[net.audio] CD reflections

rjn@hpfcmp.UUCP (rjn) (01/02/85)

[] re: observations of a new CD owner (fairly long: ~100 lines)

I recently purchased a CD player (Sony CDP610ES), 20 discs and thought I
might share my initial  impressions  with you.  Prior to this  purchase,
most of my  listening  consisted of FM and some 1/4-in.  tapes made from
LPs.  Although I have an extensive LP collection, I don't play them very
often,  due to the  hassles,  not the least of which is having to endure
the LP ritual every 20 min.  or so.

====================== The PLUSES of CD listening ======================

NOISE:  There  is no  rumble,  clicks  and  pops;  no hiss if the CD was
digitally mastered; no pre-groove echo or print-thru.

DISTORTIONS:  There  is no wow or  flutter  if  digitally  mastered;  no
compression  'breathing'; no warp wow; no  inner-groove  distortion.  It
used to very much  annoy me when I found  that the  producer  had put my
favorite  selection  on the last cut of the side - a non-issue  with CD.
There is also no tracing  distortion  on loud  passages  (I don't have a
$500 cartridge).

WEAR:  No worry about the CD wearing out eventually.  No need to wait 24
hours before replaying (to allow the groove walls to recover).  No worry
about  stylus  wear.  The first time you listen to an LP, you know "this
is as good as this disc will ever be"; with a CD "it will always be this
good".

CLEANING:  I suppose CDs  eventually  require some cleaning, but nothing
like the discwasher/zerostat (or worse) ceremony LPs require.

CONVENIENCE:  I now  change  discs  once  per HOUR and can use a  remote
control to skip  around and program the disc, not to mention  being able
to answer the phone without  running over to the  turntable  first.  CDs
are  also  portable.  Using a  "Discman"  type  portable  player  is not
inconceivable.

TAPING:  If I  decide  to make  some  tapes  from my CDs, it will be far
easier than with LPs.  Cueing is trivial and I can prevue the entire cut
to determine the optimum record level, with no wear worries.

PACKAGING:  I don't have to stock special anti-static record sleeves to
replace the LP's paper ones many LP makers use.

HANDLING:  Although I intend to be as careful with my CDs as I have been
with my LPs (some of which are 20 years  old), I have far less  paranoia
about the inevitable little mishaps.

==================== The MINUSES of CD listening =======================

ENVIRONMENT  - CDs have pointed out to me just how noisy  refrigerators,
forced  hot air and  home  computers  are.  I may have to  'upgrade'  my
residence :-)

EXPEN$E - Yes CDs are costly  today.  I expect  that they will come down
to LP  prices  within  two  years.  Even  at  today's  prices,  I'm  not
complaining.  If  anyone  offered  a  "lifetime"  LP, I would pay the CD
price for it.

PHASE  SHIFT - If the 11 uSec  delay  represented  by  multiplexing  two
signals  (each at 44KHz) is not  present in the  ENTIRE  record/playback
chain, it may be an issue.  I recently read a source which  claimed that
this is at the  threshold  of human  phase  detection.  Of course,  this
would only be perceptible on earphones, since speaker placement distance
variation  easily  exceeds  11 uSec.  I haven't  done  enough  headphone
listening to know if I can hear any phase effects.

HIGH END FIDELITY - If we assume (and perhaps we shouldn't)  that all we
need to capture are complex signals composed entirely of symetrical sine
waves whose  highest  overtone is 20KHz, a (2x)  digitizing  rate in the
vicinity  of 40KHz just won't do.  For  example,  suppose we  digitize a
pure  20KHz   signal  at  40KHz,   and  happen  to   capture   only  the
zero-crossings.  How  much  information  does  that  get us?  We need at
least 3x  digitizing  (60 KHz) to  reconstruct  a pure sine  wave, and I
suspect that actual music demands that we use at least 4x (80 KHz).

