I have often seen where the spectrum of quantization noise is assumed to be white? What is the justification for this? In what cases is it not true.
I have doubts on Davids explanation too.
Imagine a "perfect" ADC without DNL and noise and apply a linear voltage ramp to the input. Then the quanization noise should manifest with a sawtooth-waveform. Amplitude would be 50% of quantization intervall, fundamental frequency of the sawtooth would be determined by slewrate of inputsignal divided by the quantisation intervall. The spectrum of quantizaiton noise would be the line spectrum of a sawtooth waveform.
Adding DNL to this ADC should "modulate" or "jitter" the fundametal frequency. Also in real world situations the quantization noise can be "coloured", though the exact noise spectrum will indeed be very hard to predict.
If all of the ADC step sizes are uniform, the noise will be white. But real world DNL can color the noise, very difficult to predict as DNL varies from part to part. Robust designs choose an ADC that will put the system noise at least 3-5dB above the ADC noise floor to avoid the ADC dominating the system noise performance.
Well, you have provided an answer, but I am not sure it is the correct answer. Anybody interested in more information may contact me directly.
In response to Achim; the ADC output is sampled-time, so all those harmonics alias back into the Nyquist zone in random places, unless the sawtooth and ADC sample frequencies are related. This makes the noise spectrum look more uniform.
Has anyone looked at this issue when the signals to be converted are the I and Q components of MSK?
After some investigation, I think all the following conditions being simultaneously true is sufficient for the quantization noise to be close enough for practical purposes to being uniform in frequency:
Retrieving data ...