William Sommerwerck said:
Please define what _you_ mean by comb filtering.
A choppy response curve that is the result of suckout cancellations.
It may have nothing to do with any inherent problems with a
line-source driver itself, but instead simply is a result of the fact
that signals from the large, flat surface area of the line cannot all
get to the ears simultaneously. Because of this, at certain (but not
all) frequencies you get cancellation nulls.
That has nothing whatever to do with what I said. Read it again.
I am at a loss. Does the driver have to be damped by the air load? I
was under the impression that damping is mostly accomplished by
electrical characteristics and the suspension materials in the driver.
As long as a driver can deliver flat output over its
crossover-controlled bandwidth it will be sufficiently damped.
An interesting, theory, as the larger a driver is with respect to the wavelength
it reproduces, the greater its efficiency (in the sense of making a good
impendance match with the air).
Actually, its efficiency is reduced, due to the increased mass and the
beaming and comb-filtering artifacts.
Then it's not a line source, so how does it demonstrate your argument.
Actually, it was two lines, with a short tweeter line (about a foot as
I recall) and a midrange line somewhat longer (about two feet). The
tweeter line was quite narrow (about half an inch) and the midrange
line was considerably wider (several inches). It certainly qualifies
as a line-source system. Where it surpasses systems that have very
long lines (or one long line) involves the shorter distances from each
segment of the line to the ears. At longer wavelengths this is no big
deal, but a system with a very long line that is trying to reproduce
higher frequencies there are going to be powerful comb-filtering
effects. This has nothing to do with driver design and everything to
do with the size of the driver in relation to the wavelengths handled.
The point of a line source extending from floor to ceiling is that it uses the
ceiling and floor reflections to eliminate the comb-filtering effects that (I
think) you're complaining about.
I do not know how this can happen. The line cannot be the same
distance from the ears over its full length. Because of this, when
some signals from certain parts of the line are hitting the ears
signals from other parts of the line are going to be hitting it
anywhere from slightly out of phase to 180 out. As the frequency
changes the null points will change, since although the speed of sound
remains constant the out-of-phase points will be different at
different frequencies.
At least at those frequencies where the
wavelength is shorter than the effective length of the line.
This is the opposite from the way it is. Where the wavelengths are
longer than the line the time between negative and positive sweeps is
so long that for all intents and purposes the line is in phase over
its entire length. However, at shorter wavelengths the the phase
characteristics gradually shift as you move away from any point of the
line that is not the closest to the ears. The gradual delay causes the
incoming signals to gradually go out of phase with those that are
coming from the closest point on the line. As you move further up or
down the line they go back into phase (but be 360 degrees out,
actually) and then further up or down they will start going back out
of phase.
Look at it this way. Let's say that you are hearing a 5 kHz signal
coming from the center of the line. As you move up and down the line
several inches the source is further from your ears and so that part
of the signal is delayed slightly in time and the signal is not fully
in phase with the signal coming from the closest part of the line. At
some point up and down the line that signal will be 180 out with the
part coming from the closest section, and so you would get a
substantial cancellation dip. At higher or lower frequencies the dip
would be elsewhere, and the series of cancellation dips makes the
at-ear response curve resemble a comb filter.
Uh -- yes, they do, if by accuracy you mean very low coloration. There is a
qualitative difference that does not show up in a simple frequency-response
curve.
What causes this? Certainly, standard HD and other distortion products
are sufficiently low with dynamic speakers. And many dynamic speakers
can go a LOT lower in frequency and with considerably less distortion
down really low. A subwoofer will correct this limitation with
flat-panel jobs, of course, and many have dynamic woofers, anyway.
Actually, my theory is that some listeners rather like the choppy
frequency response (which makes some frequencies stand out) and treble
rolloff. They also like the reduced reverberant-field strength and the
fact that most of what reverberant-field strength there is is
reflected from the front wall and therefore somewhat delayed. This is
a pleasant artifact that has nothing to do with the reproduction
issue.
This terrible fallacy needs to be put to death once and for all.
