My mistake ... my noise level was far too high. Additionnally, the noise with amplitude lower than LSB must be added to a full-scale range sine wave to be dithered amongst the various quantization bins. If anyone is interested, I updated http://jmfriedt.org/quantization.pdf which (I believe) demonstrates the impact of quantization on the noise floor.
JM -- JM Friedt, FEMTO-ST Time & Frequency/SENSeOR, 26 rue de l'Epitaphe, 25000 Besancon, France February 16, 2020 5:36 PM, "jmfriedt" <jean-michel.fri...@femto-st.fr> wrote: > I wanted to demonstrate quantization noise (V_q^12/12/fs noise floor) > with GNU Radio and somehow my demonstration is failing. I am trying > http://jmfriedt.org/untitled.png (or http://jmfriedt.org/untitled.grc) > following the chart of the Quantizer. Somehow I had already noticed > that the Qt Freq Sync is displaying a spectral density (because keeping > 1:N will not raise the noise floor by N as would be expected from aliasing), > but here the noise is obviously quantized (see time domain) but the > spectral density in the frequency domain does not show the impact of > quantization. > Is this expected ? > > Thanks, JM > > -- > JM Friedt, FEMTO-ST Time & Frequency/SENSeOR, 26 rue de l'Epitaphe, > 25000 Besancon, France