[net.audio] Reply to Herb Chong

ark@alice.UUCP (Andrew Koenig) (04/23/85)

Herb Chong made the following claim in a prior article:

> The BOSE tends to be very forgiving of suboptimal sources, and might
> lack the resolution to differentiate between the players.

I asked him for evidence to support his claim.  He responded by giving
me arguments (not evidence) to support a completely different claim --
namely, that he thinks Bose 901's have significant design problems.
Here's what he says, and here are my comments on his remarks.

> for starters, by its very design, the Bose 901's are nonlinear phase
> transducers.

I'm not sure what this statement means.  Is this somehow supposed
to summarize what follows?

> the primary sound is from the rear speakers which reflect
> off of the walls and back into the listening area.  great cancellations
> occur between the back sound wave and the one from the front speaker.

This is true of any loudspeaker in any room with partially reflecting
walls, floor, or ceiling -- in other words, any room but an anechoic
chamber.  Believe me, having an anechoic chamber for a living room
would be very uncomfortable indeed!

It is far from obvious to me that these problems should be any more of
a problem with 901's than anything else.

> there are 8 drivers in the back.  even if each were an ideal point
> source, the combined is anything but, and after reflecting off of
> uncontrolled wall surfaces with absorption curves that can be wildly
> different between setups, flat freqency response can't be guaranteed
> even with equalization.

So far, you haven't said anything that isn't true of any loudspeaker.

> which brings up another point, the supplied
> equalizer.  the drivers in the 901's are 5 inches nominal diameter.

I think they're four inches.

> equalization must be supplied to bring up the low end and the high end
> to get flat response.  this also means that the 5" drivers are being used
> well above 10K.  yes, plenty of boost will even allow a 10" woofer
> to play a 15K tone decently, but how much power do you have to supply
> when the natural roll-off of the driver is 6dB/octave above 5K?

Answer: about 10 dB.  No problem so far.  Especially when you realize
that almost none of the energy in music is above 5K.

> so you put
> high power voice-coils in, as Bose did, which increases mass, which 
> decreases the natural roll-off frequency.

No, you ignore the problem, because it isn't a problem.

> fortunately, he uses a sophisticated
> ducting system to achieve decent low frequency response without equalization,
> but then massive phase shifts are produced, and the air from the ducts
> are such high velocity that you can sometimes hear them whistling (depends
> upon how much other frequencies are being reproduced at the same time).

And here I thought I had birds in the yard! :-)

The surface area of nine 4-inch drivers is the same as the surface
area of one 12-inch driver.  Nine 4-inch drivers in a carefully-designed
ported enclosure should therefore have about the same bass problems (or
lack thereof) as a 12-inch woofer in a bass-reflex cabinet, something
that we see all the time.

> the direct/reflecting principle is very interesting and has some theoretical
> advantages, but they are hardly ever realized in practice.

Perhaps the biggest advantage is that it is much easier to design
an active equalizer that goes ahead of the power amp than it is
to design a passive crossover network that comes after the power amp.
Yes, it is possible to use an active crossover network, with the
attendant requirement for multiple power amplifiers, but that gets
much more expensive.

> the equalizer
> removes some of the variables, but using electrostatic elements and
> properly designed conventional woofer would have been better from the
> point of view of reduced driver mass and little or no equalization
> in the high frequencies.

Evidence, please.

> of course, electrostatic elements have a whole
> set of problems of their own, but they are more appropriate for another
> discussion.  the type of drivers that Omar Bose should have used given
> his design requirements were horrendously expensive back when he started
> and notoriously unreliable.  quads were about the only option.

Every loudspeaker design has its problems.  If there were a
design with no problems, no one would build anything else!
Therefore, even if your comments indicate serious problems
with the Bose 901 design (which they do not), this information
is only relevant if:

	1. the problems actually affect the ability of
	   people to hear differences between other
	   components when using these speakers, and

	2. other speakers do not generally have other
	   problems that similarly prevent people from
	   hearing differences between other components.

By the way, you spelled Dr. Bose's given name wrong.

herbie@watdcsu.UUCP (Herb Chong [DCS]) (04/24/85)

once again, i am forced to defend a position that is self evident to
anyone who has ever done any thinking about speaker design theory.  as
it happens, i have done a lot more than thinking.  i have had the
opportunity to work with Drs. Lipschitz and Vanderkooy (see this month's
J Audio Eng. Soc.) and many others such as Paul Barton (speaker design
engineer, Luxman).  i don't make claims based upon unsupported
evidence.  however, i also don't keep my journals at work and don't
feel that quoting thousands of lines from a journal is neccesarily
going to clear the air.