Although  my hearing  is no longer  good  enough to have any  complaints
about CD sound, some of you golden ears are evidently hearing SOMETHING.
Keep complaining; it should be possible to develop a fully compatible 88
KHz (or better) CD that would play at 44KHz on existing equipment and at
a higher rate on a newer generation  machine (one simple way would be to
put the alternate  samples on the other side, read by a second laser and
digitally sync'd).

========================== The bottom line =============================

* I can finally listen (at home) to the MUSIC and not the MEDIUM.

* The CD system has the lowest hassle coefficient in audio.

* For a first implementation of a new consumer music technology, CD is a
  remarkable  technical   compromise.  The  first  music  cassettes  and
  4/8-track  cartridges  were a step  backward.  CDs are far superior to
  all other mass-produced media (I don't consider DBX disc/tapes and PCM
  tapes to be mass-produced).  CDs are probably on a par with audiophile
  LPs (all things considered) and have superior ease of use.

Bob Niland                                           [hplabs!]hpfcla!rjn
Hewlett-Packard                    Ft. Collins                        CO

gs@mit-eddie.UUCP (Gordon Strong) (01/12/85)

I'm afraid it is.  The 2x rate is called the Nyquist Rate and is
the minimum sampling rate necessary to prevent aliasing.  It has
been proven rigorously that any waveform sampled at or above twice
the highest frequency can be reconstructed.  This is part of the
sampling theorem and is the basis of much of modern communication
theory.  Since this is the accepted minimum, it is not the same
2x rate that is usually referred to in the CD literature.  The
highest audio frequencies are 22.1KHz, so the sampling rate must
be 44.2KHz.  When CD manufacturers refer to a sampling rate of 2x,
they usually mean 88.4KHz (that's what Yamaha says for my CD-2).

The use of a higher sampling rate (oversampling) is used
so that the low pass filter used to reconstruct the original 
waveform need not have a severe slope in the transition band.
The claim is that a gentler slope reduces phase and group delay
effects.  For those that care, the CD-2 using 2x oversampling
requires a 7th order filter to reconstruct the signal.  That's
still a bit steep for some.  Several manufacturers offer CD
players that use 4x oversampling.  I haven't heard of any that
use anything higher.  I don't know if anyone has anything in
the works or even what the theoretical limit is.  If anyone
has some information on this, I'd like to hear it.

Gordon Strong
{decvax!genrad, ihnp4}!mit-eddie!gs
GS@MIT-XX

jon@boulder.UUCP (Jon Corbet) (01/12/85)

[Finally! a use for my EE degree!]

>From: rjn@hpfcmp.UUCP (rjn)
>HIGH END FIDELITY - If we assume (and perhaps we shouldn't)  that all we
>need to capture are complex signals composed entirely of symetrical sine
>waves whose  highest  overtone is 20KHz, a (2x)  digitizing  rate in the
>vicinity  of 40KHz just won't do.  For  example,  suppose we  digitize a
>pure  20KHz   signal  at  40KHz,   and  happen  to   capture   only  the
>zero-crossings.  How  much  information  does  that  get us?  We need at
>least 3x  digitizing  (60 KHz) to  reconstruct  a pure sine  wave, and I
>suspect that actual music demands that we use at least 4x (80 KHz).

	Wrong.  In an ideal system, one needs to sample at no more than twice
the highest frequency component.  This is known as the "Nyquist criterion."
In reality, one needs to go a little faster, since we have not invented the
perfect low pass filter yet...this is why CD's use something closer to 44K.
This is theoretically enough to reconstruct PERFECTLY any signal that does
not have components greater than 22KHz; by limiting themselves to 20KHz,
the CD makers are actually giving themselves some slop.

