 If you re-open FFTMON (U_5), it says the SNR has been improved much (close to 50dB) because of eight bits quantization.
If the bit length is "16", the SNR would be close to 100dB.
If the bit length is "20", the SNR would be close to 120dB and almost the same quality as the one before quantization.
That's it for this class. If you design DSP or some other hardware stuff
with the SNR close to 100dB, the bits need to be more than 16 bits. The
SNR depends on how the bit length you use for the quantization and it's
mathematically (6.02 x Bit length) dB.
See also this class to see how to quantize filter's coefficients.
|