robust biasing network

[Braski]

Hi all,

I am trying to design a biasing network for a gain boosted folded cascode (1.8V), able to work in all corner at room T, and able to work in a decent way also at -40 and 200 C. I assume constant reference current (external source).
At this point, i used simple diode-connected MOS with a fixed current flowing through them to build all the voltages. In one case, i have used MOS and resistors.

This is not so robust, and in some extreme corners (e.g. slow, -40) some devices enters triode region. The amplifier still works, but I would like to improve it.

What is the correct approach to build a robust biasing network?
Maybe not use a fixed current but derive a variable current in order to compensate for process deviation?

Please help

 

[erikl]

... Maybe not use a fixed current but derive a variable current in order to compensate for process deviation?

Yes, a simple hint: Instead of a fixed (constant) current, use a PTAT current. This helps at the extreme hot corner (slow, highest temp.) to get enough current to still manage the necessary speed, and reduces the current at the cold corner to (hopefully) stay in saturation.

BTW: (slow, -40) is a mixed corner, (fast, -40) would be the extreme corner.

 

[Braski]

Yes, a simple hint: Instead of a fixed (constant) current, use a PTAT current. This helps at the extreme hot corner (slow, highest temp.) to get enough current to still manage the necessary speed, and reduces the current at the cold corner to (hopefully) stay in saturation.

BTW: (slow, -40) is a mixed corner, (fast, -40) would be the extreme corner.

ok, i will try to evaluate a PTAT biasing and see if it helps. However, being it a test chip, the idea was to bias it from an external source, with a fixed and well known current.

thanks for the clarification. I also have some trouble in this mixed corner, due to how the complete circuit is arranged (not only the amp).

However, obtaining the voltages from simple diode MOS is quite ok?

 

[erikl]

... obtaining the voltages from simple diode MOS is quite ok?

It depends on how you control the current through the MOS diode: if it's a constant current, a MOS diode current mirror will reproduce a constant current; a PTAT current will be reproduced as a PTAT current. That's just the intended purpose of a current mirror.

 

[Braski]

ok but i was just talking about the voltages of the cascode MOSTs. However, i found the proper way to do it to obtain acceptable robustness over T and corners. Thank you for your valuable help.