|
Computer Audio Asylum Music servers and other computer based digital audio technologies. |
|
|
|
Computer Audio Asylum Music servers and other computer based digital audio technologies. |
|
In Reply to: Why So Many Conversions? posted by Todd Krieger on December 21, 2009 at 17:26:56:
> A 44.1->48 conversion at best wouldn’t sound all
> that great………. This is exactly what Microsoft kmixer
> does with CDs. . .
For some value of "exactly". A kid's record player does
"exactly" what a $50,000 vinyl rig does -- for some value
of "exactly."
I've **heard** the 44.1->48 conversion through Saracon.
It does sound "all that great". I wasn't, as I said, expecting
so much from this initial conversion (I upsample again
from 48->96, and then finally from 96->192.)
> The “click and crackle removal†suggests you’re sampling
> the music off vinyl.
No, the "click and crackle removal" I'm using on ripped CDs.
Hopefully, 99% of the time (at the extremely conservative
setting I'm using) it's doing nothing. Very occasionally,
it's getting rid of a click that would play havoc with
renormalizations of peak levels later on.
I was once using more aggressive levels (the preset suggested
for vinyl) because of what I seemed to hear as a
"smoothing" effect on the sound. Piano, for instance, got
"creamier". Yes, it was probably rounding off transients,
and I stopped doing it. I don't miss that effect, with
the latest chain -- I get the "creaminess" elsewhere, and
keep the transients.
> This is insane……… There are so many conversions (I counted five),
> it makes me cringe…….
I **knew** you, of all people, would appreciate it! ;->
> You’re in essence taking a 32-bit signal, downconverting it a
> few times to 16/44.1, and then upsampling it a couple times back
> to 24/96.
No. I can tell you're not that familiar with how Digital Audio
Workstations operate.
The **original** source is a 16-bit/44.1kHz CD, ripped to a hard
drive (with Exact Audio Copy, if that matters to you).
A DAW such as Steinberg Wavelab (or Sony Soundforge, or Adobe Audition,
or Cakewalk Sonar -- take your pick!), when it does **anything**
to the original signal (say, just changing the gain by 1 dB)
outputs a 32-bit floating-point stream (or even a 64-bit floating-point
stream, in some cases) to the next plug-in in its Effects (FX)
chain. If you were going to take the signal right out and burn
a new CD, you'd have to convert that 32-bit floating-point stream
back to 16-bit fixed point (via the application of dither, either
flat or possibly noise-shaped; it's a matter of hotly-disputed
taste what style of dither to use, and which sounds best:
Apogee UV-22, POW-R, MBIT+, etc., etc. At least its hotly-
disputed when you're dithering to 16 bits; the mastering
engineers care less about the details when you're dithering
to 24 bits, but you **should** still dither rather than
simply truncating, it's the "correct" procedure.)
If you're doing repeated processing on a digital stream, it's best
to avoid redithering until the very last step, if at all
possible. So if you've got Effect A that inputs 16 bits and
outputs 32 bits, followed by Effect B that can input 32 bits;
then it's best to keep the output of Effect A at 32 bits
to send to the 32-bit input of Effect B.
In practice, what you do will depend on the capabilities (or lack
thereof) of the boxes in your chain.
My initial processing (declipping and the dreaded click and crackle
removal) is in WaveLab, which takes in the 16 bits (at 44.1 kHz)
and **can** output 32 bits. The next thing in the chain is
Saracon, which **can** input 32 bits. So I keep the output of
WaveLab at 32 bits. Get it?
When it gets to the soundcard to be played back (through the DSPs
which are hardware boxes, and only operate in real time,
such as EDR*S), then I have to dither the 32 bits down to 24-bit
fixed at that point. I do **not** "downsample" the sampling
rate! The sampling rate is always going up, up, up!
The most radical bit-depth reduction is required by that EDR*S
system, which unfortunately is 16 bits in and 16 bits out.