--
Jonathan Corbet
National Center for Atmospheric Research, Field Observing Facility
{seismo|hplabs}!hao!boulder!jon

herbie@watdcsu.UUCP (Herb Chong [DCS]) (01/13/85)

In article <3411@mit-eddie.UUCP> gs@mit-eddie.UUCP (Gordon Strong) writes:
>I haven't heard of any that
>use anything higher.  I don't know if anyone has anything in
>the works or even what the theoretical limit is.  If anyone
>has some information on this, I'd like to hear it.

I know some people in our EE department here are working on techniques
for 10x oversampling with interpolation for digital audio signals, but
I am not sure what they are intending to do with them afterwards.

Herb Chong...

I'm user-friendly -- I don't byte, I nybble....

UUCP:  {decvax|utzoo|ihnp4|allegra|clyde}!watmath!water!watdcsu!herbie
CSNET: herbie%watdcsu@waterloo.csnet
ARPA:  herbie%watdcsu%waterloo.csnet@csnet-relay.arpa
NETNORTH, BITNET: herbie@watdcs, herbie@watdcsu

sjc@angband.UUCP (Steve Correll) (01/14/85)

>  HIGH END FIDELITY - If we assume (and perhaps we shouldn't)  that all we
>  need to capture are complex signals composed entirely of symetrical sine
>  waves whose  highest  overtone is 20KHz, a (2x)  digitizing  rate in the
>  vicinity  of 40KHz just won't do.  For  example,  suppose we  digitize a
>  pure  20KHz   signal  at  40KHz,   and  happen  to   capture   only  the
>  zero-crossings.

I believe the Nyquist criterion requires the sampling rate to be *strictly*
greater than (and therefore not equal to) twice the highest frequency of the
information. You are quite correct that sampling a 20kHz sinusoid at 40kHz
might capture only the zero-crossings, and thus lose information, even in
an ideal system. But sampling at greater than 40kHz (even infinitesimally
greater) will in theory represent the 20kHz sinusoid without loss of
information.

Incidentally, here's some hope for lower prices. Laury's in Chicago and
Sam Goody's in Washington DC were recently selling the older Telarc CDs for
$10. Maybe in another year or so all CDs...
-- 
                                                           --Steve Correll
sjc@s1-c.ARPA, ...!decvax!decwrl!mordor!sjc, or ...!ucbvax!dual!mordor!sjc

newton2@ucbtopaz.CC.Berkeley.ARPA (01/14/85)

You do *not* need a 3X sample rate to capture a 1X signal. If you sample
at "40 kHz" (as you say), you can capture every frequency (sinewave or
whatever) *less than* 20 kHz. For every signal *less than* 20 kHz, a little
thought will confirm that a 40 kHz equipartition cannot fall only on zero
crossings.

Actually, I take back the "or whatever" in the paragraph above. 

The sampling theorem isn't really up for debate, is it? Or should
we leave it to the same folks who reveal truth about creationism and
other subjects so subjective they need to be rescued from the "dogma"
of science.

Please excuse this intemperate flame, but it troubles me to see elementary
stuff so poorly disseminated and understood on a medium steeped in high
technology.

piety@hplabs.UUCP (Bob Piety ) (01/14/85)

> The use of a higher sampling rate (oversampling) is used
> so that the low pass filter used to reconstruct the original 
> waveform need not have a severe slope in the transition band.
> The claim is that a gentler slope reduces phase and group delay
> effects.  For those that care, the CD-2 using 2x oversampling
> requires a 7th order filter to reconstruct the signal.  That's
> still a bit steep for some.  Several manufacturers offer CD
> players that use 4x oversampling.  I haven't heard of any that
> use anything higher.  I don't know if anyone has anything in
> the works or even what the theoretical limit is.  If anyone
> has some information on this, I'd like to hear it.
> 
> Gordon Strong
> {decvax!genrad, ihnp4}!mit-eddie!gs
> GS@MIT-XX

It seems to me that the disc itself was created with a given sampling rate.
How can a player change this?