OK, are you saying that you do not hear all portions of the line as a
direct-field signal? If that is the case, the only part of the line
you hear is that directly opposite the ears, with all the other sound
from the line passing over your head or under it. I do not think you
will find this to be possible. Rather, you hear the whole line, but
because the signals coming from the various parts of it are delayed in
time in relation to the part of the line that is closest to the ears,
those phase-generated comb-filtering effects are very real.
A line source "longer" than the wavelengths being reproduced projects a
vertically planar wavefront. The flatness of wavefront is due to destructive
interference among the spherical components of the source (considering the
sources as made of gazillions of spherical point radiators).
If I am "missing" something here, please tell me what it its.
Well, if the line were made up of those gazillions of point radiators
and we suddenly and temporarily shut down all but the top and bottom
you would still hear the top and bottom playing. Turn the central
section back on and you would still here them, but this time, because
they are different distances from the ears, much of what they
reproduce would be phase shifted from what is coming from the nearer
parts of the line. The line cannot generate a wavefront that can be
heard simultaneously from its entire surface.
You're missing the point completely. In a planar wavefront, the parts of the
wavefront "above" and "below" the listener never reach his ears at all (until
they're reflected off the side and rear walls)! That's precisely the point.
Well, you are basically saying that segments above and below the
nearest part of the line have their signals cancelled out by those
central signals. This is true, but only at certain frequency points.
Hence the comb-filtering effect.
But -- DUH! DUH, DUH, DUH! That's true for ANY driver with a broad radiation
pattern. The "direct" output of a 2" dome midrange driver that doesn't reach the
listener strikes the ceiling, walls, and floor and combs with the direct sound
(if the sound lasts long enough).
But the direct-field signal does get there as a coherent wavefront.
Sure, reflected sound arrives later, but those reflected signals are
coming from all over and they also are continuously delayed over a
long period of time, relatively. With the line-source radiator, you
are getting a comb-filtered direct-field signal. Except at the
crossover points, you do not get that with conventional systems.
One of the biggest arguments in favor of a floor-to-ceiling line source (whether
made of planar or cone/dome drivers) is that it _minimizes_ these sorts of
effects. I think you have this exactly backwards.
Nope. The line cannot be the same distance from the ears over its
entire length. Sure, some of the output is nulled by cancellation
effects. However, some parts of the line are in phase with the nearest
segment (360 degrees, 720 degrees out, etc.) and those frequencies are
what make up the tops of the comb-filtering curve. The nulls make up
the bottoms of the dips. All points in between are the result of
partial cancellations. Lipshitz illustrated this clearly in his AES
paper.
You do at some frequencies. The line cannot cancel all the sound
coming from the segments above and below the nearest section. Only
some parts points can be cancelled. The result is the comb filter. The
dips and peaks only disappear where the wavelengths are longer than
the length of the line. This generally means the lower midrange. Above
that you get comb filtering. Lipshitz actually outlined this pretty
clearly.
WRONG, WRONG, WRONG. You don't understand what a planar source it. See above.
Right, right right. The cancellation effect you note does not happen
continuously and uniformly at all frequencies. It happens in a choppy
manner.
Now I understand the fallacy in Mr. Lipshitz's calculations.
Read his article. I gave a reference. Actually, the source of the
article was the result of a debate he had with Roy Allison in the
Boston Audio Society magazine "Speaker." Allison pointed out the
deficiencies with long, uncontrolled length line sources and Lipshitz
contested that point of view. He then went on to do some research and
his AES paper ended up supporting what Allison had said.
Not in the region where the source is longer than the wavelengths being
reproduced.
Backwards. The comb-filtering effects disappear at lower frequencies,
but become more emphatic as the frequency climbs.
He believes that because his reasoning -- and consequent mathematical model --
is wrong.
I find it interesting that you contest the findings of an audio icon
like Dr. Lipshitz, particularly since you have not read his paper.
Contact the AES for a copy of the lecture and demonstration he
presented.
Yes -- if you use invalid reasoning to create the mathematical model.
Get a copy of the Lipshitz paper (with John Vanderkooy): "Acoustic
Radiation of Line Sources of Finite Length," presented at the 81st AES
convention, November, 1986. Available from the AES as preprint 2417
(D-4).
Howard Ferstler
