3rd Order sigma-delta Modulator's noise shaping

[neoflash]

I have no idea why the residue signal (difference between input and output of modulator)'s noise shape has a flattened region at high-frequency band?

Any suggestions?

 

[drizzd]

The quantization noise amplitude response for the 1st order sigma-delta modulator is sin^2(pi f/fs), which is rather flat at +/- fs/2. Is that what you mean? The amplitude response can be computed rather easily from the difference equations (cf., for example, this toy paper [1]).

[1] **broken link removed** (german)

Added after 2 minutes:

Ah, ok. I didn't see the attached figure before. That looks significantly different from the first-order one. Can you specify the model for 3rd-order sigma-delta modulator?

 

[neoflash]

I figured out that the NTF is indeed quite flat in that band. It is a 3rd order modulator with cascaded integrator and feed-forward path.

 

[FvM]

Above a certain frequency, the SD modulator loop gain gets < 1 and you see the unmodified quantizer noise at the output, which can't be but white.