kimr@tektronix.UUCP (Kim Rochat) (07/29/83)
In the July "High Fidelity" and the August "Audio" there are test reports
on 4 digital disk players. Both reviewers had gotten the Philips test disk
which has square wave and impulse tests. All fours reviews published
oscilloscope photographs of the 1 Khz square waves and impulse tests. The 2
players reviewed in Audio demonstrated a large amount of ringing on
the tops of the square waves and after the impulse. (I was vastly amused
because the 2 DAD reviews were separated by a review of a B&O cartridge whose
1 Khz square wave response was vastly superior to either of the DADs). The 2
units reviewed in High Fidelity showed much better square wave performance
(of course one of them was the Philips/Magnavox player, explaining what the
tests were doing on the disk). Also, one unit in each magazine showed
inverted phase on the impulse test. Audio also measured 2Khz/20Khz phase
shift at 180 degrees for one unit.
My question to all you digital-philes is "Are you going to pay any attention
to these test results?" Presumably, all digital players) sound alike
(read perfect) and you're supposed to choose one on the feature/price ratio.
Are you going to buy the one that rings less and has correct phase and minimal
phase shift? Are you going to audition the competing units and decide which
one "sounds" best?
Good listening,
Kim Rochat
tektronix!kimrmichaelk@tekmdp.UUCP (Michael Kersenbrock) (07/29/83)
How much phase shift does a "typical" speaker system introduce ( I have AR9 and B&O M100 speakers if you know those numbers) as compared to the phase shift in a CD player? Kim Rochat mentioned 180 degrees at 20 KHZ for some CD player (I have a SONY). It seems to me that if I move my head back (from the speakers) about 0.3 inches, that I have caused a 180 degree phase shift at 20Khz, but less than 10 degrees shift at 1KHz, so does a 0.3 inch headmovement have the same effect as the total error in the CD player process or am I missing something? Anyway, how much phase shift is caused by the speakers, and how much by headmovement, and how do these compare in maginitude to the shift caused by a CD at very high (audio) frequencies? Mike Kersenbrock Tektronix Microcomuter Development Products Aloha, Oregon
ee461@rochester.UUCP (VLSI class) (08/01/83)
Mike Kersenbrock asked recently: "How much phase shift does a
"typical" speaker system introduce (...) as compared to the phase
shift in a CD player? (...) It seems to me that if I move my head
back (from the speakers) about 0.3 inches, that I have caused a
180 degree phase shift at 20Khz, but less than 10 degrees shift
at 1KHz, so does a 0.3 inch head movement have the same effect
as the total error in the CD player process (...) ?"
A few weeks ago I started to look for new speakers for my system,
and did a little literature research followed by some speaker
auditioning after finding out that some speaker manufacturers
advertise their products by using terms like "coherent
phase and time response" (which had no meaning to me at the
time). What I've found out about this may provide an answer
to Mike's question and it may be of interest to anybody who is
considering buying new speakers.
First, it appears to me, that moving your head does not cause any
differences in the percepted sound; you introduce a time delay
rather than a phase shift. Hence the sound reaches your ears with
exactly the same relationships between various frequencies as
it would if your head was 0.3 in. closer to the speakers. It would
make a difference only if the acoustic field was a standing wave,
and with most music material it won't happen.
Much closer effect to a phase shift between different frequencies
(180 degrees was reported for some CD players) can be obtained
by reversing the polarity of one speaker in the stereo pair. This
is an exagerration, but the basic effect is the same - the
sound appears "smeared" and originating everywhere, as the
directionality is lost.
In almost every speaker on the market, several drivers dedicated
to reproduce different frequency ranges are used. The separation
of these ranges is never ideal; e.g., if the crossover frequen-
cy between the woofer and the midrange is 500 Hz, you may ex-
pect that the range 400-600 Hz will be reproduced by both
drivers. If there is a phase misalignment, the cones of both
drivers do not move synchronously and the result is much the
same as with the speaker polarity reversed, although to less extent
because of the narrower frequency band. I think that the effect
of a continuous phase shift observed in CD players should be
similar but less noticeable than the effect introduced by speak-
ers. This, however, does not mean that it will be negligible
when good speakers are used. Personally, I find the lack of spa-
tial definition in the sound image rather annoying and I'd prefer
that my CD player would not contribute any phase shifts.
