about constant gm bias

[jecyhale]

Hi, all

I have designed a constant gm bias, and the gm=1/R
So, my question is how to confirm the value of R is a constant?

We know the R will change with PVT more than 20% usually.

Thanks.

 

[ieropsaltic]

The constant gm bias won't supply constant gm, but constant gm*R and since R is very process variant, you won't get a constant gm across process corners. You would want to use this kind of bias at applications where constant gain is required where gain=gm*R.

If you're in real need of constant gm indep on R, you may need to trim the resistors used at constant-gm circuit, but this is very expensive.

 

[jecyhale]

The constant gm bias won't supply constant gm, but constant gm*R and since R is very process variant, you won't get a constant gm across process corners. You would want to use this kind of bias at applications where constant gain is required where gain=gm*R.

If you're in real need of constant gm indep on R, you may need to trim the resistors used at constant-gm circuit, but this is very expensive.

Thanks.

So I have used three current refference, the iPTAT, VBG/R, and constant Gm.
I think this three current reference are all process variant.
isn't it?
So