bhs@siemens.UUCP (Bernard Schwab) (11/28/84)
( line eater, line eater, eat me a line. Oh, while your at it, one for me too)
Of late, there has been discussion of the oversampling technique employed by
Philips, represented so ably in this country by Magnavox.
Well, there have been contradictory comments made, as well as plenty of I Dont
Knows. Thus, I decided to follow the matter to it's logical conclusion, and
consulted the Philips Technical Review, vol. 40, No. 6, 1982, pages 174-179.
There, those of you with similar Review before you will find an article
entitled "Digital to Analog conversion in playing a Compact Disc", written by
Messrs. Goedhart, van den Plassche, and Stikvoort. ( I wonder how names like
Atherton and Vesquerez must sound to the Dutch ). As this article is about as
close to the horse's mouth as one would wish to get ( figuratively, of course,
I would never call Philips a horse, while about Magnavox, however,...but now I
digress), I decided to quote some excerpts which may clarify the discussion
mentioned above.
So, for example, above mentioned gentlemen write: ( I qoute ) ( obviously )
"...This entire spectrum must not be passed on to the player amplifier and
loudspeaker. Even though the frequencies above 20 kHz are inaudible, they would
overload the player amplifier and set up intermodulation products with the
baseband frequencies or possibly with the high frequency bias current of a tape
recorder. Therefore, all signals at frequencies above the baseband should be
attenuated by at least 50 dB."
" To produce such attenuation, an analog filter after the digital-to-analog
converter will inevitably have to contain a large number of elements and
require trimming. "
At this point, let me interject that the Gentlemen know what they are doing,
the name of R. J. van den Plassche is often found in technical journals/
digests in conjunction with advanced a-d conversion techniques, the other two
names do not immediatly ring a bell with me. But, again I catch myself
digressing, so, herewith, I continue transcribing:
"In addition, a linear phase characteristic is required in the passband so that
the waveform of pulsed sound effects will not be impaired. In the Philips
Compact Disc player these requirements are met in a different way, by means of:
- fourfold oversampling of the signal in the digital phase,
-a digital filter operation,
-a hold function after the digital-to-analog conversion,
-a third order Bessel filter in the analog-signal path."
Now here comes the crux to the discussion:
" A digital transversal filter is used for the filtering after oversampling. To
understand the operation of the filter, we can think of it as consisting of 96
elements, while the delay in each element is (176.4 * 1000) ** -1s, i.e. a
quarter of the sampling period or 1/4 Ts. Four times in each period the filter
takes up new data. At three of these four times, the content of this data is
zero, since the oversampling is done by the introduction of intermediate
samples of value zero. This means that only 24 of the 96 elements are filled at
any one time. The contents of each element are multiplied by a coefficient c.
The filter provides data at a rate of 176.4 kHz; each number is the sum of 24
non-zero multiplications. In this way the filter always calculates three new
sample values at the locations of the zero samples."
"The practical version of the filter is in fact somewhat different from the
version referred to in the above explanation. In practice the filter consists
of only 24 delay elements and a 16 bit word remains in each element for a time
Ts. During this time Ts the word is multiplied four times by a coefficient c,
which is different for each multiplication. The products are also summed four
times during the time Ts and passed to the output. The frequency at which these
summated values appear at the output is therefore 4/Ts = 176.4 kHz) again."
"The coefficients are numbers with 12 bits. Each product has a length of
16+12 = 28 bits. The numbers have been chosen in such a way that the summation
of 24 products does not introduce extra bits, so that the filter output
consists of 28 bits with no rounding off....."
Comprende? As additional bedtime reading, our Authors recommend following:
B. A. Blesser "digitzation of audio: a comprehensive examination of theory,
implementation, and current practice" J. Audio Engng Soc. 26, pp. 739-771,
1978.
Whew, this is long. In fact, I am sure that some netters will have seen how
many lines the article has, and will simply have skipped it.
By the by, the above article is one of four in this particular edition of the
Technical Review, the others, titled: Error Correction and Concealment in the
Compact Disc System, The Compact Disc Digital Audio System, and Compact Disc:
System Aspects and Modulation, should be able to tell you more than you can
tell the girlfriend about digital audio. How facilly you can then justify the
purchase of one of these new-fangled machines with the killer beam inside.
Bernard H. Schwab
Siemens RTL
105 College Rd. East
Princeton, NJ 08540
(609) 734-6592
...!siemens!bhs
"Young man, you are drunk!"
"Madam, you are ugly. However, tomorrow, I will be sober."
commonly atributed to Winston Churchill (his younger
days)