| ??? 05/16/08 00:40 Modified: 05/16/08 01:29 Read: times |
#154806 - Sorry, but you are entirely wrong... Responding to: ???'s previous message |
Christoph said:
If you apply proper rounding, the LSB will stabilize. No, this has nothing to do with rounding, but with the unpredictable behaviour of noise, which is what finally makes the LSB flickering. Christoph said:
First set of 8 samples: 0111 1111 1000 0000 0111 1111 1000 0000 0111 1111 1000 0000 0111 1111 1000 0000 ------------ 11 1111 1100 Divide by 8: 0111 1111 Second set of 8 samples: 0111 1111 1000 0000 0111 1111 1000 0000 0111 1111 1000 0000 0111 1111 1000 0000 ------------ 11 1111 1100 Divide by 8: 0111 1111 The LSB does not flicker anymore. How should it ? A moving average of 8 samples has a zero as 1/2 the sampling rate. It will completely eliminate the flicker frequency. (oh, and this also works if you don't decimate. None of the output samples will have the LSB flicker, they will all be 0111 1111 if you do not round and 1000 0000 if you do round before the division by 8). You don't see the forest for the trees! Yes, of course, with that kind of fabricated noise your LSB flickering will disappear. But, welcome to real life, Christoph, in reality the byte stream is somewhat different, perhaps something like that:  Oops, Christoph, the LSB flickering is still there! Christoph said:
While your programming advice is usually very sound, you could benefit from reading up on digital signal processing. Nice attitude... Kai |
