[net.audio] CD tutorial by Rogers Re: Warning technical stuff follows

mat@hou5d.UUCP (M Terribile) (02/18/84)

I have a Phillips (Magnavox) CD player -- the cheapest I could find at
$499 -- and I have noticed no unusual qualities in the sound.  On the
other hand, I should point out that on very quiet passages, the mechanical
noise of the player (about the volume of a cat breathing) is noticeable,
so the more interesting sonic effects may be getting overwhelmed.  The
motor is quietest at the beginning and end of a disk, and noisiest about
3/5 the way through a 40-minute recording.

The noise on the signal with nothing playing, or in a silent passage
is below the noise floor of my pre-amp.  The pre-amp is spec'd at 86 dB
below a reference input of (I think) 1 mV, so things are pretty quiet.

My machine is the Magnavox top-loader, and I am not sure if it has the
filtering scheme that has gotten so much discussion here.  I have heard
at audiophile salesman at Tower Records (check this place out, by the way)
say that ``Phillips botched the digital filters''.   It IS very easy to use.

Incidentally, I have disks by Telarc, CBS, Denon, and DG and the Telarc disks
seem consistantly to be the best.  One Denon disk came in a German-language
case, but with a Japanese brochure!  I can eventually make sense of some of
the German, but I am have no hope on the Japanese.  All of them have good
disks, but only Telarc seems to avoid really botching the recording from
time to time.

					Mark Terribile
					hou5d!mat

gregr@tekig1.UUCP (02/20/84)

Yes Shaun, you are correct the Sony doesn't use Bessel filters.  I will 
severely discipline my fingers for typing this error for which my brain
now steadfastly refuses to acknowledge responsibility on its part.  Geez,
how do these things happen?  Anyway, as I later pointed out the phase 
response of the Phillips approach is indeed closer to ideal than the 
Sony approach.  (Although not necessarily audibly so.)  Also I now see
Sony advertising claiming an 11th order filter rather than 9th in their 
units.  Does anyone have a reference to a technical article by Sony filling 
in the details of their filtering system?

Since you mentioned a desired out of band rejection of -70 dB in your 
comments it started me thinking about the rejection capabilities of the
two systems.  According to Phillips the out of band rejection for its 
digital filter is about -50 dB at cutoff with rejection decreasing fairly 
linearly to only -30 dB at 88.2 kHz.  The D/A converter holds the sampled 
values at its output between samples thereby introducing another filter 
response of the shape (sin x)/x with its first zero at 176.4 kHz.  This 
contributes only an additional -3 dB of rejection at 88.2 kHz.  Since the 
3rd order Bessel filter (yes Bessel) used by Phillips following the digital 
filter has a -3 dB frequency of 30 kHz the combined rejection of these three 
filter sources must be fairly constant between 30 kHz and 88 kHz at 
approximately -53 dB.  This nearly constant rejection occurs because the shape 
of the 3rd order Bessel filter increasing rejection with frequency is nearly 
inverse of the shape of the digital filters decreasing rejection with frequency
(up to 88.4 kHz).  The minor rejection of the (sin x)/x response up to
88 kHz simply fills minor differences between the other two.  Between
88.4 kHz and 154.3 kHz things are much better as both the digital filter
and the Bessel filter provide increasing rejection and the (sin x)/x
response is around -17 dB at the upper end.  Above 154.3 kHz however the
original audio signal reappears as sidebands around 176.4 kHz without 
any attenuation due to the digital filter.  At this point and beyond
the Bessel filter and the (sin x)/x response will continue to provide
at least 50 dB of attenuation.

So whats the point of all this?  Simply this, unlike the Sony approach
where the analog filters have a constantly increasing rejection with
frequency (up to some practical limits of the components of course),
the Phillips approach provides a fairly constant rejection of only
-53 dB or so over the complete bandwidth of the audio aliases.  Since
the aliases are not harmonically related to the original the presence
of these signals must certainly be more objectional than harmonic 
components at an equivalent amplitude.  The aliases themselves are too
high a frequency to hear but they can create intermodulation products
that could be audible if they have sufficient amplitude.  So at what 
level does this become audible?  Could it be heard if the aliases were
only attenuated by 50 dB or so?  When we perform standard THD
tests (which really measure all undesirable components, not just those
harmonically related) we must also be measuring these components.
Is it not possible to assume then that we could find the same measured
THD value more objectionable when associated with a digital player
of the Phillips type, than with a player of the Sony type which would
be more likely to have its THD comprised of the conventional harmonically
related components as with analog equipment?  I really don't know the answer
to these questions, YET!  I intend to try and find out.  I don't know
if this explains why some people think the Phillips approach sounds
inferior to the Sony approach or not.  I, like Shaun, have heard these
comments but I have no idea if they are factual or not.  I have not yet
even heard a Phillips type player.  For all I know the Phillips and Sony
players may be sound audibly identical and all this is simply an engineering
curiosity.

My reference for information about the Phillips system is the Phillips
Technical Review, Vol 40, No 6.  Has anyone else with more of a digital
signal processing background thought about these issues?  What say the
experts on the two systems?  I invite technical corrections, and technical
comments.  I'm bored with the rantings of the subjectivists.


					Greg Rogers