Return to EQ index
All information in these pages is copyright (c) 1989-2002 by Roger Nichols. All rights reserved. Permission for personal reference only, and may not be reproduced by any method without written permission.
I Need a Digital Shrink
by Roger Nichols
I feel like talking about 96k converters. You won’t tell
anyone about this conversation, right? Should I lie at this end of the couch?
You see, I have always been into hi fidelity. Since I built my first stereo
out of wax paper and mud in 1959, I have been striving for the ultimate in
audio reproduction. With each improvement to my system I would re-listen to
my entire record collection and wonder at all of the new sounds that I uncovered.
Things were simpler then. If you wanted the best record changer
you bought a Gerard. If you preferred a single play turntable you probably
chose a Rek-O-Kut belt driven turntable with a Grado arm and cartridge. Speaker
choices were easy too. JBL Ranger Paragon, Klipsch Patrician, Altec Voice
Of The Theater, sand filled Wharfdales, University co-axial speakers…
and maybe a pair of Ionovac gas plasma tweeters. Amplifier choices were IECO
kits, McIntosh, Leak, and if you were lucky, a JBL SG-520 Graphic controller
pre-amp. Yeah, those were the days.
What does this have to do with 96k converters? Well, I was
getting to that. Back in the good-old-days if you read the specifications
of a piece of audio gear and looked at the frequency response curves, you
could pretty much tell what it was going to sound like. That isn’t true
any more. As an example, let’s say that two different companies made
an analog to digital converter. One said that their specs were: dc to light
±.001dB, S/N -124dB, THD .0001%, internal clock jitter <3 picoseconds.
The other company said: 1Hz-100kHz±.1dB, S/N –118dB, THD .001%,
jitter <10picoseconds. Which one would sound better?
Based on the specs by themselves, the first one should sound
better. But isn’t that like comparing a car that goes 200 miles per
hour with one that goes 240 miles per hour? There is no way that you are going
to be able to tell the difference driving down the freeway in Miami at 110.
Oh, 96k converters? Yeah, yeah, I’m getting to that.
What about the specs that aren’t specified? What about the stuff that
actually makes one converter sound different than another? In both the consumer
arena, and in professional applications, nobody ever talks about the “other”
specifications. What about pass-band ripple, power supply stability, low frequency
noise, dc hi-pass filtering by low frequency roll off or 2Hz digital servo,
converter linearity, nyquist filter design, zero crossing distortion, and
reference voltage accuracy.
You know, one of the problems facing us now is the quality
of digital audio and mostly CD players in the early ‘80s. 16bit converters
were 12bit or even 10bit linearity. Low-pass filters were not very good and
it was easy to hear artifacts in the digital audio. Another problem was emphasis.
Most of the early CDs were released with emphasis. This was a pseudo noise
reduction system that boosted high frequencies in the analog domain just before
the conversion to digital. The highs were then rolled off in the analog domain
just after the conversion back from digital. Every CD player, and later, DAT
manufacturer designed their own de-emphasis filter, which made every CD player
sound different when playing back the same CD.
Finally, 96k Converters
Well, I feel sort of funny about all this 96k stuff. I have
mixed to 96k and 44.1k and on real good systems you can hear a slight difference.
But you shouldn’t be able to hear a difference. The shortest cilia in
your ear won’t detect anything higher than 20kHz. It turns out that
it is not the content above 20kHz, it is what the low pass filter does to
the audio just below 20kHz. If I up-sample a 48k mix to 96k and play it on
the 96k system, it sounds better by the same amount. Now, there was no new
material produced in the transfer, but if you analyze the output of the analog
converters while comparing the mixes, you can see frequency response and phase
shift differences on the test equipment. This is because of the slope of the
nyquist filter. When a filter is that steep, it causes level and phase distortion
that you can hear.
Granted, the material recorded at 48k has already gone through one filter
during the A/D conversion, but if you up-sample and play it back through a
96k converter, you have only used the harsh lo-pass filter once instead of
twice. I have seen advertisements for high-end audio gear that touted that
process in their expensive CD players, but I dismissed them as foolish un-substantiated
claims. Now that I have done my own mixes and made my own tests I can see
what is going on.
You can see this difference also if you find that old CD player
that has been stashed away in the basement. The older the CD player the better
the test. Compare the sound of the old CD player with the sound of a new CD
player using the same CD. The new CD player sounds better. This is because
over the last 20 years filter design has improved and the lo-pass filter in
the new CD player is much better and much smoother than the lo-pass filter
in the old player. This is the same sort of thing that happens when you play
back a regular CD on a CD player with up-sampling 96k converters, or play
a DVD-Audio disc of the same material that was up-sampled for DVD-Audio release.
I was one of the guys yelling about record companies that were
up-sampling 44.1k CD releases for DVD-Audio. They were just doing it to put
out more product, but it turns out that they were accidentally doing something
good. I feel much better now that I have gotten all of that off of my chest.
Now, 192k… that is a different story. Nothing about 192k
can be better, can it? What? We’re out of time? I have to come back
next month? That’ll be HOW MUCH!!!!
---------------------------------------------------------------------------------------------------
Roger needs help with his hi-fi obsession. A few years ago he took his family on a trip to Hope, Arkansas. His family thought they were going there to see President Clinton’s house, but instead he took them to see the original Klipschorn factory.