high noise floor in measured 12b ADC

[bobo123]

Hi,

I'm testing a 12b pipeline ADC. It's been designed to have a thermal noise floor below 72dB, however when I test the chip I end up getting a noise floor no better than 66dB (~ off by one bit). The histogram of the noise looks gaussian and in the frequency looks virtually flat (with a small tone here and there). Anyone have any suggestions as to what could be causing the higher than expected white noise floor? Much much appreciated!

 

[Gundam001]

Could you part name or part number of the ADC, so that we can review..

 

[Verba]

Can be noisy supply, bad routing, clock jitter.....
What is your test circuit ?

 

[bobo123]

the test setup is basically the ADC connected to a logic analyzer (TLA714), with batteries used as power supplies. I've shorted the inputs together and to the common mode voltage to isolate the noise of the ADC from that of the input buffer to the chip (AD8138). When I sample the ADC at 1ks/s (the ADC is rated for >50MS/s), I notice many large tones in my noise floor (with no ADC input). Is it possible that FM radio noise can be coupling into my circuit and having a significant effect? (BTW my PCB is a 4 layer FR4 board). Thanks!

 

[nathanee]

how many sample number you take to calculate FFT?
for a 12 bit ADC, you should take 2^13 sample at least.

 

[bobo123]

I'm taking 2^16 points.

 

[bcl]

Please check if your input is not exceeding the ADC input full-scale rating? Do you have grounding problem in the PCB design?