Quad ESL2805 [was: And so...]
 

On Fri, 05 Aug 2011 09:43:55 +0100, Eiron
wrote:

On 04/08/2011 16:28, Jim Lesurf wrote:
In aweb.com, Rob
wrote:


I had a pair of 988s for a year or two and did find them pretty
astonishing. It's difficult for me to explain, but 'uncoloured',
'unresonant', 'precise', pin sharp' perhaps. Bass more in the 'tuneful'
than 'deep' camp, and I could never pair them with a REL sub I had
successfully.

I've managed to get a sub working OK with the 988s I use in the living room
'AV' system (just stereo, of course). But never found any happy combination
of a sub with the old 63s in the hifi room. I guess this is partly a matter
of different room acoustics, and partly a matter of different source
material.

In the main, two things led to selling them. I did find that I had to
sit still, pretty much on an axis, or else the tone/image would shift to
a much greater extent than I've found with box speakers. The sound
remained consistently good, I just found the effect tiresome.

I recognise what you mean. Even optimally set up I experience something
similar. I tend to regard them as the 'biggest pair of virtual headphones
in the world' in some ways. The best result means having your head in about
the right location to get your ears in the right places wrt the invisible
large headphones floating in space.

So the 'virtual point source' doesn't really work?


The virtual point source was an interesting claim, but technically
naiive. Had the speaker been a complete sphere, the virtual point
source at its centre would have been spot on.

With a speaker of limited size, the effect is the same as a true point
source in another room, and a speaker-sized hole in the intervening
wall. It is easy to see how things go wrong off-axis. In other words,
if you aren't perfectly lined up, the point source is simply not
visible through the hole, and you lose all the highs.

d

Quad ESL2805 [was: And so...]
 

In article , Don Pearce
wrote:
On Fri, 05 Aug 2011 09:43:55 +0100, Eiron
wrote:

On 04/08/2011 16:28, Jim Lesurf wrote:
In aweb.com, Rob
wrote:


I recognise what you mean. Even optimally set up I experience
something similar. I tend to regard them as the 'biggest pair of
virtual headphones in the world' in some ways. The best result means
having your head in about the right location to get your ears in the
right places wrt the invisible large headphones floating in space.

So the 'virtual point source' doesn't really work?


The virtual point source was an interesting claim, but technically
naiive. Had the speaker been a complete sphere, the virtual point source
at its centre would have been spot on.

IIRC Quad were said at one time to be working on a 'ball' ESL. But hardly
surprising that this didn't lead to a product. Not an easy thing to make,
even compared with the 63!

With a speaker of limited size, the effect is the same as a true point
source in another room, and a speaker-sized hole in the intervening
wall. It is easy to see how things go wrong off-axis. In other words, if
you aren't perfectly lined up, the point source is simply not visible
through the hole, and you lose all the highs.

Its a bit more complex than that for the ESL63 and its children,
essentially for two reasons.

Firstly, the phasefront is quite distinctly curved over much of the
freqency range. So the 'apparent point source' center is close to the
'hole'.

Secondly, the size of the hole also scales with the wavelength, and the
hole has graduated edges.

Plus, of course, the fields then exhibit what people useually think of as
diffraction spreading to a wider angle as they move away from the array.

The result isn't much like 'geometric ray' behaviour. Nor like a
rectangular aperture with a uniform (plane wave) field.

As a result you don't lose 'highs' as quickly as you might expect if
assuming behaviour like a rectangular hole the size of the speaker passing
sound from a distant pointlike source. And the change in behaviour as you
go off axis is much smoother.

FWIW I did once visit Quad and discuss with PJW having a go at modelling
the speaker radiation in terms of Gaussian Beam Mode analysis. This is a a
form of analysis for field propagation I and others used a lot for the beam
diffraction behavour of EM radiation and antennas, arrays, etc. However I
never got around to this as other applications of the method provided more
money at the time. ;-

Somewhere or other I've got some plots that PJW gave me showing the
dispersion versus frequency of the 63. This is perhaps rather more
revealing than the usual set of polar plots when considering the overall
behaviour. I'll see if I can dig a plot out and shove a copy onto the
website so people can see what the behaviour is like.

