Hallo,

I have a question please regarding the decoupling and bypassing of supply noise because of the bonding inductances.
The question is how to judge the quality of the compensation network.  Shall the xf transfer functions from all the inputs to the tested node be all below the 0dB which I tried a lot but still some are above in certain frequency ranges?
In the transient domain, If the signal swings once before it settles (one over and one undershoot) with a certain amplitude and settling time. I found that this single swing limits the sensitivity of the circuit, this means that I have to make sure that the original signal is large enough. Is it Ok? Is it normally that the supply bouncing limits the sensitivity of  a circuit or is it possibly to completely get rid of it?

Best regards,

aamar

Hi Sumit,

Thanks a lot for your reply, yes I think you got parts of what i meant. I meant actually with the bounce the supply noise resulting from the injected currents in the supply terminals inductances.
My case is as follows, I am designing an optical receiver, the system includes comparators which are implementing class AB outputs leading to current spikes through the supply rails. As soon as I use the inductances on the supply inputs starts the signals to swing. What I learned is to divide the problem as follows:(1) to search for source of the spikes which is the comparators in my case. (2) To search for the blocks which can amplify the supply noise, in my case was the TIA, and to try to isolate these blocks using star connected RC pairs. (3) To connect bypass capacitors which can bypass a part of the current injected by the class AB output stages. I have done all of this on both VDD and GND and checked the XF analysis to check that all blocks have a gain of less than zero (but still I can't). I have even tried to improve the PSRR of the TIA and the voltage post amplifier.

What I have heard once is that as long as the signal settles in no more than one cycle it is ok. But I can find now that the swing of the oscillating cycle is comparable to the signal amplitude for low inputs which deteriorates the BER. So I was not sure If there would be a solution to this noise or it is normal and we have to adapt to it.


Thanks a once again

Best regards,

aamar