From: Dan Dugan <>
>
> Walt wrote about reading Pohlmann,
>
>
>>I'm still working my way through sections of the book. And this section
>>I need to reread a few more times to get it all. But, if nothing else
>>he's definitely saying that anything like true 24 bit accuracy is going
>>to require a extremely high quality A/D section. Something that's not
>>too likely in lower price points. Makes me wonder about the inexpensive
>>stuff claiming 24 bit. They may be producing 24 bit, but not doing it
>>accurately.
>
>
> It's like when digital recording started. The code was 16-bit, but
> the A/D converters were only good for 13 or 14. 13 bits was as good
> as the best analog tape (w/o Dolby or dbx) and had lower distortion
> and flatter frequency response. It was problematic when quiet
> environments failed to dither the lower bits, and low-level sounds
> were munched or dropped out. I have a CD from the early days that has
> that problem.
>
> In a few years 14-15 bit converters were available, and we learned
> how to dither. Digital came of age and left analog in the dust.
>
> Now with 24 bit codes we again are working with converters that don't
> really deliver the whole range of the code, maybe 21 bits of audio in
> a 24-bit word. But 21 bits makes a dynamic range of 126 dB, and -any-
> environment or preamp noise will dither those low bits.
Of course the practical value of even 126 dB in day to day recording is
questionable. Certainly it's going to be the very rare combo of
environment/ mic/ preamp that's going to give that a workout in the
portable usage of nature recording.
And, what's more, the limitation this time is not design improvement,
it's the very components of the design, not just environment or preamp.
You can design a converter that spits out 24 bits, but that's not what's
meaningful.
Pohlmann seems to set the practical level at the equivalent of 20 bits,
and is very pessimistic about seeing any improvement. if so, the 24 bit
equipment is wasting 4 bits per sample, they won't contribute to the
accuracy. And that's under ideal conditions.
It would seem valuable for there to be a recorder that stored the actual
20 bit accuracy. More time could be stored in the same memory, and every
step of the storage/archive process would be that much cheaper. But, of
course any such recorder would be beaten in the marketplace by one that
put a higher number on it's label. As any sales department knows.
In spite of all that I've seen predictions of 48 bit and higher
recorders. Seems pointless. Like the faster and faster computers people
buy just to chat on email. Since sales departments tend to drive new
product choices I'm certain we will see the bit wars though. If they are
not already upon us.
This, of course, says nothing about processing, where the efficient
samples should be temporarily converted to much higher levels to cut
down on processing errors. What we are talking about here is the actual
value of the conversion from analog to digital. When one is looking at
specs for equipment it also means you need to be very clear what the bit
numbers they give mean. For the sales department they are almost
certainly going to give you the highest bit depth found anywhere in the
machine, regardless of how important it is.
Personally, I think the linear system is very wasteful. Regardless if
needed the bits are used. Future high end systems seem to be headed
toward non-linear or floating point which will be more targeted. Using
these other systems may result in more quality for the storage used. Or
more quality for the bandwidth available. If anyone can hear it. Each of
these alternatives has it's own limitations and noise characteristics so
it seems a little hard to predict where it will all go. Assuming of
course that anyone still cares about these increasingly fine distinctions.
As I've noted before, Pohlmann is worth a read. A great lot to think
about in there. Presented in a reasonably readable factual manner.
Walt
________________________________________________________________________
________________________________________________________________________
|