The phase shift in multi-driver speakers is introduced mainly by the
crossover network, where usually filters of 3rd or 4th order are
used. The consequence of this is that phase shifts up to 270 or
360 degrees are obtained. Although 360 degree shift at some
frequency means that the cones in two speakers move synchronously
at this frequency, one of them lags a full cycle behind the
other, which will have unpleasant transient effects (and essen-
tially all music consists of transients). For example, when you
feed a single voltage pulse into the speaker, you get two sound
pulses from it, one from each driver. This effect is diminished
by the fact, that one from the two signals with large phase shift
between them is simultaneously attenuated by the high order
filtering network. However, lower phase shifts are less at-
tenuated and still measurable and audible.
A single driver speaker can be designed so that it will not in-
troduce any phase shifts. This is mainly because of the
fact, that the phase shift is a relative quantity and
having only one driver gives you no reference point. I am sim-
plifying here; as the membrane in the speaker is not rigid,
various frequencies are reproduced differently, resulting in some
phase shift as well as in efficiency differences and
non-uniform frequency response. Exactly these problems cause
that most speakers use different drivers for different
ranges (as it is easier to design a driver working
correctly over a narrow bandwidth), and crossover networks to
separate ranges reproduced by different drivers. And there
comes the phase shift.
Ideally, one driver should be enough for the perfect reproduc-
tion, and there are some speakers using full frequency drivers
with no need for the crossover network. Bose 901 (or older
2201) are an example; I think that Tannoy also used to make one
driver designs; also I've heard about some laboratory standard
reference speaker from Bruel & Kjaer that was to be mounted in
6x6 ft board (no enclosure!) and used in anechoic chamber for
microphone calibration.
Most of the problems with the phase response is attributed to the
high order crossovers. Theoretically, only the 1st order cross-
over can be designed so as to provide flat phase response (a
proper driver alignment is also necessary). The problem is that
the drivers used in such a speaker must be designed for perfect
reproduction of a much wider frequency range than when high ord-
er, steep slope filter is used.
Final remarks of this lengthy submission: I recall only two
brands that use first order crossovers, which, by the way, are
exactly the two brands I am having a hard time to choose
between. These are: Vandersteen IIC and Thiel 03a. Their phase
response is flat to within 10 degrees. Most of the other
manufacturers do not publish the phase response as usually they
have nothing to be proud of. I recall some reviews where the phase
response flat to within +-40 degrees was considered to be quite
an achievement.
I would recommend auditioning Vandersteen's or Thiel's to anybody
who is considering spending $990 or more on a pair of speakers.
(Vandersteen's are $990 a pair, Thiel's are $1100 a pair). I have
compared them to IMF Reference <some number I don't recall> selling
at some $3600 a pair or so, and they were clearly better than the
IMF's in imaging, and most of all, in "naturalness" of the sound.
I'd say that Vandersteen's give the reproduction that is the closest
to a live music I've ever heard. Try listening to the "Sheffield
drum record" - you'll be able to distinguish every drum in the
percussion set by it's position (in all 3 dimensions and without
Carver's sound holography). On the same day but a few hours later,
I also heard KEF105.2 and B&W802, (some $2800 a pair). There
is no comparison - Vandersteen's outclass them by a wide margin
(n.b. KEF and B&W use 4th order crossovers). All faults of KEF's
are a material for a separate submission. Anyway, B&W (advertised
as: Best in the World) stands rather for: Boomy Woofer, at
least this was my impression.
Krzysztof Kozminski
(ee461@rochester)newman@utcsrgv.UUCP (Ken Newman) (08/03/83)
There is an excellent article in the July 83 High Fidelity, p. 17, called "CD sound: Trouble in Paradise?", by E. Brad Meyer, president of the Boston chapter of the AES. It seems to hit some interesting nails on the head regarding poor sound quality from some cd disks. He talks mainly about recording techniques being unsuitable for cds. The ringing in square waves from cd players is caused I believe by the very steep filters used to attenuate response above 20 kHz. Remember also a square wave is a very severe test signal and a good square wave at frequency f indicates good response out to roughly 10f. Also the square waves present on most test records probably have some ringing themselves. While we're talking about phase response in speakers, an interesting technique is/was used in the Ohm F speaker. These things have a single driver, a bullhorn-shaped cone essentially that opens downward. The cone is made of several different precision materials of carefully controlled densities, and is excited at the top around its perimeter, causing a circular wave to travel down the cone. Short wavelengths dissipate quickly, and radiate from the top, long wavelengths travel the whole cone and radiate from the full length. The result is a completely phase-coherent cylindrical (360 deg) sound front. What's even more amazing is that this apparently worked (they were kind of inefficient though). K. Newman decvax!utzoo!utcsrgv!newman