.) The drain transconductance measures the amount of drain current increase caused
by the increment in the drain bias.
3.) Finally, the bulk transconductance consideres the effect of back-gate bias on the drain current

Perhaps, the g values corresponding to 2) resp. 3) are not included in the g-calculation of the simulator.
In this context it is, of course, important whether you are simulating the pure MOSFET or the MOSFET with a load resistance.
Moreover, there may be a difference between an "intrinsic" transconductance and the transconductance measured between the terminals of the DUT (influence of internal resistances).
Finally, here is a good reference fpr MOS-modelling:
iwailab.ep.titech.ac.jp/pdf/mnakagawa/mnakagawa_mthesis.pdf

Regards
Lutz[size=12][/size]

Title: Re: Different simulation result for gm and d(Id)/d
Post by Ken Kundert on Oct 24^th, 2007, 9:41am

Is there any source degeneration? That would be included in dId/dVg but not in gm.

-Ken

Title: Re: Different simulation result for gm and d(Id)/d
Post by genuineintel on Oct 24^th, 2007, 8:38pm

The error might come from the AC part of Gm model, I guess. :-/

Title: Re: Different simulation result for gm and d(Id)/d
Post by toseii on Oct 25^th, 2007, 7:01pm

Hi,

When computing the d(Id)/d(vg) you are also considering the effect of the output impedance of the device, while when computing the gm you get the actual value of the controlled output current source "with the output shorted" (no effect of finite output impedance) in the two port model.

tosei

Title: Re: Different simulation result for gm and d(Id)/d
Post by RobG on Oct 25^th, 2007, 7:59pm

I noticed the same thing a few weeks ago... I've been designing analog for 15 years so I don't think ;) I was doing something dumb. I did a parametric sweep of a current source into a diode connected device. The slope dI/dVgs did not match gm. It might have been the Rs and Rd terms - I never thought to pursue that. There might be affects from Vds if that is changing too. It will be a while before I can get back to it...

Be aware some companies, including one that begins with N, have a sub-circuit for their mosfets that includes more than just a mosfet. This could make gm different than dI/dVgs

rg

Title: Re: Different simulation result for gm and d(Id)/d
Post by nxing on Nov 13^th, 2007, 5:26pm

I think the difference is coming from the "big" gm and "small" gm. means for big gm, you get the gm at every DC point, for small gm, you get actually the derivative at that point. they are not necessary the same.

Title: Re: Different simulation result for gm and d(Id)/d
Post by jimwest on Nov 15^th, 2007, 6:56pm

Hi Julian,
It's the difference between the gm(just the parameter of the mos) and the effective gm (the ability of converting the voltage to the current).
There are guys who will affect the drain current such as vds.
When plotting gm at the OP, the effect of vds will not be involved. But when plotting the dI/dv, everything which will affect the drain current is introduced.

Kinda Regards,

Jim