a simple question for pipeline adc

[fencl]

Hi all,
it's about pipeline adc s/h stage .since i use full differencial structure for it that means vin and vip,after sample and hold ,got voutn and voutp.
we all know sfdr is important to adc performance.so i got sfdr of s/h.the question is sfdr value: voutp=58db,voutn=58db(influence by even harmonic,special 2nd).but voutp-voutn>100db.
i'm confused about it, since adc will convert voutp-voutn to digital code.is it means i only need care voutp-voutn ? but it seems voutn and voutn itself sfdr is also important ? so how to estimate it ,such for 12bit adc.i find most paper don't give their sfdr come from singal voutn or voutp or vout=voutp-voutn ?

BRS

 

[JoannesPaulus]

the second harmonic, and the even harmonics, will go down in a differential system (especially in simulation).
SFDR is important as it is a measure of linearity (somewhat correlated to the INL). Many papers do show DNL/INL therefore SFDR is not needed.

 

[fencl]

the second harmonic, and the even harmonics, will go down in a differential system (especially in simulation).
SFDR is important as it is a measure of linearity (somewhat correlated to the INL). Many papers do show DNL/INL therefore SFDR is not needed.

Yes ,you're right about SFDR with INL(nearly 20log(2^b/INL)),and SQNR related to DNL.
i just care about simulation spec result ,not refer to testing after tape out,so we care sfdr rather than INL/DNL.
but my question is, how these paper's SFDR data come from. singal out? or diff out after minus?since SFDR is quitely different by 2 ways datas.
We can find if sfdr>80db is quite well from many paper.so i feel this "SFDR" is most probabily means single voutp or voutn.is it right?(diff out after minus is so easy to this value)
but if it is right.i have a question,since adc finally convert diff out after minus to digital code,why we need so care about single out ?
pls help me!

 

[JoannesPaulus]

No, you only care about the differential signal (how can you measure only one of the two?). Moreover, papers usually show their best results, so one more reason to show the differential signal results only!

 

[fencl]

No, you only care about the differential signal (how can you measure only one of the two?). Moreover, papers usually show their best results, so one more reason to show the differential signal results only!

Hi, JoannesPaulus
All i mentioned here is about S/H stage for pipelined adc. so i just measure data after S/H, diff sin input with diff sin output. it's the same way for one or differential signal(they're all sine signal).
Since it's easy to get high SFDR of S/H(>90db) for differential signal ,so i suspect what they mentioned or what we should also care at least contain single end data.could you share more experience about S/H of adc ?
thanks and regards