CMRR and PSRR of Fully Differential Difference Amplifier

[deveshkm]

Hi,
I had connected the FDDA in unity gain feedback mode and calculated the gain from Vdd to output.
The value obtained is -300 dB!.
Please comment if this is the correct set-up and whether the value can be justified.


For CMRR, I connected the input terminals to the same sinusoid and calculated the CM- DM gain.
This value was added to unity gain( low value in dB) to get the CMRR.


Please comment if this is the correct set-up and whether the value can be justified.

 

[frankrose]

I would check PSRR vs. frequency in AC analysis and run Monte Carlo mismatch. They can screw PSRR, but I don't see problem with your setup. I just don't understand why you call the vcvs output as 'outcm'. It should be a differential output, not a common mode.
The sign of your rejection is not so good, however lots of the engineers calculates PSRR with negative sign in dB. I think it is wrong, because the "Rejection" of supply noise should be high as possible, so I prefer the positive dB values.
From same reason for CMRR I would calculate it as CMRR = Au_diff/Au_cm = Au_diff_dB - Au_cm_dB

 

[KlausST]

Hi,

300dB is a factor of about 1,000,000,000,000,000.
I'd say this is not a measured value, because it is impossible to measure such a high ratio.

Klaus

 

[deveshkm]

I would check PSRR vs. frequency in AC analysis and run Monte Carlo mismatch. They can screw PSRR, but I don't see problem with your setup. I just don't understand why you call the vcvs output as 'outcm'. It should be a differential output, not a common mode.
The sign of your rejection is not so good, however lots of the engineers calculates PSRR with negative sign in dB. I think it is wrong, because the "Rejection" of supply noise should be high as possible, so I prefer the positive dB values.
From same reason for CMRR I would calculate it as CMRR = Au_diff/Au_cm = Au_diff_dB - Au_cm_dB

Thanks for responding
I kept vcvs label as "outcm" to indicate the CM - DM gain
I would recheck the value after Monte Carlo mismatch

 

[frankrose]

Hi,

300dB is a factor of about 1,000,000,000,000,000.
I'd say this is not a measured value, because it is impossible to measure such a high ratio.

Klaus

It looks absulutely as simulation result, and diffrential circuits can produce numbers like this in CAD for PSRR or CMRR. I agree this is not a real circuit's PSRR.

 

[deveshkm]

Hi,
This time I performed an xf analysis in cadence virtuoso.
I used a single setup, with output mentioned in the analysis.
FDDA was connected in unity gain mode
Each time I chose a source, i got the corresponding plot
common mode (DC) voltage for common mode gain (-ve)
differential voltage (signal) for differential gain (unity in this case)
small signal voltage at Vdd for gain from vdd to output (-ve)

cmrr = diff-mode gain(0 dB) - common mode gain = 115 dB
psrr = diff-mode gain (0 dB) - (gain from vdd to output) = 143 dB

Can these values be considered reliable?

 

[frankrose]

143dB very high, simulate with process+mismatch variation. 115dB not small either. How much is the frequency?
But as I mentioned differential circuits can produce numbers like this in CAD for PSRR or CMRR.

 

[deveshkm]

Thanks for the feedback.
Please comment on the method used

 

[frankrose]

Your method is good, just note the frequency and a suggestion to find the worst case values. For example the CMRR can depend on the DC value of vcm input too not just on PVT and mismatch.

 

[deveshkm]

Your method is good, just note the frequency and a suggestion to find the worst case values. For example the CMRR can depend on the DC value of vcm input too not just on PVT and mismatch.

Hi,
These values were taken at 100 Hz