Conference cookie::notes$archive:cd_v1

See the special CD review issue put out by "Audio" for an article debunking
"Digital Myths", these among them.

Same issue contains several technical reviews, each of which has a
graph of distortion vs frequency.  All CD's do NOT sound the same!!

From:	50245::ORA          "Ora Jaervinen CSC Munich"   21-DEC-1983 04:20

CD sampling, quantization and dynamic range
-------------------------------------------

There seems to be lots of confusion and misunderstanding about the CD
technology (PCM coding). Digital audio recording is fairly new, but the
technology has been around for a long time e.g. in telephone networks.

Sampling frequency
------------------

The sampling frequency used is 44.1 kHz. In theory, this should be enough
for a bandwidth of 0-22.05 kHz if we had ideal low-pass filters. In practice,
the filters are not ideal, and the frequency band is limited to 20 kHz at the
upper end. Note that the sampling process (in theory again) does not introduce
any distortion into the signal.

Statements like "44.1 kHz is too low to reproduce fast transients" are
nonsense. An analog recording with a bandwidth of say 20 kHz cannot reproduce
any faster transients than a digital recording system with a bandwidth
of 20 kHz. There may be musical instruments with spectral components
above 20 kHz; whether you can hear the difference if these are cut off is the
question. As most of you know, the human ear is limited to about 20 kHz
(and for most of us who have served DEC for many years the limit might
more likely be around 15 kHz).

Quantization and dynamic range
------------------------------

There are many figures floating around. If you ask 10 people, you would
get 10 different answers (the highest figures from the CD manufacturers,
the lowest probably from British analog HiFi freaks).

The most common figure is based on the fact that using 16 bits you have
a range of -32768 to +32767. The lowest absolute value being 1 you can
thus represent a voltage range of 1 to 32767 which means 20*LOG10(32767) dB
or about 90.3 dB.

If you look at this closer, you can e.g. reproduce a sine wave with a
peak-to-peak amplitude of 32767 units. But try to reproduce a signal with an
amplitude of say 2 units, and it is hopelessly distorted.

This leads many people to the conclusion that the system is useless because
the distortion RISES towards lower signal levels (quite in contrary to analog
recording techniques). Certainly if you record a evry low-level sine wave on
a digital medium the distortion is high in terms of percentage or dB.

I read an article in a British Hifi magazine where the reviewer had a CD test
disc with a 1 khz signal recorded at a level of something like 85 dB below
theoretical maximum. Playing this back and turning the volume control all the
way up to hear the signal at all, he could hear distinct audible distortion.
What he forgets is that under normal conditions, with a certain amount of
ambient noise, that signal would not be audible at all even at very high normal
listening conditions. Assumimg say an SPL of 100 dB (rather loud), a headroom
on the disc of 5 dB that sine wave would have an SPL of 20 dB.

There's one more philosophical question: what is distortion and what is noise?
Both are components that were not there in the original signal. One would
probably define distortion as something deterministic, depending on the input
signal (no input, no distortion) and noise as something random.

With digital recording this is not so easy; the quantization distortion,
though certainly dependent on the input signal, is fairly random (in PCM
telephone technology, one usually talks about "quantization noise" and not
"quantization distortion"). As previously mentioned, the distortion (or noise,
as you wish) in a digital recording rises towards lower levels when expressed
in percentage. But if you look at the TOTAL foreign components of a reproduced
PCM signal, they remain more or less constant regardless of the signal level.
If you compare the two systems with this in mind, the results are quite
different: Think of recording a signal at 50 dB below maximum usable level
on PCM and on an analog medium: the PCM medium still has a good ratio of
original signal to foreign components, the analog one would have a pretty bad
ratio. Of course the noise is very different in both cases; in the analog
recording it is mostly "real" noise or hiss; in the digital one also, but to
a great extent quantization distortion or noise. As to which type of noise
is more easily audible I wouldn't dare to say.