Bob

piety@hplabs.UUCP (Bob Piety ) (01/14/85)

> >  HIGH END FIDELITY - If we assume (and perhaps we shouldn't)  that all we
> >  need to capture are complex signals composed entirely of symetrical sine
> >  waves whose  highest  overtone is 20KHz, a (2x)  digitizing  rate in the
> >  vicinity  of 40KHz just won't do.  For  example,  suppose we  digitize a
> >  pure  20KHz   signal  at  40KHz,   and  happen  to   capture   only  the
> >  zero-crossings.
> 
> I believe the Nyquist criterion requires the sampling rate to be *strictly*
> greater than (and therefore not equal to) twice the highest frequency of the
> information. You are quite correct that sampling a 20kHz sinusoid at 40kHz
> might capture only the zero-crossings, and thus lose information, even in
> an ideal system. But sampling at greater than 40kHz (even infinitesimally
> greater) will in theory represent the 20kHz sinusoid without loss of
> information.


If the sample signal has no frequency components above 20KHz, whether
naturally or filtered, then capturing only the zero-crossings of this signal
by sampling at 40KHz CAPTURES ALL THE INFORMATION THERE!!  If there were
perturbations on the 20KHz sinusoid, there would have to be a 
frequency-component there ABOVE 20KHz!  If there isn't, then ALL the
information has been captured.  NYQUIST WASN'T WRONG!


Bob

newton2@ucbtopaz.CC.Berkeley.ARPA (01/15/85)

It is most certainly not true that

	Sampling a 20 kHz sinewave at its zero crossings (at exactly 40 kHz
	sample rate) captures "ALL the information". It omits certain
	minutia, e.g., the *MAGNITUDE* of the sinewave.

	I may not be a golden ear, but I *do* require a dynamic range greater
than 0 dB from my digital audio system!

gino@voder.UUCP (Gino Bloch) (01/15/85)

> > The use of a higher sampling rate (oversampling) is used [to simplify
    analog filtering]
> It seems to me that the disc itself was created with a given sampling rate.
> How can a player change this?
I wondered the same thing and asked the right person.  Fictitious samples are
added to increase the number of samples, then digital filtering and analog
filtering are applied.  Intuitively, you might expect the fictitious samples
to be interpolations of the samples already present, but what is done is to
intersperse zeros.  The subsequent digital filtering makes this OK, but to
prove it you have to write `FFT' on your hands with magic markers and then
do some hand-waving.  Find a friend who actually knows digital signal
processing (I don't, as I'm sure you can readily tell) and ask them (see
net.nlang) to explain it.
-- 
Gene E. Bloch (...!nsc!voder!gino)
Extend USENET to omicron Ceti.

wro@noscvax.UUCP (Michael Wroblewski) (01/16/85)

There is a 5(?) oversampling deck from one of those Japanese firms with a
four-letter word for a name (Aiwa?  Akai?).  I read it in one of the "above
ground" audio magazines in the new products section about a month back...
Somehow, a 5X oversampling doesn't sound right, but the manufacturer claimed
that the correction circuitry would commit only one error in twenty years!

Mike Wroblewski

There is a 5X(?) oversampling deck from one of the Japanese firms with a        four-letter word for a name (Akai?  Aiwa?).  Iread it in one of the "above
 I read it in one of the "above    
 ground" audio magazines in the new products section about a month ago....

mat@hou4b.UUCP (Mark Terribile) (01/16/85)

Oversampling:

	Yes, there is a certain sampling rate on the CD.  (Gee, didn't we do
this a while ago ...?)  The player inserts either one or three samples IN
BETWEEN the ones it gets.  These are created by a mathematical process from
the existing samples, and the lower bit (or two bits) of the sample get
removed.

