Hi all,

I recently read a paper where the baseband amp used unrestricted biasing to get high IIP3: i.e. no tail current source. The improvement in IIP3 is tremendous. The only problems i could see with the design were:

1) the circuit had no immunity to process variation

2) due to variation its possible the bias point would change and therefore the biasing of the subsequent stage would be affected.

These problems can be worked around though. So is there a really good reason why we would want to add a tail current given the huge loss in IIP3?

thanks,
Aaron

What was the circuit topology (citation)?
A pair of common source amps isn't that much better for IP3 than a differential pair; however, there are transconductor circuits which are extremely linear over a wide swing range but the noise performance isn't very good.

As you noted, IP2 is an issue and common mode rejection is too if you are worried about substrate coupling.
Thanks

Isn't it the bipolar differential pair will normally have better linearity compared to common emitter configuration? I will think that the same applies to FET circuit also.

chase

Dear Aaron,

here are some other reasons why you don't want to omit the tail current source:

1) PSRR becomes (much) worse
2) CMRR becomes (much) worse

If you want to linearize a CMOS differential pair, you could also use resistors (in the source region).

Regards

Peter

Dear Aaron,

placing a buffer with a tail current source after your amplifier (without current source) will not improve your PSRR/CMRR performance, since once the 'disturbance' is added to your nominal/ideal signal, it will proceed through the system (through your buffer etc.) so you can not get rid of it this way.

I have seen this kind of things before (e.g.omitting current sources in CML logic) and it's true, there are some benefits, but PSRR/CMRR are very important when you integrate your macro in a SoC.

Succes!

Regards

Peter