2 Feb
1999
2 Feb
'99
8:34 p.m.
No problem. The tube CD player simply has an AES/EBU
(or S/P-DIF)
coaxial
digital output. If you want a tube D/A to go with it,
that's a separate
product.
The tube D/A would likely not use oversampling, but rather an 8-pole
analog filter.
If people want a tube oversampling D/A, that can be a more advanced,
larger,
and more expensive model.
I agree with Tony. From an audio point of view, chips vs. valves (tubes,
for those colonies who've forgotten how to speak English) should make no
difference except in the analogue sections. On the digital side they have
no more than old computer technology hack value.
That said, oversampling and filtering is _not_ hard. People nowadays think
only in terms of digital, discrete working and analogue, continuous
working. IMHO the place to do oversampling is in between - in the
Analogue, Discrete domain.
How: Make several A-to-Ds (or one and some sample and hold circuits if
you're a cheapskate). The simplest case is linear interpolation which uses
only two. I would envisage precision ladder networks or something similar
for these. Call them A and B (and C, D, etc.)
Have some analogue circuits which use analogue summation techniques to
derive (say) A, 0.75A+0.25B, 0.5A+0.5B and 0.25A +0.75B. Switch between
them. You now have 4* oversampling, with simple linear interpolation.
(After those four valuse, B becomes the new A and the next value from the
digital side becomes B...)
More complex filters can be implemented with more A-to-Ds (or a series of
sample/holds or a bucket brigade etc) and more complex maths in the
combinations.
IIR filters can be implemented with a sample and hold that remembers the
last _output_
And so on. (but dion't forget the conventional analogue filter on the
output)
No DSP required.
Fun,eh?