How to improve the PSR of my fully differential op?

[icfans]

Hello, everyone. I am now designing a Class AB folded cascode fully differential opamp, and the Class AB is implemented by floating voltage source.
The spec require a PSR(Avdd=Vo/Vdd) of single end less than -50dB, but I can only get about -46dB in tt corner.

Here is my question:
1. How to improve PSR? what's the most important parameter that will affect PSR?
2. In some corners (e.g. -40 degree, ss, 0.9vdd), the PSR decrese very much to about -28dB, why?

Need for help! Thank you!

 

[erikl]

1. How to improve PSR? what's the most important parameter that will affect PSR?

PSR doesn't depend (so much) on parameters, but much more on the circuit topology. Most of the PS interference usually is injected to the output through stages without current sources, so probably your single-ended class AB output stage is the culprit. You can easily find out, if you separate the power supply, inject the interference stimulus only to the power supply of a single stage, and provide a clean supply to the other stages, then run another AC (or XF) analysis for PSR, and compare.

2. In some corners (e.g. -40 degree, ss, 0.9vdd), the PSR decrese very much to about -28dB, why?

PSR decreases (to |absolutely| lower negative dB-values) with decreasing loopGain = openLoopGain/closedLoopGain. OpenLoopGain decreases usually with ss conditions and with lower power supply (due to lower headRoom) - but not necessarily with lower temperatures - and so the loopGain and with it the PSR decreases.

So how to get higher PSR:

• increase openLoopGain
• spend a current source to the most PSR-responsible stage (which of course will cost you some output voltage range, if it is the last stage).

 

[icfans]

PSR decreases (to |absolutely| lower negative dB-values) with decreasing loopGain = openLoopGain/closedLoopGain

thank you, erikl. But I don't understand how PSR is realted to loopGain, can you explain that?

I wondered whether PSR is related to Rout and prasitic capacitor from vdd to Vout?

Thank you!

 

[erikl]

I don't understand how PSR is related to loopGain, can you explain that?

Like any signal - desired (the normal input signal) or unwanted - as noise, general EMI, injected on power supply or whereever, and circuit properties like non-linearity are attenuated by negative feedBack, the responsible factor being the loopGain; read the books!

I wondered whether PSR is related to Rout

Lower Rout increases the PSR stemming from the last stage (voltage division ratio).

and parasitic capacitor from vdd to Vout?

Larger parasitic capacitor from vdd to Vout decreases the PSR (voltage division ratio, again).

 

[icfans]

Like any signal - desired (the normal input signal) or unwanted - as noise, general EMI, injected on power supply or whereever, and circuit properties like non-linearity are attenuated by negative feedBack, the responsible factor being the loopGain; read the books!

Got it. But now what I am really talking about is PSR=Avdd, NOT PSRR.
So consider my fully differential op-amp, the single-end PSR is not related to the Open Loop Gain. Am I right?

 

[erikl]

PSR=Avdd

Not understandable for me, sorry!

 

[icfans]

erikil, thanks all the same

 

[jgruszynski]

Remember what PSRR is. It's always useful to go back to basics!

Power is common mode while your signal is differential mode. PSRR is approximately Adiff / Acm, so make circuit changes that alter this ratio in the right direction.

For example, diff pair current mirrors assure that common mode gain is heavily degenerated (thus low) because current sources have high impedance while diff gain is high and have no degeneration feedback at all. So whatever you can do to improve the ideality of current sources at high frequency will generally improve the PSRR, for example. Failing to look at the input impedance of current sources over frequency is a common newbie mistake that can hose PSRR.