[SOLVED] fully differential amplifier and CMFB

[greenjuice]

I'm designing fully differential amplifier to utilize at bandgap reference circuit.
And I studied CMFB for fixing output common mode level of amplifier.
But I have heard CMFB also drives input voltages of amplifier which is virtual short.
It is true?

Actually I already made fully differential amplifier.
But if i connect amplifier with PTAT, CTAT generator, CMFB make output common mode level constant but i can't make its virtual short.
This is my circuit. Tell me my problem. Thank you in advance.

 

[FvM]

I doubt that a fully differentially amplifier is right in this place. Unfortunately the circuit can't be analyzed without knowing the PTAT/CTAT output characteristic.

To work in the shown circuit, PTAT and CTAT output node must be bidirectional current sources but they most likely aren't.

 

[greenjuice]

This picture is my ptat/ctat generator which use vbe and variation of vbe from different bjt emitter junction width. And of cource my ptat and ctat node have bidirectional current.

As i know, if i design it well, its input voltages should be virtual short.

But it isn't. Why? and Could you teach me how can i make specific virtual short which i want using CMFB.

this CMFB which i use can also regulate input voltage level?

 

[FvM]

O.K., not bidirectional current sources, but complementary N and P current sources. So the fully differential amplifier can work with suitable dimensioning.

Which amplifier input and output node voltages do you observe? Why do you think that the amplifier doesn't implement a virtual short?

 

[greenjuice]

Oh. I'm sorry that where it is not virtual short in advance. It is Fully differential amplifier, not single ended output amplifier that couldn't make virtual short.
I observed PTAT CTAT node which is input voltages of fully differential amplifier. Actually to make virtual short, I think high open loop gain and negative feedback is essential. My folded cascode open loop gain is about 90dB at input range of upper 0.5V. So I think it's ok. But it wasn't virtual short. I think i am missing something to make this.

 

[FvM]

Of course a fully differential amplifier will create a virtual short between both input nodes. As you say, sufficient open loop gain is required. What's the open loop gain when your amplifier is loaded with feedback resistors? It has only a single gm stage.

 

[greenjuice]

This is my open loop phase margin of fully amplifier. And If I connect resistor 100k along feedback loop and check closed loop gain, that gain is about -20dB. clearly, this is cause of not virtual short. So I thought If i connect feedback loop, then it make another poles. so it make phase margin not get. So. I make resistor above 1M. But It is not virtual short. My input voltage(+) is always 1.87v and input voltage(-) is always 1.62v.

is there something wrong with my design?