	The data is then processed by digital and analog filters as though
it were created at the higher rate.  This raises by an octave or two the
frequency at which the anti-aliasing filters must cut off.  That in turn
changes the guardband (between the highest frequency of interest and the
alias that must be removed) from a couple of kHz to tens of kHz.  The filters
can then be designed for minimal phase damage and amplitude deviation in the
pass band ... but that pass band can extend up to (above) the new sampling
frequency.  This allows the filter designer (correct me on this, Bill Mitchell)
a place where he can put the anomalies in the filter's response (which MUST
occur).

	Y'see, there are a bunch of things that are just incompatable in
filter design.  The proper design of filters includes the usual kind of
engineering tradeoffs on these things, which are:

	Even response in the passband

	Slope of the filter (how quickly it cuts off)

	Quality of cut-off in the stop-band (if you need to get 100 dB S/N,
	you can't let lots of crud leak out here)

	Phase shifting of frequencies in the passband.

	Minimal noise created by the filter (the more electronics you add,
	the more noise you get).

If you can make the transition between passband and stopband big, you get
room to trade the other stuff off.

	Without oversampling, your top frequency of interest is 21 kHZ, your
sampling rate is 44 kHz, your alias frequencies begin at 23 kHz.  That gives
the filter designer just 2 kHz to play with on frequencies of 21 kHz.  Quite
a challenge!

	With 4x oversampling, your sampling rate goes to 176 kHz, and your
alias frequences begin at 155 kHz, but the top frequency of interest is
still 21 kHz.  We've gone from a guardband that was less than one-tenth the
bandwidth of the highest signal requency to one that is over seven times that
bandwidth!  Or, as a wise mathematician once wrote: ``somewhere ... a miracle
occurs''
-- 

	from Mole End			Mark Terribile
		(scrape .. dig )	hou4b!mat
    ,..      .,,       ,,,   ..,***_*.

shauns@vice.UUCP (Shaun Simpkins) (01/16/85)

> > The use of a higher sampling rate (oversampling) is used
> > so that the low pass filter used to reconstruct the original 
> > waveform need not have a severe slope in the transition band.
> > The claim is that a gentler slope reduces phase and group delay
> > effects.  For those that care, the CD-2 using 2x oversampling
> > requires a 7th order filter to reconstruct the signal.  That's
> > still a bit steep for some.  Several manufacturers offer CD
> > players that use 4x oversampling.  I haven't heard of any that
> > use anything higher.  I don't know if anyone has anything in
> > the works or even what the theoretical limit is.  If anyone
> > has some information on this, I'd like to hear it.
> > 
> > Gordon Strong
> > {decvax!genrad, ihnp4}!mit-eddie!gs
> > GS@MIT-XX
> 
> It seems to me that the disc itself was created with a given sampling rate.
> How can a player change this?
> 
> Bob

Comment on higher than 4x oversampling: 4x oversampling means that a sample
has to be delivered to the D/A every 5.6uS if separate DACs are used or every
2.8us if only one is.  Today's consumer-grade 16 bit DACs settle in 4uS,
preventing faster than 2x oversampling w/DAC sharing @ 16 bits (YAMAHA) or
4x oversampling with dedicated DACs @ 16 bits (KYOCERA, NAK, etc.).   Philips
uses only 14 bits - hence, their D/A settles faster and can be time-shared at
a 4x rate.  If you're willing to perhaps quintuple the price of the DAC and
add another $200 to the retail cost of your player, 16 bit D/A converters are
available with 350ns settling time, permitting 32x oversampling if you are
crazy enough.  At this rate, harmonics would be centered at multiples of 1.4MHz
and a 1-pole filter could be used for reconstruction.  However, another problem
would arise - and probably does now, too: the cycle time of the uP doing all
the error correcting and filtering would be unbelievablly small.  If my memory
serves me correctly, most machines use a master clock of 3.58MHz (TV color burstfrequency and 81 times the base data rate on the CD) with instruction processing
rates of perhaps 1/5th that.  Each sample delivered to the DAC represents at
least 4 processing cycles (I would assume); result - you can't do more than
4x oversampling with 3.58MHz clock rates.   Finally, cheap uPs max out at
around a 4MHz clock.

