How does ping-pong amplifier cancel offset?

[Pipeline]

By paper "A rail-to-rail ping-pong op-amp"(JSSC sep 1996), I do not understand how the ping-pong offset cancellation works.
Since the Vos will be stored and feedback to an input of another input pair, but this pair will work open loop, so the stored Vos will be amplified and get A*Vos at the output. how can the offset be cancelled?
Thanks!

 

[Humungus]

In ping pong operation you have two identical amplifiers each one mounted in a configuration allowing auto-zero of its input offset. Then, circuit operates cyclicly by zeroing one amplifiers during one clock perio, while the other is connected to the main loop, then, during the next clock period, the zeroed amplifier is connected to the main loop, while the other amplifier is zeroed.

As you swith from one amplifier to the other, while the second amplifier is being zeored, this is called ping-pong.

 

[Pipeline]

Thanks, and I know the principle.
Now my question is there are TWO input pairs of ONE identical amplifier, one for signal and the other for offset feedback. but when it works as an amplifier, only the signal input pair is closed-loop, and the other input pair is actually open-loop.
how can we use open-loop to do cancel offset? I do not understand this in the paper.

 

[timer]

hi pipeline,
could you post the paper or the link to the ping-pong paper
thanks,
bye

 

[Pipeline]

here is the paper:
https://ieeexplore.ieee.org/iel1/4/11273/00535417.pdf

 

[willyboy19]

The link you put there can only be accessed by IEEE member. Can you upload here in this forum? I can certainly help you out after reviewing the paper.

 

[timer]

i'll upload the paper for you by sunday.
bye

 

[Humungus]

Pipeline,

The amin input is used as signal input and, as you said, it is likely to be part of a closed loop. The second input is used for offset compensation, so some storing device, say a capacitor, bias this secondary input in order to get the total amplifier's offset close to zero. no problem if this secondary input operates in open loop, provided that the leakage currents through the switches used to zero the amplifiers do not introduce so much drift during the amplification phase. The ping-pong frequency depends of this leakage.

 

[Pipeline]

Thanks Humungus.
I understand in open-loop, the correction input can cancel the offset of signal input. But when we use this amp closed-loop, the signal input pair is not open-loop, so is there any change then? that's where I am confused.

BTW, anyone can explain how to determine the clock frequency and the offset-storing capacitors in details? thanks!

 

[Humungus]

Thanks Humungus.
I understand in open-loop, the correction input can cancel the offset of signal input. But when we use this amp closed-loop, the signal input pair is not open-loop, so is there any change then? that's where I am confused.

Me too. I don't really understand your question.

Each amplifier has two inputs: one for signal and one for offset nulling.

In one phase, one amplifier is nulled, while its differential input is shorted to the input voltage (to have the right common mode), while the other amplifier is connected to the signal loop, with its nulling input connected to capacitors that stored the offset of the precedent nulling phase. Once the amplifier is nulled, the signal input is connected differentially to the signal loop, while its nulling signal is connected to capacitors which stored its offset. I don't know what el se to say. Did you get it this time?