- Joined
- Feb 23, 2016
- Messages
- 20,771
- Likes
- 37,635
Here is a graph from Stereophile's review of a Project Box S2.The 42 dB shift makes sense, but why would it apply to the pass band only and not also to the stop band?
Notice the red level of his - 4 dbFS pink noise is -40 dbFS until the stop band. And how the stop band noise level is about the same as when he is using the 0dbFS 19.1 khz tone in blue. That is what is really happening.
I'm thinking REW reset its 0 db level when you told it to generate 0 db pink noise. And that in fact its 0 level should read -42 dbFS or close to that. When you tell it to generate 0 db tones it is also 0 dbFS. So while done this way it looks different the actual levels have to be like in JA's graph above. You know it can't have 0 dbFS levels at each frequency over the band of the pink noise because that total signal would be well into clipping if it were true. It would be like having 100 hz at 0 dbFS plus 200 hz at 0 dbFS plus 300 hz at 0 dbFS so and so forth with that actually true at every single hz. The total signal would be way too hot.
BTW, it would be a good idea to use - 4db FS for the noise signal like JA does. Pink noise has some variation in level (think peak to RMS ratio) of up to 10 db. Most of it is only a little over 3 db. So doing this at -4 dbFS means rarely would there be any clipping. Done at 0 dbFS clipping will occur now and again. This is most noticeable when sampling the analog result at a different sample rate than the signal is generated or when oversampling is done.
Here is a 44.1 khz white noise 1 second each at -4 db, -3 db, -1 db and 0 db. Then the same resampled to 384 khz. If your ADC and DAC are separate you of course could adjust level at the ADC to mitigate this. But the lower level reduces the odds of inter-sample overs that some DACs don't handle correctly. Red is clipping of course in this image below.
Last edited: