PrinceGaz wrote:
>Apologies for the very bad joke attempt in the "subject" field.
Mmm...
>What if a 22.05KHz sine wave is recorded such that the 44.1KHz
>sampling points fall exactly (or close to) where the waveform is near
>zero.
A 22.05KHz waveform can't be recorded because it is not less than half the
sampling rate. In thoery, any frequency under this can be digitised
accurately but in practise some leeway is allowed to compensate for our not
mathematically perfect hardware. Using 44.1KHz sampling, the practical and
implemented limit is 20KHz.
>Won't it just sample zero or next to zero every time, such that
>regardless of sampling methods or filtering, the end result will be
>a zero (or near zero) output?
Yes, but that won't happen for frequencies below or above 22.05KHz. For the
ones above, they will alias to frequencies the same amount below at the D-A
stage, which is why it is so important to remove these frequencies before
the signal is digitised.
>Is a
>wider bit-width employed and/or interpolation used to make it easier
>on the following analog filters?
A wider bit width isn't used for oversampling, that is left to schemes like
SBM, etc. Interpolation is used to generate more samples in the digital
domain, and the D-A is operated at a multiple of the actual sampling rate to
accomodate the interpolated samples. It is done entirely to ease the task
of the analog filters. Very steep slope filters introduce significant phase
errors which are undesirable. By oversampling, we gain several octaves to
apply gentle slope and more phase correct filtering. The interpolation is
performed by DSPs and will accurately plot a sine wave at any frequency
below half the sampling rate - there is no need to record these samples on
the meda. For more in depth information about how this is achieved, do a
search for "Nyquist theory" on the web.
-cb
-----------------------------------------------------------------
To stop getting this list send a message containing just the word
"unsubscribe" to [EMAIL PROTECTED]
