The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Analog Design >> how to extract flicker noise co-efficient
https://designers-guide.org/forum/YaBB.pl?num=1157965337

Message started by avlsi on Sep 11^th, 2006, 2:02am

Title: how to extract flicker noise co-efficient
Post by avlsi on Sep 11^th, 2006, 2:02am

HI friends,

I would like to extract flicker noise co-efficients for NMOS and PMOS in 0.18 u m process.
Can u tell me,what is the procedure to do it in cadence(any automatic)
or else any test circuit.

Thanks in advance.

Title: Re: how to extract flicker noise co-efficient
Post by ACWWong on Sep 11^th, 2006, 2:09am

avlsi wrote on Sep 11^th, 2006, 2:02am:

I'm not quite sure what you mean by "extract".
The flicker noise coefficients that Cadence(spectre?) will simulate are given in the foundry model files. Parameters like kf, af, noimod etc. You can look up the meaning and how they are used in the spectre model guide in your cadence installation path.

Title: Re: how to extract flicker noise co-efficient
Post by avlsi on Sep 11^th, 2006, 2:35am

Hi,

I want to know the Kf of NMOS and PMOS. How will calculate using some circuit,as we do for Vt,Kp,Lambda.

We plot some curves and calculate for hand calculations. Similarly,is there any procedure to get the flicker noise co-efficient.

Thanks

Title: Re: how to extract flicker noise co-efficient
Post by loose-electron on Sep 11^th, 2006, 10:10pm

If I understand the question, correctly -

You need to go into the lab and measure the transistor flicker noise in a defined test envronment, for multiple sizes of W and L and then fit the noise parameters measured to create the actual model parameters.

Is that what you are looking for?

Jerry

Title: Re: how to extract flicker noise co-efficient
Post by avlsi on Sep 11^th, 2006, 11:05pm

In fact, My question seems a bit misleading.

I meant, I want a circuit to simulate the flicker noise and then calculate from the graphs, Kf(flicker noise co-efficient)

We can get the flicker noise co-efficient from Model file itself.

Thanks for your comment.

Title: Re: how to extract flicker noise co-efficient
Post by Andrew Beckett on Sep 12^th, 2006, 1:08pm

Why would you want to do this? Extracting a model parameter from simulated graphs, when you have the models, seems a bit of an odd thing to do...

Andrew.

Title: Re: how to extract flicker noise co-efficient
Post by sck236 on Sep 13^th, 2006, 4:57pm

For hand calculation, it may be useful if I can extract Kf from the simulated grapgh.

I heard that simulators use rather complex parameters of BISM for calculating noise.

This parameters are not useful in hand calculation.

Does anyone has answers? This question is also what I want ask.

Title: Re: how to extract flicker noise co-efficient
Post by ACWWong on Sep 14^th, 2006, 2:50pm

sck236 wrote on Sep 13^th, 2006, 4:57pm:

It depends on which exact model of BSIM you are using and what setting your model gives (eg. flkmod or noimod etc) as to what euations it uses.
So for example BSIM4 (quite modern) model with flkMod=1 gives:
Flicker noise = (Kf*gm^2)/(Cox*W*L*Frq^Af), which is pretty straight forward especially as the model file will tell you what Af and Kf are.
But if you insist on simulating it, just do a noise simulation (voltage driving the gate of a common-source fet will do for example) and plot the the noise, and correlate (or "extract" as you say) Af and Kf, you'll end up with the same or close to those in the model file..... so i'm still not sure what the point would be ?

Title: Re: how to extract flicker noise co-efficient
Post by avlsi on Sep 20^th, 2006, 5:02am

Hi guys,

I would like to enlighten you about the flicker noise co-efficient.

There are two formulas for flicker noise
1. simple using Kf and Af
2. using q,noia,noib,noic,ef .. etc

Generally for noimd= 1, simple formula is used. This is not accurate.
So for noimod flag = 2,3 , we use the complicated expression.

This u can refer in BSIM3 manual.

Can anyone tell me a circuit, to simulate this noise and extract the Kf. Because not all foundries give Kf and Af. they give noia,noib,noic,Em,Ef.

Thanks in advance.

Title: Re: how to extract flicker noise co-efficient
Post by ACWWong on Sep 20^th, 2006, 5:33am

avlsi wrote on Sep 20^th, 2006, 5:02am:

There is no need to enlighten me. I know, and in fact mentioned it in several of my previous posts to look up the kf and Af noia noib noibc Em Ef etc in the model file and look up the equations in the model guide.
IF YOU LOOK AT YOUR MODEL FILE AND THE VALUES FOR KF, AF, MOIA, NOIBM, EM, EF ARE NOT SET THEN IT IS POINTLESS TO SIMULATE BECAUSE YOU NOT GET ANYTHING!! HOWEVER THEY ARE SET IS WHAT ANY SIMULATION WILL GIVE YOU!

If noimod is set to 2 or 3 it will still use the parameters given in the model file. Again if they are not set, then you will get nothing.

Also as i said in my previous post you can simulate a common source fet driven by a voltage source on the gate and do a noise analysis. You will get the noise plot which, HOWEVER THE MODELS HAVE THE FLICKER NOISE PARAMETERS SETUP, will give you a noise you want. You can then correlate the slope and intercpet of the graph using simple maths to correlate back to the PARAMETERS IN THE MODEL FILE.

rgds,
aw

Title: Re: how to extract flicker noise co-efficient
Post by avlsi on Sep 20^th, 2006, 5:43am

I didnot litterally mean "enlightning"

I take back that word.

Ok, u have told u can get Kf and Af from model file. But I am not having them in my model file. So wht can i do is simulate and then extract.

Thanks for ur concern.

Title: Re: how to extract flicker noise co-efficient
Post by ACWWong on Sep 20^th, 2006, 6:12am

avlsi wrote on Sep 20^th, 2006, 5:43am:

so you don;t have af or kf, so the model will be using noimod 2 or 3 in which case noia noib etc must set for any simulation to valid. you can do a noise simulation of a common source fet and plot the output noise. the slope will give you an equivalent af, the intecept point will help you equate an equivalent kf. of course the noise you get is bias/area dependant. perhaps do a family of sims to see if using noia noib etc. give youa stable equivalent kf, af of a simple model.