+Herb Chong made the following claim in a prior article:
+
+> The BOSE tends to be very forgiving of suboptimal sources, and might
+> lack the resolution to differentiate between the players.
+
+I asked him for evidence to support his claim.  He responded by giving
+me arguments (not evidence) to support a completely different claim --
+namely, that he thinks Bose 901's have significant design problems.
+Here's what he says, and here are my comments on his remarks.

design is the key to any problem.  if i wanted to measure the length of
a piece of metal, i don't use a rubber band with markings on it.  if a
design inherently makes it impossible to measure something, then that
design is useless for distinguishing that something.  another example:
suppose i wanted to measure a micro-accelerometer which produces
voltage in the range 0 to 1 mV.  if the noise in my system is 10 mV,
then there's no way i am going to get accurate, reliable, and most
important, repeatable, values for the output voltage which is less than
the rated noise of my design.  i have to use a better design with lower
noise to resolve what i am interested in.

+> for starters, by its very design, the Bose 901's are nonlinear phase
+> transducers.
+
+I'm not sure what this statement means.  Is this somehow supposed
+to summarize what follows?

it means that if i stick a square wave in, i don't get one out, not
even a bandlimited one.  it means that the phase shift is not a linear
function of frequency.  it means that the effects of the phase shifts
that are being sought are going to be lost in the massive phase shifts
already in the design.  any difference between the two players is 
going to be primarily in the phase differences at high frequencies with
minor fluctuations in frequency response less in magnitude, although
perhaps more audible.

+> the primary sound is from the rear speakers which reflect
+> off of the walls and back into the listening area.  great cancellations
+> occur between the back sound wave and the one from the front speaker.
+
+This is true of any loudspeaker in any room with partially reflecting
+walls, floor, or ceiling -- in other words, any room but an anechoic
+chamber.  Believe me, having an anechoic chamber for a living room
+would be very uncomfortable indeed!

you missed the point.  the two sets of drivers are reproducing the same
frequencies at about the same amplitude only centimeters apart.  the
cancellation pattern is horrendous and superimposed upon that are the
cancellations due to interference from the reflected sound wave.  in a
normal speaker system, there is no cancellation due to a nearby driver
(30 to 50 cm away), only the reflected sound from nearby walls.  you
may claim that there is due to diffraction effects of the corners of
all speaker cabinets, but i submit that these are neglible in
comparison to having a driver only a short distance away.  an ordinary
speaker can have the cancellations from reflected sound moved to a
nonobjectionable frequency by speaker placement.  when the drivers are
part of the same cabinet, you can't change the spacing.

+> there are 8 drivers in the back.  even if each were an ideal point
+> source, the combined is anything but, and after reflecting off of
+> uncontrolled wall surfaces with absorption curves that can be wildly
+> different between setups, flat freqency response can't be guaranteed
+> even with equalization.
+
+So far, you haven't said anything that isn't true of any loudspeaker.

most loudspeakers do not depend upon reflected sound from walls as the
primary sound source.  they want to minimize that as much as possible
and use only direct sound, which is much more easily controllable and
suffers less from interference effects of all those reflecting sound
waves from irregularly shaped objects of different absorption
characteristics.

+> which brings up another point, the supplied
+> equalizer.  the drivers in the 901's are 5 inches nominal diameter.
+
+I think they're four inches.

if you look in this month's stereo review, 5" nominal.  effective 4.5".

+> equalization must be supplied to bring up the low end and the high end
+> to get flat response.  this also means that the 5" drivers are being used
+> well above 10K.  yes, plenty of boost will even allow a 10" woofer
+> to play a 15K tone decently, but how much power do you have to supply
+> when the natural roll-off of the driver is 6dB/octave above 5K?
+
+Answer: about 10 dB.  No problem so far.  Especially when you realize
+that almost none of the energy in music is above 5K.

10 dB is still ten times the power.  some amplifiers don't take kindly
to near full power for today's highly electronic music with lots of
high frequencies.  more importantly, it puts the high frequencies closer
to the clipping point of the amplifier.  many amps don't take kindly to
overload at high frequencies, sometimes going into oscillation because 
of the very high frequency (>100kHz) generated by the clipping high
frequencies.  this also means very high power too.

+> so you put
+> high power voice-coils in, as Bose did, which increases mass, which 
+> decreases the natural roll-off frequency.
+
+No, you ignore the problem, because it isn't a problem.

you choose to bury your head in the sand.  decreasing the natural roll-off
frequency means more boosting at the high end in order to achieve flat
response.  since this is done by frequency dependent amplification
(read active filters), you are inducing phase shifts on top of whatever
was in the system anyway.  if you do it right, it cancels the phase shift
due to the driver configuration.  Bose makes no claim about phase
accuracy and given the design, can't.  the boosting also further reduces
the high frequency headroom of the amplifier.