Here are my gripes with CDs:

Hard edginess:

High frequencies that are digitized come out like square waves (similar
to those produced by a clipping amplifier).  To minimize this audible
and very irritating noise, CD decks filter out the harmonics generated.
The problem is that produces lots of phase distortion, hurting imaging.
No, 44.2 Khz is not good enough.  It isn't even a problem of frequency
response, I would be more than pleased if the high-frequency cutoff was
17 or 18khz, that's high enough even for young people.

Marketing hype:
Record company profits have been lagging the last several years, they
want to sell you your record collection a second time at three times the
price.  MidFi growth has slowed too, they want to sell this totally new
product.  The CD disk consortium tells you who is in bed together..

Marketeers determined that consumers like "high-tech" flashing lights,
bells, whistles, "features," and specs.  It is the same with cars that
talk and have "Star Wars" dashboards.  Consumers feel better about their
purchase decision when it is based on "scientific" reasons like specs.
Besides, it takes a long time to listen to all those players, its easier
to just plunk down your money.

Japenese recievers are designed to meet specs. that were set by the
marketeers.  It is alot harder for marketeers to sell based on a subjective
statement like: "It sounds good."  So, the designer's goal is not the
best sound, but the best specs..  Try looking at the 2 and 4 ohm power
outputs for Japenese recievers (if you can ever find them).  They are
all poor, which is why the bass sounds so thin and cold.  NAD is an
exception (American company, made in Japan) as might be some of the
"high end" Dennon etc.

CBS and others make recordings for teenagers.  They are the ones buying
the most records.  They have poor equipment and turn up the volume until
they get Total Harmonic Distortion.  Radio stations like constant high
volume so they have a wider range (greater advertisement reach).
So, what you get is over-miked, over-mixed, compressed 'music.'
How is the CD going to change that?  It does overcome the fact that
very few people use good turtables with good tonearms and good cartridges
and set them up correctly.  I heard a rumor that Polydor was going to
limit the dynamic range of its CDs to 65 db (same as records) because
of complaints from customers whose recievers are clipping.  So much for
increased dynamic range.  Most rock records only have 20db anyway.

By the way, S/N ratios and dynamic range are not the same thing.  A
30ips tape machine has a S/N ratio of 72 db, but people can hear 15db
below the noise floor, which gives you the same dynamic range as a CD.

Rags like Stereo Review who amplify the public desire to be modern and
sophisticated and "high-tech" (how I hate that word) by evangalizing
the "religion" of specs. for the profit of themselves, record companies,
and the manufacturers.  They almost never tell you how the equipment sounds
and always end by saying that X is a good buy if its what you want.  They
don't want to risk being honest at the expense of advertising money.

There are almost no substantial comparative reviews of CDs that tell you
the most important thing: how it sounds.  They don't usually tell you
how much phase-distortion there is, and if they do, they don't ever say
anything about it.  The reviewers don't say what the effect is, or how it
compares to other CDs.

There are alot of digital recordings comming out these days and few are
really good.  By "good" I mean good music, not good sound.  S/N ratios
do nothing to change the difference between a mediocre performance and
a great performance.  The record manufacturers have been scrambling to
produce new releases of classical music that have DIGITAL blazened accross
the jacket in order to sell another record to someone who already has that
piece.

I think that CBS, DG, and others are deliberately making bad records in
order to sell the CD concept.  Even their "audiophile" pressings can't
compare to a Japenese pressing off an analog tape.

There is alot of talk about CD as a concept, because it isn't perfected
yet.  Damn it, records arn't perfected yet!  Few people have the opportunity
to hear a good pressing on a state-of-the-art playback system.

Deep Throat said: "follow the money."  It is the same with the CD business.
There is big money in it for the manufacturers.  The no-money, no-advertising
audiophile magazines don't profit, they just want to protect good music.

eof(lame)
Mark Kaepplein (AMBER::KAEPPLEIN)

	I don't think you have to be an engineer from MIT to understand 
the possibility of quantization problems at high frequencies - although 
you might need to be one to explain them away.

