problem with gain error in S/H amp .

[manissri]

hi all..
i am desingning the 12 bit 10 msps pipelined adc. for this i am using for S/H amplifier - folded cascode fully differential with gain boost opamp. i am getting 106db gain, P.M.-61 degree, UGB-261mhz.
my input range is 2Vpp. so my lsb is 2/4096 = 488 uv. my supply is 0 - 5v
my question is when i put my opamp in unity gain and apply a sin wave of of 1Vpp with 2.5v offset. theroatically i should get around 1Vpp with 2.5 volt offset.
but i am getting a 5mv error in my output.
and my lsb is 488uv.
so having such gain (106db) why i am getting 5mv error. and how i will rectify this problem so that i get the error withing 488uv range. so that i can desing my 12 bit adc.
waiting for ur reply ..
manish

 

[philipwang]

Desr manissri,

What means when you mentioned "5 mv error"? Would you please put your output chart on here or describe your output more detailed?

Bg,

 

[sankudey]

This one is really a good problem and not comprehensively addressed in literature.
You can go through the following options and try first to locate the problem.

1. The error you reported could be a common fallacy. To measure the error we generally do (Vo-Vi) operation. But due to certain delay, the error seen is not correct. To avoid this, the operation should be done like (Vo-Vi_d) where by "Vi_d" I want to mean the delayed version of input and the amount of delay should be adjusted by observing the error curve/result for several cases.

2. One very important point is the common mode voltage. Some times it happen that the circuit performance changes (e.g., gain, references etc.) over the common mode voltage. You could check it by smaller aplitude signal. Also for higher amplitude check the current and voltages at different nodes , particularly the bias generating part.

3.Have ever checked the 3db BW of the op-amp. I am not sure, but it is said that the BW (which does mean the 3db bw in openloop) should be greater than the input signal frequency (not to confuse with sampling rate).

 

[manissri]

Dear philipwang..
actually my lsb is 488uv ( as 12 bit adc). and my common mode voltage is 2.5v.
so when i apply the 2.7v and 2.3v at Vi+ and Vi- on S/H stage ( which consists of opamp and switch cap circuit for sample and hold mode). so i am getting 2.6946 and 2.3051v which is greater then my lsb( 488uv).
in other simulation for checking the only opamp results .
i took my opamp seperatly and connect the negative input ( in-) to the postive output (Vo+). in this way i use my opamp as buffer.and apply a sin wave of amplitude 0.4vpp with 2.5 offset. with freq=5meg on positive input (in+)
i am getting the output on Vo+ , as a sin wave with offset 2.5v but the peak amplitude is 2.6946v and 2.3051v. actually i should get theratically 2.7v and 2.3v.
so there is a approx 5mv error.
my question is having the gain of 106db why i am getting the 5mv error and how i reduce that much error within the range of 488uv.

i think u got my point..
waiting for ur reply ..
manish

 

[sankudey]

As you explained the error now, the first point of my answer does not fit. I thing from 2 and/or 3 you can get the result. Just for verification, could you check once what happens to a lower frequency sinusoid. Also you should check the step response of the opamp in the 2.3 and 2.7 V range and can get answear. I think the 3db BW is giving you the trouble.

 

[philipwang]

Dear manissri,
Understand.
As requirement of 12bit pipeline, I think ~80db open-loop gain is necessary,
for design margin 90db can be selected. So your problem maybe not come from gain but from bandwidth, why not decrease the gain and increase the bandwidth?
Another advice, please pay attrntion to the CMFB, adjust its capacitor and see what happen.
Try it.

Bg,

 

[manissri]

dear philipwang..
how i check the loop stability when cmfb circuit is connected in to the opamp .
and what should be the gain (differential or common mode ) of cmfb so that loop should be stable..
pls guide ..
manish

 

[wsy979]

hi all..
i am desingning the 12 bit 10 msps pipelined adc. for this i am using for S/H amplifier - folded cascode fully differential with gain boost opamp. i am getting 106db gain, P.M.-61 degree, UGB-261mhz.
my input range is 2Vpp. so my lsb is 2/4096 = 488 uv. my supply is 0 - 5v
my question is when i put my opamp in unity gain and apply a sin wave of of 1Vpp with 2.5v offset. theroatically i should get around 1Vpp with 2.5 volt offset.
but i am getting a 5mv error in my output.
and my lsb is 488uv.
so having such gain (106db) why i am getting 5mv error. and how i will rectify this problem so that i get the error withing 488uv range. so that i can desing my 12 bit adc.
waiting for ur reply ..
manish

dear manish

i am not very clear about your S/H schemetic, and your CMFB circuit.
Much things in S/H affects the gain error, including the specifies of OTA, the sample switch and sample cap. For the close loop OTA, its gain, UGB, input and output common mode voltage and the feedback factor affect the gain-error a lot.

i advise:
1. check the CMFB and input/output common mode voltage, make sure when OTA settled the voltages agree
2. check the feedback factor and the gate-cap of OTA's input transistor, make sure
1/(f*A0)<lsb
3.think about the sample switch charge injection

btw: make sure when you measure the error, OTA has already settled

hope those help

regards

Added after 11 minutes:

to manish

i think you'd better measure the op-amp with full input range 2V, maybe at that case the error increase a lot