+> fortunately, he uses a sophisticated
+> ducting system to achieve decent low frequency response without equalization,
+> but then massive phase shifts are produced, and the air from the ducts
+> are such high velocity that you can sometimes hear them whistling (depends
+> upon how much other frequencies are being reproduced at the same time).
+
+And here I thought I had birds in the yard! :-)

the measured air velocity from those ducts can reach the range of 30 to 50
kmh.  a technical paper by Bose states that the toughest part of the
design of the 901's was to have a small duct for high air velocity
without audible effects due to air friction and unwanted resonances
caused by them.  the ducts are only about 3cm in diameter.  moving air
that fast through that small an area is tricky if it's not supposed to
make any noise of it's own.  try playing some really well recorded
organ music on a pair of 901 IV's like i did and you'll see what i
mean.  

+The surface area of nine 4-inch drivers is the same as the surface 
+area of one 12-inch driver.  Nine 4-inch drivers in a carefully-designed 
+ported enclosure should therefore have about the same bass problems (or 
+lack thereof) as a 12-inch woofer in a bass-reflex cabinet, something 
+that we see all the time.  

Bose uses two ducts.  this means smaller cross-sectional area for
each.  this also means higher probability of a duct resonance at a
very audible (250 Hz to 5 kHz) range.  most ducts have large areas
and are much shorter than the Bose ducts.  air velocity is lower and
duct resonance frequency is very high in comparison.  Bose uses a long
throw design voice-coil.  more air is moved for greater output at low
frequencies.  this also increases the air velocity compared to a short
throw design.  also, you obviously haven't heard rushing sound from a
bass reflex system due to air in the port.  organ music tends to induce
this.  so do warped records.

+> the direct/reflecting principle is very interesting and has some theoretical 
+> advantages, but they are hardly ever realized in practice.  
+ 
+Perhaps the biggest advantage is that it is much easier to design 
+an active equalizer that goes ahead of the power amp than it is 
+to design a passive crossover network that comes after the power amp.  
+Yes, it is possible to use an active crossover network, with the
+attendant requirement for multiple power amplifiers, but that gets
+much more expensive.  

how much active equalizer network design experience do you have? i
don't have a lot, but i also know it's a lot trickier than it looks.
once you have that active equalizer in the pre-power interconnection,
you have to use all 901's or all other's.  it is a trickier step to do
a comparison of different speaker systems, but then that wasn't what
you were doing.  

have you looked inside a Bose equalizer?  i have.  i wouldn't use most
of those parts for a cheap transistor radio.  they happen to be in a
key connection in the amplifier, between the pre-amp and power-amp
(assume ideal configuration).  that sucker is noisy.  i maintain
greater than 85 dB SNR from ALL my inputs to speaker out relative to
rated output.  this is unweighted.  the only exception is my MC head
amp, which is 80 dB.

+> the equalizer 
+> removes some of the variables, but using electrostatic elements and 
+> properly designed conventional woofer would have been better from the 
+> point of view of reduced driver mass and little or no equalization 
+> in the high frequencies.  
+ 
+Evidence, please.  

simple.  an elctrostatic driver is inherently a linear phase driver and
has a flat frequency response well into the 50kHz range.  roll-of at low
frequencies is directly proportional to driver area.  the only problem
is the load presented by a typical electrostatic driver is one a
transistor amplifier is not designed to handle.  the reactive component
of the impedance is too high.  a typical electrostatic element the size
of the back of the a 901 would weigh about 25 gm.  also remember that i
stated restrictions on the substitution.

+> of course, electrostatic elements have a whole 
+> set of problems of their own, but they are more appropriate for another 
+> discussion.  the type of drivers that Omar Bose should have used given 
+> his design requirements were horrendously expensive back when he started 
+> and notoriously unreliable.  quads were about the only option.  

+Every loudspeaker design has its problems.  If there were a 
+design with no problems, no one would build anything else!  

901's have created more problems than they solved.

+Therefore, even if your comments indicate serious problems 
+with the Bose 901 design (which they do not), this information 
+is only relevant if:  
+ 
+       1. the problems actually affect the ability of 
+          people to hear differences between other 
+          components when using these speakers, and 

linear response in all aspects is required.  it is not to difficult to
show, as i have done, that 901's are inherently nonlinear phase and
amplitude.