	I'll start with a case that's extreme and only really valid for 
illustration (and otherwise getting the basic notion pounded down 
straight) - imagine what can happen to a 22.05 KHz signal being sampled 
at 44.1 KHz: depending on the timing (the phase relationship, to be a 
little more unintelligible) the signal may appear to be completely 
present or completely absent.  If the 44.1 KHz clock happens to catch 
the 22.05 KHz signal right on the crest and trough it will be recorded 
perfectly; if it happens to catch the 22.05 KHz signal at each zero 
crossing it will not be recorded at all.  If the signal is slightly more 
or less than 22.05 KHz it will fade in and out.

	Of course, the 22.05 Khz signal will be somewhere on the way out
(depending on the low pass filter on the CD player (Philips has a nifty
plan to break each clock interval into 4 subintervals and interpolate 
the data across these, making possible a less sharp, less phase shifty 
low pass filter)) and inaudible to most.  But at this point I think you 
can visualize the low frequency amplitude modulation that can occur near 
submultiples of the sampling frequency (22.05, 14.7, 11.025, 8.82 Khz, 
etc.), the depth of which decreases with decreasing frequency.

	Other interesting things happen at say, clock/4.6666 (9.45 Khz).
Because we are dividing the sampling rate by 14/3rds, a pattern will
repeat every three cycles in the reproduced signal, effectively
amplitude modulating it at Freq/3 (3.15 KHz).  If you remember beat
frequencies or heterodyning, you'll know this results in sidebands at
6.3 and 12.6 Khz.  6.3, 9.15 and 12.6 work out to a note, a fifth and an
octave.  Less pleasant intervals might be produced by real world
frequencies recorded on the disc. 

	Two things that won't, from experience, cause problems: square 
waves at high frequencies and square waves at low amplitudes.  Sit down 
with a friend's synthesizer sometime and play with harmonic content and 
listen to what happens as you increase the frequency or decrease the 
volume - in either case the character of the tone becomes less 
distinguishable.

	Another issue is how much these considerations apply to music
(as opposed to sine waves).  I have frequently read reviews that
complain of harshness at high frequencies in digital recordings.  Also,
Paul Klipsch (for what it's worth) claims there is an audible difference
between reel to reel recorders biased at 38 KHz and those biased at 100
KHz (there must be a reason people go to the trouble of biasing at
100KHz). My personal attitude is that good high fidelity digital
recordings should be made with sampling rates and resolution clearly out
of the range of debate - say 80 or 100 Khz and 18 bits.  Then various
prices of player can reproduce various amounts of niceness, according to
the buyer's budget. 

					Robt. Praetorius

Please don't confuse quantization and sampling. Quantization DOES cause some
distortion or noise even in the theoretically ideal case, sampling does not.
(I think it was Mr. Shannon who found that out).

Perhaps it is easier to visualize it like this: When you sample a signal
at 44.1 kHz, the result is a square wave with a frequency of 22.05 kHz and
rather random looking amplitudes (remember we are not talking about 
quantization but sampling). The spectrum of this signal contains the original
signal at 0 - 22.05 kHz and indefinite sidebands of it centered around
multiples of 44.1 kHz. These are filtered out and what is left over is the
original signal!

One must also remember that you can only reproduce sine waves if the
frequency of the signal is above 11.025 kHz. (anything else than a sine wave 
would have harmonics above 22.05 kHz). Also remember that the original signal
(before sampling) relly MUST NOT contain anything above 22.05 kHz, otherwise
your sidebands (see above) would overlap each other and you could not 
reconstruct the original signal with a low-pass filter. This means that the
incoming signal must be carefully filtered too (and as little noise generated
after filtering as possible, because its spectrum is not limited!). In practice
of course with non-ideal filters you have to limit your input signal to less 
than 22.05 kHz; I think CDs use 20 kHz which in my opinion is unnecessary high
(at least if the sampling frequency is 44.1 kHz).

In any case, the advantages of CD by far outweigh the disadvantages for me.
One must keep in mind that a GOOD analog player is more expensive than a 
CD player and you still have the problems of record wear, scrathes, care .....
Of course CD makes neither the music nor the master recording better, and there
are certainly lots of bad CDs on the market; but how many bad analog discs
are there??????????

