ADC Signal To Noise Ratio (SNR) Confusion

[Mechatrommer]

while searching for 1MSps ADC Chip datasheet. i look at SNR spec and most of them (at least 2 datasheet) showing similar figure. Below is the figure i got from 1MSps MAX153EWP (datasheet here). from what i know, that figure means its only usefull at 195.8KHz (KSps?) then what about and why it is speced at 1MSps? if the SNR at freq other than 195.8K are useless. or maybe i understand it wrongly, does it means the adc is feed with 195.8KHz signal, and the adc conversion is speced at around -80dB noise floor? what i expect is the SNR graph if the chip is to be feed by a range of input frequencies, say 0-1MHz sine input signal, and the graph will show the noise floor for each input frequency? is it too costly for that?

since i'm at this, actually i'm looking for 10bit 1MSps ADC 1Channel input with parallel interface output, SOIC or smaller pitch. still dont have luck finding the right one. if you know some chip to recommend, or have worked with it before, i will appreciate your suggestion. thanks.

edit: why 10bits? because i want to use 8bit only and since i expect there will be noise creeping in the low 2 bit, so i think i will skip the 2 LSB bit and get cleaner 8 bit in MSB, am i delusional here?

[ejeffrey]

That is showing the ADC sampling at 1 MS/s with a full-scale input signal of 195.8 kHz sine wave.  You always must measure SNR of an ADC with some signal.  Many ADCs have less than 1 LSB of random noise so if you connect constant DC voltage to the input you may measure zero noise due to the quantization error. 

There are two main things you can tell from this type of plot.  The first is the baseline.  That is the broadband noise that is independent of signal (as long as your input is properly dithered).  The second is harmonic distortion.  The peak at 390 kHz is the second harmonic of the test tone.  In principle you could also see spurious tones: narrow peaks that are not harmonically related to the test signal, but there don't appear to be any that rise above the noise floor here. 

[Mechatrommer]

thanx jeff. so what about test signal for other frequency? say at 500KHz? can i safely assume the same plot? -80dB other frequency SNR, and one harmonic at 1MHz at a bit lower than -60dB?