On 4/3/07, Abhijit Menon-Sen <> wrote:
> At 2007-04-04 01:02:30 -0000, wrote:
> >
> > This process samples the analog voltage from the preamp at a very high
> > rate (usually 2.8 million samples per second). The output is one bit
> > only: a zero if the analog voltage is lower than the previous sample,
> > or a one if it is higher.
>
> When you first posted about the Korg recorders, I read their white paper
> on 1-bit recording, and I failed to understand this part. How come it's
> not important to specify the voltage differential more precisely?
As I understand it each single "sample" is not intended to encode an
actual value, just the delta from the last one, of course (when the
encoded value is not equal to the current value you nudge up or down).
I think the idea is that the sample rate is so high relative to audio
rate deltas that you can always represent them -- with an order of
magnitude or two extra samples to spare! The value that are output on
reproduction are only implied (by aggregating from a known reference
point at a fixed sample rate).
What I don't get is how true silence is encoded, wouldn't it be a 1.4
Mhz sawtooth... and how error correction could work to re-reference to
any absolute levels (I assume there's nothing analogous to "key
frames" in MPEG encoding...).
best,
aaron
--
quietamerican.org
oneminutevacation.org
83% happy
9% disgusted
6% fearful
2% angry
|