Quote:
Originally Posted by lucid
Not if you want to be descriptive of what exactly is happening during the shift. It would also help identify the transitions better. It is possible that the first shift is longer than 35 ms for instance. But the noise needs to be reduced first.
|
I guess it depends on what the signal or event is...
I'm thinking it's the shift, not the noise...
and as swamp stated, he got good transitons and reversal of direction...
obviously more is better...but the Nyquist theorem is valid...it's how they cram so much signal on the net, phones, TV, etc. and not lose any information...
it would be better, as he stated, if filtering was applied...but that has risk too, filtering out signal along with noise...but if you only filter stuff outside 2x's the bandwidth you should not lose any information...
it's one reason 44.1khz was used for cd's, basically 2x's the range of human hearing of 20khz...
signal theory is manipulated in the f domain, not t....that way you can see the f's where the energy (information) lies...
I'd like to see the raw data transformed into the frequency domain with matlab
I bet the first one would have a spike at say 3f and f, discard the 3f...filter
result in t; flat line with an asymetric sine wave at the shift
the second one would have a spike at say 10f and f, discard the 10f...filter
resul in t; flat line with a pulse at the shift