how to realize a current reference with constant gm*R?

[carpa]

Because of the variance of the technology, gm and R are both variable in reality. If constant gm*R are necessary, where gm is the transconductance of a NMOS transistor and R is its load resistance, how to realize such a bias circuit which can keep gm*R relative constant. My purpose is to use mulitple times of R as one part of the bias circuit, and then gernate the bias current (i.e, gm), to reduce the technology variance of gm*R.

 

[chinito]

You can bias your circuit with constant-gm bias circuitry. Since your circuit's bias current is mirrored from the constant-gm circuit, you can be assured that circuits gm will be relatively constant.

 

[xwcwc1234]

The current reference will be suffered the accuracy .

 

[micel99]

maybe David Jones' book has some relevant discussion.