Is SNR an important parameter, if the signal is DC?

[jebaspaul]

Hi,

I got into a strange doubt while doing signal conditioning at the input of ADC. The Signal to Noise ratio is defined as the ratio of average signal power to the average noise power.
SNR = Psignal/Pnoise ; Where P is average power
(or)
SNR = (Asignal/Anoise)^2 ; Where A is rms value

Let us take the second equation for our analysis. Consider the signal as DC. The signal line has some high frequency noise which is not filtered out(it means that there is no LPF ) before reaching the ADC. According to the definition of SNR,

Signal RMS value = Adc (It is a DC value)
Noise RMS value = An (say, the noise is a sine wave with some higher frequency)
SNR = (Adc/An)^2 --> (eqn1)

case 1. Consider the sampling frequency which is lower than the noise frequency. - Will that calculated SNR affect the ENOB of ADC
case 2. Consider the sampling frequency which is higher than the noise frequency. - Will that calculated SNR affect the ENOB of ADC

My argument: In both the cases there is a probability of sampling the noise along with signal. In case 2, the noise aliases back to the signal frequency. Even in case 1 the probability of noise getting sampled along with signal is not zero.

So, in the scenario like case1 what kind of analysis to be carried out in order to estimate the SNR at the input of ADC.

Thanks,
Jebas.

 

[barry]

I think there's a bit of confusion here (at least, I'm confused). First of all, SNR for an ADC is specified at a particular input condition, usually full scale; it's also specified at a particular frequency. You don't ordinarily specify it at an arbitrary amplitude as you seem to want to do. Obviously, if you specify your input to be some very low level your SNR will be a lot worse than if you specify it at full-scale. The fact that your signal has some noise on it has nothing to do with the SNR or ENOB of the device, those parameters don't change because the outside world changed. Since you're applying a non-band-limited signal to your input, I think you have analyze what the aliased signal is going look like. You'll need to know bandwidth of the ADC (which is probably a lot higher than the sampling frequency).

 

[crutschow]

As barry stated, the ENOB is the intrinsic noise level of the ADC and doesn't include any external noise from the signal.

Any noise in the signal will be included in the ADC output if it's not filtered. This noise will be seen in the output, independent of its frequency ratio to the sample frequency (the limit being the frequency response of the ADC input sample & hold circuit).

Of course you can digitally average the ADC output samples to attenuate this noise.