(I thought when I acquired it that it would be 24 bits in and
out; nobody, not even the erstwhile Audio Alchemy folks,
really remembered the details from more than a decade ago. But
it ain't -- it's 16 bits.) So, I'm using the Meridian 518
(a well-regarded processor!) to do the dithering of the
24 bits from the soundcard back down to 16 bits for the trip
through EDR*S. The sample rate has been increased to 48 kHz
at this point, so the noise-shaping algorithm has a bit
more bandwidth to work with (that's a good thing!).
The pseudo-expansion from 16 bits to 24 bits by the Meridian 518
at the other end is simply what audiophiles (as opposed
to studio engineers) were doing with 518's at home back in
the late 90's. It was discovered that if your DAC can take
more than 16 bits in, then just "filling in" the lower 4
(for a 20-bit DAC) or the lower 8 (for a 24-bit "DAC",
which is what I'm doing -- there weren't many 24-bit DACS
back then) makes the sound "better" (increases ambience --
it's a subtle effect). The Genesis Digital Lens (and
for that matter, the Audio Alchemy DTI Pro32) all used
this "trick".
The next step in the chain is the dCS Purcell upsampling
from 48 kHz to 96 kHz. I'll be trying out Saracon in place
of that, too. I may not be able to keep it up -- that
will require **two** dubs instead of just one.
> I guess if you like how it sounds, any technical criticism
> shouldn’t matter…… But then again, I'm afraid you might not
> realize what you’re hearing…………
Look, I'm not exactly a babe in the woods here. I've got
a ridiculous number of DACs in the closet. Dare I enumerate
them? MSB Platinum III. Audio Aero Prima. Lector Digicode 2.24.
Camelot Uther Mk. IV. Cary 306 SACD (and the Cary Cinema 11).
Tube Technology Fulcrum. Esoteric D-05.
Right now, **all** my systems are on an optical bus,
sourced from a computer. The master clock from an Apogee
Rosetta 200 is fed back to the sound card (an EMU 1212)
through 192 kHz ADAT optical. Each system "taps" off
that long Hosa ADAT-over-Toslink bus via its own
Apogee Big Ben.
> All you need to do is sample your original source signal
> (the very first step) at 24/96 (if it’s analog), and play
> it back at the native rate. No conversions necessary. The
> playback will be a *lot* cleaner.
OK, now we've seemingly strayed into the realm of digitizing
LPs (or recording live music). That's not what I'm talking
about here. I'm talking about upsampling (and processing
along the way, through some "magic boxes" -- Audio Alchemy
and Perpetual -- and some "magic software" --
Burwen Bobcat) 16-bit/44.1 CDs. You know, those familiar
silver coasters, that we around here never play anymore
except to pop 'em in the computer for a quick rip.
> If you’re a brave soul, you should try playing the same source
> with zero conversions.
Speaking of zero conversions, I also have three (count 'em!)
zero-oversampling DACs. A modest Scott Nixon, a pretty
modest Audio Note DAC-1 (kit), and a totally immodest
Audio Note Signature 4.1x.
I have not, however, listened to a "naked" CD at home
since 2005, when I first discovered the joys of this sort
of manipulation on the computer. At that time, I was
using Eximius DVD2One. It opened my eyes. Funny thing
is, though, when I posted my discovery here
http://db.audioasylum.com/cgi/m.mpl?forum=pcaudio&n=45351&highlight=DVD2One&r=
I got a lot of responses along the lines of (yeah, so?
We've been doing that sort of thing for **years**) ;->
http://db.audioasylum.com/cgi/m.mpl?forum=digital&n=111132
http://db.audioasylum.com/cgi/m.mpl?forum=digital&n=111172
And guess who posted a great big harrumph in response
to my article back then?
http://db.audioasylum.com/cgi/m.mpl?forum=digital&n=111131
http://db.audioasylum.com/cgi/m.mpl?forum=digital&n=111133
This post is made possible by the generous support of people like you and our sponsors:
Follow Ups
- RE: Why So Many Conversions? - Jim F. 09:50:58 12/22/09 (0)
Top of Page ]
[ Contact Us ]
[ Support/Wish List ]
[ Copyright © 2025, Audio Asylum, all rights reserved ]