[I've just done this and the result is at
http://jcgl.orpheusweb.co.uk/temp/half.png
note this is a 160kb png.]

Slainte,

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc http://www.audiomisc.co.uk/index.html

Quad ESL2805 [was: And so...]
 

On 05/08/2011 13:08, Jim Lesurf wrote:

Who was it who claimed to use four stacked ESL63s either side of his
desk as the world's biggest pair of headphones?

Pass. ...No, I don't mean 'Nelson Pass'. I mean 'Mastermind Pass'. ;-

http://groups.google.com/group/uk.re...697cfd53d2dc55
if you want a laugh or some duff advice about ESL placement.

--
Eiron.

Quad ESL2805 [was: And so...]
 

On Fri, 05 Aug 2011 15:43:11 +0100, Jim Lesurf
wrote:

In article , Don Pearce
wrote:
On Fri, 05 Aug 2011 09:43:55 +0100, Eiron
wrote:

On 04/08/2011 16:28, Jim Lesurf wrote:
In aweb.com, Rob
wrote:


I recognise what you mean. Even optimally set up I experience
something similar. I tend to regard them as the 'biggest pair of
virtual headphones in the world' in some ways. The best result means
having your head in about the right location to get your ears in the
right places wrt the invisible large headphones floating in space.

So the 'virtual point source' doesn't really work?


The virtual point source was an interesting claim, but technically
naiive. Had the speaker been a complete sphere, the virtual point source
at its centre would have been spot on.

IIRC Quad were said at one time to be working on a 'ball' ESL. But hardly
surprising that this didn't lead to a product. Not an easy thing to make,
even compared with the 63!

With a speaker of limited size, the effect is the same as a true point
source in another room, and a speaker-sized hole in the intervening
wall. It is easy to see how things go wrong off-axis. In other words, if
you aren't perfectly lined up, the point source is simply not visible
through the hole, and you lose all the highs.

Its a bit more complex than that for the ESL63 and its children,
essentially for two reasons.

Firstly, the phasefront is quite distinctly curved over much of the
freqency range. So the 'apparent point source' center is close to the
'hole'.


Any idea how close they got it? I know they used annular panels and a
tapped delay line, but I've never found out how much delay they used.

Secondly, the size of the hole also scales with the wavelength, and the
hole has graduated edges.

That helps a lot.

Plus, of course, the fields then exhibit what people useually think of as
diffraction spreading to a wider angle as they move away from the array.


Well, that would be true of the hole too.

The result isn't much like 'geometric ray' behaviour. Nor like a
rectangular aperture with a uniform (plane wave) field.

As a result you don't lose 'highs' as quickly as you might expect if
assuming behaviour like a rectangular hole the size of the speaker passing
sound from a distant pointlike source. And the change in behaviour as you
go off axis is much smoother.


My own experience is that the highs roll off quite progressively - but
also rather quickly compared to a dome tweeter.

FWIW I did once visit Quad and discuss with PJW having a go at modelling
the speaker radiation in terms of Gaussian Beam Mode analysis. This is a a
form of analysis for field propagation I and others used a lot for the beam
diffraction behavour of EM radiation and antennas, arrays, etc. However I
never got around to this as other applications of the method provided more
money at the time. ;-


These days it would be done with a 3-D array of points covering the
entire field together with finite element analysis. You can achieve an
arbitrary degree of accuracy, but you do need a lot of computing
power.

Somewhere or other I've got some plots that PJW gave me showing the
dispersion versus frequency of the 63. This is perhaps rather more
revealing than the usual set of polar plots when considering the overall
behaviour. I'll see if I can dig a plot out and shove a copy onto the
website so people can see what the behaviour is like.

[I've just done this and the result is at
http://jcgl.orpheusweb.co.uk/temp/half.png
note this is a 160kb png.]


