what's the disadvantage of this CMFB circuit?

[cqmyg5]

Please check the attached figure...

Most op-amps use a pair of identical resistors or switched-cap for CMFB circuit; what's the drawback of the attached alternative form CMFB circuit without resistors or caps. Some book say it has nonlinear characteristic, what's the "nonlinear" meaning?

 

[gsheng]

When Vo+ or Vo- swing too high, differential pair enters large signal region,ie ,one branch cut off, the other flows total tail current.

 

[cqmyg5]

When Vo+ or Vo- swing too high, differential pair enters large signal region,ie ,one branch cut off, the other flows total tail current.

I don't understand "large signal region", if Vo+ too high, Vo- may be too low, the CMFB still work, right? what's the relation between "large signal region" and "nonlinear"?

 

[OpAmp]

This common mode feedback reduces the output swing of the amplifier from lower end. I.e. the CMFB is not effective if the output voltage is less than: Vss+Vgs(Q)8+Vds,sat (Q7)

 

[cqmyg5]

This common mode feedback reduces the output swing of the amplifier from lower end. I.e. the CMFB is not effective if the output voltage is less than: Vss+Vgs(Q)8+Vds,sat (Q7)

But the CMFB with a pair of resistors has same problem, right? the output swing problem is not special for this type CMFB circuit.

 

[rfsystem]

This CMFB circuit could accept differential input signal in magnitude less than

2*sqrt(2)*VDSAT

where VDSAT is the saturation voltage if the diff pair NMOS. Without substrate effect the feedback target is ideal so there is no CMD level change if a pure differential signal is applied. But because of the substrate effect some CMD distortion is introduced. Using NWELL separated PMOS will help here.

The resistor based regulation is able to process unlimited differential voltages and is inherent linear. So for full swing output stages it is more useable.

 

[eezou]

sc cmfb is preferred

 

[cqmyg5]

This CMFB circuit could accept differential input signal in magnitude less than

2*sqrt(2)*VDSAT

where VDSAT is the saturation voltage if the diff pair NMOS. Without substrate effect the feedback target is ideal so there is no CMD level change if a pure differential signal is applied. But because of the substrate effect some CMD distortion is introduced. Using NWELL separated PMOS will help here.

The resistor based regulation is able to process unlimited differential voltages and is inherent linear. So for full swing output stages it is more useable.

Do you mean this type of CMFB circuit using input diff PMOS is linear?

 

[rfsystem]

If you exchange all devices, NMOS to PMOS and vice versus, you can connect the NWELL of the PMOS to the source. Via this the substrate effect is eleminated. But before doing that you should consider simulation of the CMD nonlinearity. Possible does the NMOS performance fits your requirements.

 

[kenlino]

to know the "nonlinear" of this CMFB ckt ,you can run a dc simulation,then you can find there are two step in the wave.

if you design sc ckt you can use sc CMFB,resistors CMFB circuit also can be used, but it will decrease the Gain and also consume extra power.

 

[cqmyg5]

to know the "nonlinear" of this CMFB ckt ,you can run a dc simulation,then you can find there are two step in the wave.

if you design sc ckt you can use sc CMFB,resistors CMFB circuit also can be used, but it will decrease the Gain and also consume extra power.

Thanks! I still have questions:

1. Does the "a dc simulation" mean sweeping one input from 0 to vdd?

I have tried to sweep vi+ from 0 to vdd, and sweep vi- from vdd to 0 at the same time, but the result of .DC simulation shows that the wave of vfb is a flat line, there's no "two steps".

2. How to increase the Gain with resistors CMFB circuit, other than by increasing the value of resistors?