Fully-differential Opamp Step Response problems

[wandola]

Dear all, pls help me with this problem. I encountered this problem since last week. But I still don't know the reason...

I have designed a fully-differential Opamp based on folded-cascode structure in a 60nm CMOS process. VDD = 1.2V. I choose common-mode voltage to be 600mV.

The circuit is shown below.


The Opamp AC simulation is also shown below:


The open loop DC gain is about 70dB. GBW = 6MHz. Itotal = 97uA. PhaseMargin = 57 Degree.

Then I put the opamp in a feedback configuration to test the step response.



The resistor R = 200Kohms, C = 150fF.

The signals are marked in red square and waveforms are shown in Fig below.



Looking at the step response simulation results, I am totally confused. What has happened to the Opamp? The inverting and noninverting inputs of the fully-diff opamp should remain at Vcm, which is 600mV. The output should be a step like response showing the settling time...

But my waveforms look weird. Can someone tell me what is wrong with my opamp?

Thanks a lot

 

[pavel_adameyko]

Hi, wandola.

Looking on the picture #3, it seems to me that DC operating point of amplifier is wrong because of 12.16 mV at the output.

As i understand, you control current which feeds input diffpair by your CMFB. So if input common mode voltage is low current source transistor appears in triode region - your CMFB wouldn't work, as in your case (common mode voltage is 206.1mV - isn't it very low ???). So my advice is to increase input voltage from 400 mV and see if at DC output is 600mV. If everything will be right, measure step response.

Regards, pavel.

 

[erikl]

... Can someone tell me what is wrong with my opamp?

The footpoint of your vip source (V8 ) is at 600mV (= Vcm), but the footpoint of your vim source (V10) seems to be at 0V.

 

[wandola]

... Can someone tell me what is wrong with my opamp?

The footpoint of your vip source (V8 ) is at 600mV (= Vcm), but the footpoint of your vim source (V10) seems to be at 0V.

Nope... im afraid u made a mistake. The footpoint of V8 is ground. The 600mV is the voltage level at the Vcm output point. Because this number shown in the picture is so close to the footpoint of V8, you misunderstood it.

Added after 8 minutes:

Hi, wandola.

Looking on the picture #3, it seems to me that DC operating point of amplifier is wrong because of 12.16 mV at the output.

As i understand, you control current which feeds input diffpair by your CMFB. So if input common mode voltage is low current source transistor appears in triode region - your CMFB wouldn't work, as in your case (common mode voltage is 206.1mV - isn't it very low ???). So my advice is to increase input voltage from 400 mV and see if at DC output is 600mV. If everything will be right, measure step response.

Regards, pavel.

Hi Pavel, thanks a lot for your help.

Pic#3 is the testbench for step response. I did it in cadence. I also saved the DC operation points. So the values you can seen in Pic #3 are the initial DC values of circuit.

I have two points to discuss with you. My understanding for fully-diff Opamp is that, my inverting and noninverting inputs of the Opamp should remain at common-mode level during large signal, which is 600mV. Am i right?

The output should not always at common-mode voltage in large signal, which is also 600mV in my design. For output voltage, Vout+ and Vout-, we want the DC voltage of these two terminals to be common-mode voltage. That's why we have the CMFB ckt.

So regarding your argument " to increase input voltage from 400 mV and see if at DC output is 600mV", I don't think this point is correct. Pls justify me if im wrong.

btw. still open for discussion.

 

[pavel_adameyko]

Hi, wandola

At first, I want to know if I was right about operation of CMFB circuit - picture so small that i cann't read labels. So you avaraging output, compare it with reference voltage and influence gate voltage of input diffpair current source transistor - is this right ?

If I was right above and CMFB operates in the way i said, then i illustrate more what i said in previous post. I meant that CMFB would work well if input common-mode voltage would be high enough to current source transistor works in saturation. And i think that input common mode voltage in your circuit (400 mV) is low. So my advice was to increase input common mode voltage level.

I agree with you that output shouldn't always be equal to common-mode voltage, but average of Vout+ and Vout- should be equal to common mode voltage level (600 mV).

Regards.