I don't think there are too many of us left who would be daunted by a
160kb png ;-)

Kind of difficult to interpret though. Can you shed some light on what
we are looking at?

d

Quad ESL2805 [was: And so...]
 

On Fri, 05 Aug 2011 16:31:48 +0100, Eiron
wrote:

On 05/08/2011 13:08, Jim Lesurf wrote:

Who was it who claimed to use four stacked ESL63s either side of his
desk as the world's biggest pair of headphones?

Pass. ...No, I don't mean 'Nelson Pass'. I mean 'Mastermind Pass'. ;-

http://groups.google.com/group/uk.re...697cfd53d2dc55
if you want a laugh or some duff advice about ESL placement.

******* - you just made me see something written by Jute. I clicked
off it quickly enough to cause no harm, luckily.

d

Quad ESL2805 [was: And so...]
 

In article , Don Pearce
wrote:
On Fri, 05 Aug 2011 15:43:11 +0100, Jim Lesurf
wrote:


Firstly, the phasefront is quite distinctly curved over much of the
freqency range. So the 'apparent point source' center is close to the
'hole'.


Any idea how close they got it? I know they used annular panels and a
tapped delay line, but I've never found out how much delay they used.

I'd need to go back and reexamine the data. This was something I looked at
*decades* ago, so have forgotten all the real details I'm afraid!


FWIW I did once visit Quad and discuss with PJW having a go at
modelling the speaker radiation in terms of Gaussian Beam Mode
analysis. This is a a form of analysis for field propagation I and
others used a lot for the beam diffraction behavour of EM radiation and
antennas, arrays, etc. However I never got around to this as other
applications of the method provided more money at the time. ;-


These days it would be done with a 3-D array of points covering the
entire field together with finite element analysis. You can achieve an
arbitrary degree of accuracy, but you do need a lot of computing power.

Yes. However the advantage of GBM is that for beam-like situations you
actually get results much more easily, quickly, and more accurate for many
cases. In effect one computation specifies the entire 3D complex field
pattern along and across the beam.

Somewhere or other I've got some plots that PJW gave me showing the
dispersion versus frequency of the 63. This is perhaps rather more
revealing than the usual set of polar plots when considering the
overall behaviour. I'll see if I can dig a plot out and shove a copy
onto the website so people can see what the behaviour is like.

[I've just done this and the result is at
http://jcgl.orpheusweb.co.uk/temp/half.png note this is a 160kb png.]


I don't think there are too many of us left who would be daunted by a
160kb png ;-)

Kind of difficult to interpret though. Can you shed some light on what
we are looking at?

Consider the top graph as an example. The horizontal axis is fairly
obviously frequency. The vertical is the angle between the listening
(measurement) location and the speaker axis. i.e. 'how far off axis' you
are.

The lines are contours of sound level wrt the on axis sound level. So for
example the line with '-3' at the lefthand end is for where the off axis
level is 3dB down on the level on boresight axis at that frequency.

At 315 Hz the -3dB lines (note they are paired either side of the axis) are
about 35 deg off axis. i.e. at 315 Hz the speaker's radiation patten has a
3dB half-width angle of about 35 degrees.

At 5kHz the -3dB lines are about 30 degrees off axis. So at 5kHz the
halfwidth is 30 degrees.

This indicates that over a frequency range from a few hundred Hz up to
around 5kHz the angular width of the radiation pattern (to the -3dB points)
stays remarkably similar at about +/- 30 degrees. The change in beamwidth
is only about 10 percent for more than a factor of 10 change in frequency.

In effect at each frequency a vertical 'slice' though the plot shows you
the shape of the radiation pattern. It is rather more uniform with
frequency than a a large hole with a plane field, and also rather more
uniform that many conventional speakers. TBH I doubt many speaker makers
would like people to see their equivalent results! :-)

This means that one feature of the 63 design is that the radiation off at
angles (which you hear later via transverse room reflections) is tonally
more similar to the direct sound near axis than is the case for many
conventional speakers. In particular, no 'jump' in the axial/surround ratio
at the crossover point between a woofer and a tweeter.