As far as the second comment above, it's been answered already: input data rate
doesn't change but `interpolated' data exits the processor at 4x.

The wandering squash,

				Shaun Simpkins

uucp:	{ucbvax,decvax,chico,pur-ee,cbosg,ihnss}!teklabs!tekcad!vice!shauns
CSnet:	shauns@tek
ARPAnet:shauns.tek@rand-relay

-- 
				Shaun Simpkins

uucp:	{ucbvax,decvax,chico,pur-ee,cbosg,ihnss}!teklabs!tekcad!vice!shauns
CSnet:	shauns@tek
ARPAnet:shauns.tek@rand-relay

ron@brl-tgr.ARPA (Ron Natalie <ron>) (01/17/85)

> 
> It seems to me that the disc itself was created with a given sampling rate.
> How can a player change this?
> 
Sampling has a funny definition here.  They mean number of numbers output
by the DtoA per second.  The problem is that the D to A transits from one
number directly to another number.  This sharp transition generates all
kinds of higher frequencies which must be filtered out (as a result of this
44kHz square wave you've just created).  You have to be really careful with
your filter to get rid of the things that are a result of the transition
without mucking up the encoded signal.  So, lets interpolate and generate
three intermediate steps in between each of the original samples.  Now
the step transitions are generating 176 kHz squarewaves, and you can be
a little sloppier with your filter because rather than being 22kHz above
the real signal you are 154 kHz above it.  You are infact doing some of
the low pass filtering in the digital domain because it's easier (i.e.
cheaper) to build that way.

-Ron

rentsch@unc.UUCP (Tim Rentsch) (01/20/85)

In article <boulder.264> jon@boulder.UUCP (Jon Corbet) writes:
>[Finally! a use for my EE degree!]
>
>>From: rjn@hpfcmp.UUCP (rjn)
>>HIGH END FIDELITY - If we assume (and perhaps we shouldn't)  that all we
>>need to capture are complex signals composed entirely of symetrical sine
>>waves whose  highest  overtone is 20KHz, a (2x)  digitizing  rate in the
>>vicinity  of 40KHz just won't do.  For  example,  suppose we  digitize a
>>pure  20KHz   signal  at  40KHz,   and  happen  to   capture   only  the
>>zero-crossings.  How  much  information  does  that  get us?  We need at
>>least 3x  digitizing  (60 KHz) to  reconstruct  a pure sine  wave, and I
>>suspect that actual music demands that we use at least 4x (80 KHz).
>
>	Wrong.  In an ideal system, one needs to sample at no more than twice
>the highest frequency component.  This is known as the "Nyquist criterion."
>In reality, one needs to go a little faster, since we have not invented the
>perfect low pass filter yet...this is why CD's use something closer to 44K.
>This is theoretically enough to reconstruct PERFECTLY any signal that does
>not have components greater than 22KHz; by limiting themselves to 20KHz,
>the CD makers are actually giving themselves some slop.
>

A small correction if I may.  The nyquist theorem states that sampling at
frequency 2F allows reconstruction of all information with frequency F or
less, but only if the samples are infinite precision.  Since finite
precision (16 bits) is used, the actual fact is that information gets
fuzzier (and so reconstruction gets worse) as the frequency gets closer to
the nyquist limit.

newton2@ucbtopaz.CC.Berkeley.ARPA (01/21/85)

	is that information gets fuzzier (and so reconstruction gets
	worse) as the frequency gets closer to the nyquist limit."

	Like so much opinion presented as "the actual fact" on net.audio,
this is just hogwash. "Fuzziness" is independant of frequency, up to or
even above the Nyquist limit, unless you consider fully precise aliased
components (resulting from sampling too seldom) to be fuzziness. Noise
and distortion are functions of the number of bits per sample and linearity.

