Rail to Rail Op Amp Input stage problem

Dear everyone

I recently have read a biosensor paper. The author adopts a rail to rail folded cascoding structure to design a fully differential amplifier. The schematic is attached below.

My questions are
1. What is the function of the circuit as I circled in red? As far as I know, we need to provide a constant gm at the rail to rail input stage but I really have no idea why this circuit can make it.
2. Why the sizes are different in the bias circuit as I circled? I assumed that the cascoding branch should provide 2uA current so the size should be the same.
3. Why the input stage is like a differential amplifier?

Thanks in advance for all the discussions and help!

phdseeker said:

1. What is the function of the circuit as I circled in red?

Click to expand...

Mirror current generation.

phdseeker said:

As far as I know, we need to provide a constant gm at the rail to rail input stage but I really have no idea why this circuit can make it.

Click to expand...

R2R inputs usually -- i.e. without additional circuitry -- don't provide constant gm. Within the middle range of the power supply, gm nearly doubles compared to the ranges close to GND & VDD, because both complementary input stages contribute to the transconductance.

phdseeker said:

2. Why the sizes are different in the bias circuit as I circled? I assumed that the cascoding branch should provide 2uA current so the size should be the same.

Click to expand...

In the broken circle? This is part of an often called "Monticelli" quiescent current regulation method for the output stage. I guess the size ratio (15) came off by simulation / try & error method. And probably has to match with the size ratios (30 resp. 5) of the cascode transistors.

phdseeker said:

3. Why the input stage is like a differential amplifier?

Click to expand...

Isn't that quite normal for an opAmp?

phdseeker said:

1. What is the function of the circuit as I circled in red? As far as I know, we need to provide a constant gm at the rail to rail input stage but I really have no idea why this circuit can make it.

Click to expand...

It provides some form of gm control, though not necessarily constant. Imagine if the input common mode is way above Vcm, the bottom left transistor essentially steals most of the 16uA of tail current from the lower branch.
In contrast, the top left transistor is in cut-off, giving the full 16uA of tail current to the upper branch.

checkmate said:

It provides some form of gm control, though not necessarily constant. Imagine if the input common mode is way above Vcm, the bottom left transistor essentially steals most of the 16uA of tail current from the lower branch.
In contrast, the top left transistor is in cut-off, giving the full 16uA of tail current to the upper branch.

Click to expand...

Thanks, Checkmate
Your explain turns me up. The source follower is used to set the constant gm into the circuit.
Is the common mode feedback in this circuit efficient?What is the function of the floating current source circled in slash lines?

---------- Post added at 17:34 ---------- Previous post was at 17:27 ----------

Hi, erikl

Thanks for the valuable reply.
You may have found that this circuit is the folded cascoding structure. Normally the input should be just NMOS or PMOS with tail.
Here, it uses the diff pair at the input side. It seems it doesn't need current to be provided by the cascoding part.