The behaviour breaks down about 5kHz because you are now just using the
central element of the array (the center disc) and so the angles then fall
with frequency in the usual way.

Slainte,

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc http://www.audiomisc.co.uk/index.html

Quad ESL2805 [was: And so...]
 

On Fri, 05 Aug 2011 17:36:53 +0100, Jim Lesurf
wrote:

In article , Don Pearce
wrote:
On Fri, 05 Aug 2011 15:43:11 +0100, Jim Lesurf
wrote:


Firstly, the phasefront is quite distinctly curved over much of the
freqency range. So the 'apparent point source' center is close to the
'hole'.


Any idea how close they got it? I know they used annular panels and a
tapped delay line, but I've never found out how much delay they used.

I'd need to go back and reexamine the data. This was something I looked at
*decades* ago, so have forgotten all the real details I'm afraid!


FWIW I did once visit Quad and discuss with PJW having a go at
modelling the speaker radiation in terms of Gaussian Beam Mode
analysis. This is a a form of analysis for field propagation I and
others used a lot for the beam diffraction behavour of EM radiation and
antennas, arrays, etc. However I never got around to this as other
applications of the method provided more money at the time. ;-


These days it would be done with a 3-D array of points covering the
entire field together with finite element analysis. You can achieve an
arbitrary degree of accuracy, but you do need a lot of computing power.

Yes. However the advantage of GBM is that for beam-like situations you
actually get results much more easily, quickly, and more accurate for many
cases. In effect one computation specifies the entire 3D complex field
pattern along and across the beam.

Somewhere or other I've got some plots that PJW gave me showing the
dispersion versus frequency of the 63. This is perhaps rather more
revealing than the usual set of polar plots when considering the
overall behaviour. I'll see if I can dig a plot out and shove a copy
onto the website so people can see what the behaviour is like.

[I've just done this and the result is at
http://jcgl.orpheusweb.co.uk/temp/half.png note this is a 160kb png.]


I don't think there are too many of us left who would be daunted by a
160kb png ;-)

Kind of difficult to interpret though. Can you shed some light on what
we are looking at?

Consider the top graph as an example. The horizontal axis is fairly
obviously frequency. The vertical is the angle between the listening
(measurement) location and the speaker axis. i.e. 'how far off axis' you
are.

The lines are contours of sound level wrt the on axis sound level. So for
example the line with '-3' at the lefthand end is for where the off axis
level is 3dB down on the level on boresight axis at that frequency.

At 315 Hz the -3dB lines (note they are paired either side of the axis) are
about 35 deg off axis. i.e. at 315 Hz the speaker's radiation patten has a
3dB half-width angle of about 35 degrees.

At 5kHz the -3dB lines are about 30 degrees off axis. So at 5kHz the
halfwidth is 30 degrees.

This indicates that over a frequency range from a few hundred Hz up to
around 5kHz the angular width of the radiation pattern (to the -3dB points)
stays remarkably similar at about +/- 30 degrees. The change in beamwidth
is only about 10 percent for more than a factor of 10 change in frequency.

In effect at each frequency a vertical 'slice' though the plot shows you
the shape of the radiation pattern. It is rather more uniform with
frequency than a a large hole with a plane field, and also rather more
uniform that many conventional speakers. TBH I doubt many speaker makers
would like people to see their equivalent results! :-)

This means that one feature of the 63 design is that the radiation off at
angles (which you hear later via transverse room reflections) is tonally
more similar to the direct sound near axis than is the case for many
conventional speakers. In particular, no 'jump' in the axial/surround ratio
at the crossover point between a woofer and a tweeter.

The behaviour breaks down about 5kHz because you are now just using the
central element of the array (the center disc) and so the angles then fall
with frequency in the usual way.