	There *is* a fecund source of fuzziness WRT to digital audio,
but I'll leave it as an exercise for the reader to pinpoint its origin.

	Getting grumpy again (sorry)

	Doug Maisel

lauck@bergil.DEC (01/21/85)

>
>Re: a 2x sampling rate is not good enough to reconstruct a sine wave.
>
>I'm afraid it is.  The 2x rate is called the Nyquist Rate and is
>the minimum sampling rate necessary to prevent aliasing.  It has
>been proven rigorously that any waveform sampled at or above twice
>the highest frequency can be reconstructed.  This is part of the
>sampling theorem and is the basis of much of modern communication
>theory.  Since this is the accepted minimum, it is not the same
>2x rate that is usually referred to in the CD literature.  The
>highest audio frequencies are 22.1KHz, so the sampling rate must
>be 44.2KHz.  When CD manufacturers refer to a sampling rate of 2x,
>they usually mean 88.4KHz (that's what Yamaha says for my CD-2).

A perfect sine wave could be recovered if sampled at 2X plus epsilon,
but this would require a perfect filter.  (Sampling at exactly 2X won't 
work, for example, all the samples might be zero.)

The problem is that music doesn't consist of perfect sine waves and isn't
bandlimited to 22.1Khz, hence the need for anti-aliasing filters.  There
are two problems with these filters: 1) even if perfect they may filter out
musically significant information and 2) they can't be perfect.  Their
imperfections consist of amplitude and phase variation in the passband and
failure to completely eliminate the stop-band.  Similar problems exist in
the reconstruction filters on playback, but these filters have a less
difficult role, especially on players with two DACs where the digital
waveform before filtering is a staircase and not a collection of pulses.

The advantage of digital filters is primarily one of implementation;  the 
critical parameters of the steep filter can be implemented more precisely.  
In addition it is possible to build linear-phase digital filters, but these
may not be the answer since they have amplitude ripple in the passband.

As long as theoretical arguments are going to be applied where practical
arguments should be used, consider this:  the only causal signal which is 
bandlimited is the zero signal.  Such a signal could be processed perfectly 
by a digital recorder, but would only be useful to reproduce a single piece 
of music (by John Cage).

      Tony Lauck

               ...decvax!decwrl!rhea!bergil!lauck

gjphw@iham1.UUCP (01/22/85)

   So, okay, what bandwidth can we accept to provide good transient
 response and imaging?  Is it so difficult to design and build chips
 that can avoid introducing coloring into the audio range, given that
 most amplifiers only attempt to be linear for continuous sine waves
 between 20 and 20k Hz?  What bandwidth would be acceptable for
 recording - 20 to 45k, 1 to 50k?  Why not have designed a recording
 process that would have performed all the sampling and filtering magic
 far above the audio band, rather than allowing Carver an opening for
 his image enhancer?

   Also, what was so difficult about logarithmic encoding?  This would
 have yielded a uniform distortion due to quantization with signal
 strength rather than the *greater at low levels, less at high levels*
 that is obtained with the current linear encoding.

   I have no answers but only questions, yet I suspect the marketing
 department rather than the engineering department.  These issues are
 probably discussed in the relevant trade journals, but I do not read
 them.

-- 

                                    Patrick Wyant
                                    AT&T Bell Laboratories (Naperville, IL)
                                    *!iham1!gjphw

gregr@tekig1.UUCP (Greg Rogers) (01/22/85)

One more into the fray .....

