how to design a robust current mirror?

Now I am using the standard biasing which includes a MOSFET transistor and resistor for the current mirror biasing

Although the cascode current mirror can often be used to increase the output impedance and the precision of the current mirror. but what I am considering about now is a robust current supply

But with such kind of current mirror, the current supply biasing T1 and R is very sensitive to the process variations, which I can observe from worse case simulations.

So I am wondering how can I obtain a current mirror which is very robust

i'm also interested in this. previously i've asked the same question as you. you can go . there are some feedback from forumer here.

-evilguy-

Lets separate different terms. Now u have current mirror T1,T2 and R acts as current source. In your case the current mirror (e.g. cascode, or even simple mirror with VdsT1=VdsT2) is robust to process and supply variation to some degree. But ur current source gives current = (Vsuply-VgsT1)/R. Very, very sensitive.
Clearly, that u need more robust current source rather than robust current mirror. There are a lot of methods to make current source: PTAT,CTAT, beta multiplier, Vbe reference, bandgap based, swithed capacitor based, bandgap with external resistor. Stable current source is necessary not for all application. E.g. for switched capacitor ckts or Gm-C filters ur reference current must have some temperature dependence to make slew rate constant or constant transconductance.
I don't know ur design goals. I'll give u some reference for familization. U can look for another information on the forum.

current mirrors are used to duplicate currents, not to generate them.

You need current sources to generate robust currents, then you can use current mirrors to duplicate them.

The picture you uploaded is indeed a combination of current source and a current mirror. The current source formed by R & T1 is too sensitive to process variation.

i've found a paper on this. anybody has the soft copy of this paper. can someone upload it. it might help us. thanks

Low Voltage Current Reference with Temperature and Process Variation Compensation

A. Bendali and Y. Audet (Canada)
Keywords:

-- Current reference, bandgap reference,low voltage, transimpedance amplifier.
Abstract:
This paper proposes a new 1V-voltage CMOS current reference which provides a small dependence of the output current on temperature and process variation. The proposed topology is based on a current-mode bandgap structure that uses a transimpedance amplifier. In order to obtain a small sensitivity of the output current with respect to changes in process parameters and temperature, a threshold voltage (VTH) compensation technique of an output nMOS transistor is used. It produces not only a mutual temperature compensation of mobility and VTH variations, but also reduces the effect of VTH process variation on the output current. SPICE simulations using the TSMC CMOS 0.18m confirms the theoretical analysis. Variations of the output current over a typical temperature range (-40 to 100 C) were simulated to be less than 0.9%, while output changes due to process variation is less than 1%.

https://www.actapress.com/PaperInfo.aspx?PaperID=18455

dear evilguy,

can you pls upload the paper.

thanks in advance.

you caan see the Vt referenced source or bootstrap source. it's very simple do implement it and robust