Slainte,

Jim

OK, got that. Why didn't they think to use a log horizontal scale
rather than that muddle?

It is the bit over 5kHz that is the problem though. That centre panel
is a fair bit bigger than your average tweeter and will get rather
beamy.

d

Quad ESL2805 [was: And so...]
 

On 05/08/2011 17:36, Jim Lesurf wrote:

The behaviour breaks down about 5kHz because you are now just using the
central element of the array (the center disc) and so the angles then fall
with frequency in the usual way.

I thought the annuli were fed from a delay line rather than a crossover....

--
Eiron.

Quad ESL2805 [was: And so...]
 

On Fri, 05 Aug 2011 21:25:11 +0100, Eiron
wrote:

On 05/08/2011 17:36, Jim Lesurf wrote:

The behaviour breaks down about 5kHz because you are now just using the
central element of the array (the center disc) and so the angles then fall
with frequency in the usual way.

I thought the annuli were fed from a delay line rather than a crossover....

They are. Jim is referring to the fact that the central portion
doesn't get any smaller with rising frequency, and at 5kHz its size is
starting to become appreciable.

d

Quad ESL2805 [was: And so...]
 

In article , Don Pearce
wrote:
On Fri, 05 Aug 2011 21:25:11 +0100, Eiron
wrote:

On 05/08/2011 17:36, Jim Lesurf wrote:

The behaviour breaks down about 5kHz because you are now just using
the central element of the array (the center disc) and so the angles
then fall with frequency in the usual way.

I thought the annuli were fed from a delay line rather than a
crossover....

They are.

Yes. But this line does roll away the HF to the outer rings as well as
introduce delays. This is partly because some power has already been
radiated by inner elements of the array. Also because each coil in the
delay line has an extra 'shorted turn' to act as a resistance. This is a
clever idea as it also cancels the dispersion that otherwise can arise with
loaded lines. The result is to reduce frequency dependence of the time
delays between elements and stabilising the apparent phase center ( 'point
source') Those who understand transmission lines will recognise this as an
interesting application of the Heavyside Criterion which was initially
devised to get transatlantic cables to work correctly back in Victorian
days! :-)

Jim is referring to the fact that the central portion doesn't get any
smaller with rising frequency, and at 5kHz its size is starting to
become appreciable.

Yes.

In principle PJW could have added one or more smaller rings to the center
and ended with a smaller center disc. Then had few more stages on the
transmission line. However I guess he didn't do this for two reasons.

Firstly, my impression from listening is that the beaming above 5k has
little real impact on imaging. Perhaps because the reality is that there
often isn't much up there or it doesn't have much effect on perception once
the region below 5k work well.

Secondly, there is some spreading, partly due to fringe fields, partly due
to the diaphragm material. So the smallest 'spot' you can actually drive
'independently' is limited.

Also, I always find that the 63 design works best when the speakers are
angled to cross in front of me, below my ear height. This makes the image
more stable and less affected by small head movements. I guess this is for
the same reason people use a similar alignment for conventional speakers.
It allows a trade off in the change in time difference and signal amplitude
at the ears at HF so that the effects tend to cancel when you move your
head sideways a little.

So overall, I think PJW got this about right.

But of course, this will depend on the room. listener, and choice of music.
My taste is probably like PJWs. The tradiational 'BBC' sound for classical
music, etc. Indeed, I still prefer Radio 3 to many commercial recordings,
and prefer the sound on the BBC Music Mag discs as well!

Personally, my only hesitations tend to be the way the 63 design exposes
things like 'grainy' or 'scratchy' massed violin playing/recordings and the
lack of heavy bass that may suit some music.

FWIW I have been tentatively feeling that the new 2805s do have a 'drier'
bass than my old 63s. But can't do any direct comparisons, so am not sure
if this is my imagination as yet. So I may start twiddling with
positioning, etc, and maybe even sometime have another try at using a sub
to find out if that can now make an improvement rather than just adding
mud!

Slainte,

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc http://www.audiomisc.co.uk/index.html