How long can this argument about CD sampling rates continue to go on?  
Having watched this discussion the last several weeks I've never seen so 
much false, misunderstood information being used by people that clearly 
know very little about sampling and digital signal processing.  This
must be very confusing to readers who have an open mind about CD's
and are looking for some 'technical facts' to augment their listening
opinions.  I certainly don't want to silence the critic's of CD's
but I think a little responsibility is in order here.  If the critics
would go back to commenting on the sonic problems they hear, and quit
trying to 'technically explain' the causes of these problems, then 
the digital experts could quit wasting energy correcting all the 
half-truths and no-truths.  For those digital signal processing experts
that believe there are significant problems with the CD format, please
speak up and lets have an "intelligent educated" dialogue so at least
some of us might learn something relevent here.  To rephrase this 
simply  "lets all talk about what we really understand or ask questions
about what we don't, otherwise silence is golden".

By the way this isn't a new problem, the same thing happens with analog
audio.  People that don't really understand the technical issues often
seem compelled to validate their observations by giving totally false
misunderstood explanations for what they hear.  The difference is that
more people understand analog processing than digital, and its easier
in the analog domain to educate the untrained with intuitive arguments.
The result is the misunderstandings are much more quickly corrected
and continue to exist only among the hardcore 'flat earth' types.  This
in no way implies that their sonic observations are faulty, simply that
their explanations are invalid.  However, this can lead to some mighty
expensive ineffective solutions to their problems.  So you see there
is justice, the ignorant eventually pay ($$$) for leading others astray.
				
				Greg Rogers
				Tektronix
 
 
explanations 

rfg@hound.UUCP (R.GRANTGES) (01/22/85)

[]
I don't think Carver's contribution to digital disc audio has
anything to do with sampling rate or bandwidth or numbers of bits.
From what I've read it has more to do with what he did to his amp
to get the blessings of the golden ear fraternity..he made it have
a non-flat frequency response to match that of ...the Levenson (?).
THis same gimmick sells speakers. It sold JBL's for years. ...and
Altecs. ...and will go on selling things......"Hark! listen carefully!
Note that the glotch sounds clearer and more transparent in the upper
midrange...there...hear that?
"....well, by George, I do believe I do hear that...$$$."
If you have an equalizer you can play the game yourself. Just be sure you
only introduce a teensy-weensy bit of boost.

-- 

"It's the thought, if any, that counts!"  Dick Grantges  hound!rfg

rentsch@unc.UUCP (Tim Rentsch) (01/23/85)

In article <ucbtopaz.667> newton2@ucbtopaz.CC.Berkeley.ARPA writes:
>
>	is that information gets fuzzier (and so reconstruction gets
>	worse) as the frequency gets closer to the nyquist limit."
>
>	Like so much opinion presented as "the actual fact" on net.audio,
>this is just hogwash. "Fuzziness" is independant of frequency, up to or
>even above the Nyquist limit, unless you consider fully precise aliased
>components (resulting from sampling too seldom) to be fuzziness. Noise
>and distortion are functions of the number of bits per sample and linearity.
>

Sorry for the misunderstanding.  "Fuzziness" is, I admit, an imprecise term.
I chose it for its intuitive appeal.  The idea is that uncertainty in the
sample can result in error in the reproduced sine wave, and that this error
increases as the frequency gets closer to the nyquist limit.  Think of this
as another manesfestation of aperture uncertainty -- in the same way that
jitter in *when* the signal is sampled can result in error in the amplitude
of the sampled value, so uncertainty in the amplitude of the sampled value
can be interpreted as error in the time when the signal is sampled.  For a
given delta t of uncertainty in sample time, this uncertainty affects high
frequency signals more than it does low frequency signals, since perurbing a
sine wave by 1% (say) of its period will have more effect than perturbing a
sine wave by 0.1% of its period.

To set the record straight:  the explanation of fuzziness is my own.  The
existance of aperture uncertainty is sampling-theory-accepted fact.

(Hope you're not still feeling grumpy),

Tim

jf4@bonnie.UUCP (John Fourney) (01/24/85)

*** REPLACE THIS LINE WITH YOUR MESSAGE ***
The third harmonic of a cymbal???
Have you ever checked out the frequency spectrum of this inharmonic
sound source?  I would imagine it to look a lot like pink noise.