(response to response 3):

I have a JVC amplifier, and it can produce full output to 4 ohms or 2 ohms or
even lower. What's so special about Denon anyway? Jus the image, in some 
countries their marketing people have been able to create a high-end
(high-price) image for them. In this country at least the Denon products are
quite cheap (I could have bought a Denon amplifier with 50% more power for
20% less than the one I have now).

By the way NAD is not completely American neither are all of their products
manufactured in Japan (the NAD analog record player is manufactured in 
Czechoslovakia).

It is a well known fact that human hearing is able to detect sounds that are 
actually below the noise level. Whether it is enjoyable to listen to a 
recording where the pianissimo parts are 15 dB below noise is up to you.
It is true that dynamic range and S/N ratio are not the same. For my taste
however the useful dynamic range of any medium is less than the S/N ratio;
in case of your 30 ips tape deck it might be somewhere around 60 dB.
For CDs it probably is somewhere around 70 dB or so, which I think is enough
for most purposes (I don't know where you live but my neighbors would sue me
if I had a real symphony orchestra in my living room or would play music
with a dynamic range of 100 dB). People often confuse dB's as the ratio of two
things and dBs of SPL (sound pressure level) which is an abolute measure.
I would guess that most living rooms have a usable dynamic range of 70 dB or so
(you would probably have 20-30 dB SPL ambient noise and your neigbours might 
tolerate 90 - 100 dB SPL).

My complaint about Japenese equipment is that most of it is targeted towards
high-volume mid-fi, where "consumers" pay more attention to specs than to
the sound quality.  This is partly due to the fact that the stuff flys
through high volume/discount houses that are always crowded, have poor
accoustics, and generally won't spend the time with you.  Typically, the
sales force is very transient having come from used cars or shoes, and
on their way to personal computers.

In this price/spec competitive environment, the Japenese use whimpy, switched
power supplies (reminiscent of the Carver M400 that I listened to a long
time ago) that make their products sound anemic and un-dynamic.

I just used Denon as an example of a Japenese company that had a high-end
line.  I mentioned NAD because I knew it was made in Japan and had gotten
good reviews.  I don't know any stores around that carry Denon's mono
amps, so its hard to see Denon as having a high-end.

I have been reading some CD reviews, and they help explain my negative
feelings.  The reviewers consistantly complain about harshness because
of the recording/producing tecniques.  Again it shows that record
producers are selling bad recordings because they are "Digital CDs."
It worked with digital records and its working with CDs.

It is the selling of the CD "concept" that all the companies are pushing.
(sentance continued) which worries me.  All the CDs look alike.  They
are all mass market items as common as the 50 watt digitally synthesized
receiver.  Their specs are close (what difference does .01% vs .05% make
when a turntable is 3% THD?).  ALL the reviews say they all sound the same.
Why are there so many?  It seems like a new market entrant would be hard-
pressed to find any niche to differentiate himself.

All the manyfacturers seem to say that "this is only the beginning",
    (      ^  a typo, perhaps a freudian slip - unintentional)
and that they are comming out with second-generation machines, BUT NONE
OF THEM SAY THAT THERE ARE ANY PROBLEMS!!!!  

	People associate "DIGITAL" with "Computers" and perfection.
	Computers don't make mistakes with numbers.

Manufacturers use this preconception to advantage when marketing digital
CDs.  They don't want to ruin it by acknowledging that the technology
has limitations or to a less extent that there are implementation problems,
(which doesn't injure the myth as much because it is due to human design
error).

It won't be until people acknowlede sonic differences in units before
anyone will try to build any improvements.

This is a fundamental step in any disciplain for progress to exist.
If you deny the problems you are never going to solve them.

(off flame)

According to the reviewers in Audio magazine, there ARE audible differences
between CD players.  They mostly have to do with how accurate the filtering
is, and how little phase distortion gets in.

< Note 2.1 by PIXEL::DICKSON >


See the special CD review issue put out by "Audio" for an article debunking
"Digital Myths", these among them.

Same issue contains several technical reviews, each of which has a
graph of distortion vs frequency.  All CD's do NOT sound the same!!

testing