Designer’s Guide Community

dxt78

Junior Member

Offline

Posts: 15
Ohio

trouble extracting unCox
Apr 27^th, 2011, 10:22am

Hi all,

I'm trying to characterize a new process, and having trouble extracting unCox. My approach has been to plot ID vs. sqrt(VGS), then find the slope of the curve in the region where I will operate the transistor. Since ID=(1/2)*unCox*(VGS-VT)^2, the slope of the ID vs. sqrt(VGS) curve will be equal to sqrt(1/2*unCox*W/L). The unCox obtained by the above method is approximately 700uA/V^2, which seems reasonable.

My problem arises when I try to use .op in Cadence to confirm the above extracted value for unCox. I print betaeff, which I understand betaeff=(kprime)*W/L. Now, two strange things happen:

1) Cadence is giving the same value of betaeff for transistor of different sizes
2) Assuming kprime = 1/2*unCox, the unCox printed by Cadence is orders of magnitude from the value extracted.

Thank you in advance for your input and comments!

Back to top

IP Logged

Braski

New Member

Offline

Posts: 2

Re: trouble extracting unCox
Reply #1 - Apr 28^th, 2011, 3:04am

are you sure than betaeff is k'*W/L? for example, in MOS MODEL 9 (Philips) the parameter BETSQ is the unCox*W/L for a reference square transistor, so it is equal to unCox, where un is considered without vertical or horizontal field degeneration. Maybe in your models is something like that, and this is why you have the same value for different device size. However, orders of magnitude are a strange thing. Maybe betaeff is not what you intended. What kind of models do you use?

Back to top

IP Logged

Geoffrey_Coram Senior Fellow Offline Posts: 1999 Massachusetts, USA	Re: trouble extracting unCox Reply #2 - Apr 28^th, 2011, 5:31am What model are you using? Is it a MOS or JFET model (since you are "jfetuser") ?
Back to top	If at first you do succeed, STOP, raise your standards, and stop wasting your time. IP Logged

dxt78

Junior Member

Offline

Posts: 15
Ohio

Re: trouble extracting unCox
Reply #3 - Apr 28^th, 2011, 10:25am

Thanks for the replies.

Although I am a (former) "jfetuser", I am now using all MOS!

You bring up a good point about the exact definition of "betaeff". Some sources I looked at online define this as kprime*W/L, but in my case I am certain that it is equal to betaeff=k=unCox, because the parameter does not change with aspect ratio (W/L).

I am not certain as to exactly which model I am using, because the process and associated models come with very limited documentation. It's certainly possible that betaeff is simply a fitting parameter that is nowhere close to what I am assuming it to be.

Back to top

IP Logged