+       2. other speakers do not generally have other 
+          problems that similarly prevent people from 
+          hearing differences between other components.  

there are many linear phase and amplitude speakers on the market.  the
closest to ideal in this respect are the bipolar radiating speakers
like Magneplanars, Accoustat, and others.  you haven't compared with
these systems obviously or you would not be so lavish in your praise of
901's.  the 901's aren't bad sounding speakers, but they do not have
the resolution (capability to differentiate between subtle differences)
to do the job you set them out to do.  many conventional speaker
systems such as the B&W 801F's, KEF 104.2's, Allison 2's, and too many
others to name perform better than the 901's in that regard.

i have spent quite a bit of time in the last few years looking at
speaker systems in general in the $3000 CAN/pr range because i'm
looking to upgrade my B&W DM7Mk2's ($1700/pr CAN).  901's are $1800
CAN/pr with equalizer.  at no time before I bought my 7's or afterwards
have i ever considered the 901's up to my standards.  imageing was
nonexistent and detail in upper midrange particularly lacking.  it is
not possible to distinguish details using the 901's that i can with my
7's, and i know that there are better speakers than mine.  i personally
doubt that with my 7's i could hear a difference between the two ways
of converting the CD digital signal.  i am certain that with the 901's,
i could not.

Herb Chong...
Student Member, Audio Engineering Society
BASc in Systems Design Engineering

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, EARN: herbie@watdcs, herbie@watdcsu

dca@edison.UUCP (David C. Albrecht) (04/25/85)

> Herb Chong made the following claim in a prior article:
> 
> > The BOSE tends to be very forgiving of suboptimal sources, and might
> > lack the resolution to differentiate between the players.
> 
> I asked him for evidence to support his claim.  He responded by giving
> me arguments (not evidence) to support a completely different claim --
No, what we have here is a failure to communicate since, like religion
open mindedness to criticism of ones favorite piece of stereo is like
attacking ones faith.  People who cannot reason well enough to follow
a simple and easily provable argument get no respect in my book, I
certainly don't feel compelled to get numbers scribbled on a sheet
of paper to give to this person and how do you measure imaging anyway?
> namely, that he thinks Bose 901's have significant design problems.
> Here's what he says, and here are my comments on his remarks.
> 
> > for starters, by its very design, the Bose 901's are nonlinear phase
> > transducers.
> 
> I'm not sure what this statement means.  Is this somehow supposed
> to summarize what follows?
> 
I assume he means that using equalization to produce a (semi)linear
reponse from a speaker as in the 901s intrinsically produces phase
shifts in the affected parts of the frequency spectra.  Some discussion
has been made to the effect that this hurts imaging but this is definitely
not conclusive.
> > the primary sound is from the rear speakers which reflect
> > off of the walls and back into the listening area.  great cancellations
> > occur between the back sound wave and the one from the front speaker.
> 
> This is true of any loudspeaker in any room with partially reflecting
> walls, floor, or ceiling -- in other words, any room but an anechoic
> chamber.  Believe me, having an anechoic chamber for a living room
> would be very uncomfortable indeed!
> 
> It is far from obvious to me that these problems should be any more of
> a problem with 901's than anything else.
Horse pucky, a major portion of the high frequency energy of the 901's
comes off the walls, an artificially induced condition and one certainly
not true of most loudspeakers where most reflective energy is considerably
reduced in amplitude.
> 
> > there are 8 drivers in the back.  even if each were an ideal point
> > source, the combined is anything but, and after reflecting off of
> > uncontrolled wall surfaces with absorption curves that can be wildly
> > different between setups, flat freqency response can't be guaranteed
> > even with equalization.
> 
> So far, you haven't said anything that isn't true of any loudspeaker.
> 
Both parties are to some degree right but, certainly the 901's are much
worse in this respect because of their degree of reflected energy.

The major points that I have seen about the 901's

  Overdriving drivers to produce low and high frequencies produces a
  great deal of IM distortion.

  Splattering the high frequencies all over the rear wall of the
  room while giving a dispersed sound stage will certainly not
  give any ability to localize sources (i.e. the imaging sucks,
  in my opinion the Polk SDA, Carver sonic hologram, etc. approach
  is much better).  This is an OBVIOUS point to any but the most
  brain damaged.  This is why everyone is hopping on the person
  that submitted the CD review (personally I enjoyed the review
  but some people have no humor).

  They are more sensitive to placement and room furnishing due
  to their reflective nature (again OBVIOUS).

  Frequency dependent phase shifts are induced by the equalization
  network which is required to get semi-linear frequency response.
  This is possibly not an important point but it is a point.

Now this is not to say it isn't a perfectly fine speaker if you
like the spread sound stage, don't care about the lack of localization
of images, and like the way it sounds in your room.

I certainly agree that it is not a suitable speaker for monitoring since
imaging is certainly a factor in such evaluation.

David Albrecht (My opinions have always been and always will be, my own)
General Electric