Mike Coatham wrote:

"John Smith" wrote in message
...
[This reply got trapped in some e-mail limbo but has re-emerged so I'm
letting it be sent now in the hope someone remembers what were discussing]

Jim, Mike,

I bought the meter from Maplins about 3 years ago. I don't remember the
make
and I'm not at home to check. Maplins had it at £50 or £60 but claimed the
full price was around £100 and when I looked around the net it was on sale
elsewhere for around £100 so I guess it's accurate enough for my needs.

I set it up 12 inches from the centre of the speaker mounted on a camera
tripod (the meter has a tripod mount). Each speaker was measured
separately
so 4 runs of the test. I set the volume control as I would for normal
listening. The test CD (EMI) has a series of tracks playing a fixed
frequency for about 10 seconds (20Hz-20kHz). The volume coding is the same
for each track (15dBA below peak I think). For each frequency I noted the
meter reading (its a digital readout). While the TDL readings stayed about
the same from 8-15kHz the Quads dropped over 12dBA and by 20kHz they were
over 15 dBA less than the TDLs. I can't hear much above 10kHz these days
but
the meter had no problem and I knew what frequency was playing by the
track
number on the CD player.

Does this help or do you need the actual meter readings?

Thanks for the interest.

John.

"Jim Lesurf" wrote in message
...
In article , Mike Coatham

wrote:

"John Smith" wrote in message
...
Hi,

I compared my 1974 pair of ESL57s against a pair of newish TDL
speakers using a test CD with stepped tones and measured the output
with a sound level meter. The high-frequency roll-off of the ESL57s
was very noticable. The higher tones (10kHz up) were there but the
volume dropped off rapidly.
[snip]

Actually you may not have a problem with the ESL's per se but with the
listening position. The treble panels in the ESL 'beams' the sound to
one specific spot.

John: Could you tell us more about the microphone and speaker locations,
etc? Also could you give the details of what sound levels you measured
and
at what frequencies? My initial reaction is was the same as Mike's, but
it
would help us to diagnose/advise if we have more detailed/specific info.

Slainte,

Jim

Hi John,

Your meter was way too close to the Quads to get any meaningful ( accurate)
readings. Trouble here is you could be trying to fix something that isn't
broke by assuming the data you have collected is actually correct.
The Quad ESL, as I mentioned earlier certainly beams the HF content, so you
need to be far enough back to get accurate measurements. Unfortunately, I
have no idea what the original test parameters were for the ESL, but you
can bet your boots the measuring mic wasn't 12 inches from the centre .
What I suggest you do is set up the meter at your normal listing position,
and have someone tilt the ESL backwards and forwards whilst playing some HF
content. Note the meter change as the ELS is moved - and set them up at the
position where the HF content is at a peak. Then measure the whole spectrum
and see what results you get.

The method of testing ESL57 described by John Smith is not how anyone should
test any speaker.

The best way to test in an average lounge room is with a pink noise signal and a
calbrated microphone
set up in 4 different random positions around and near the listening position,
ie, say either side of the chair and at
two different heights, so mic position is say 3metres away from the speaker and
on axis.
One needs a pink noise source that is a true source of flat pink noise,
cacalibrated mic with stand, wide band amplifer,
switchable bandpass filter, peak detector and logarithmic volt meter calibrated
in +/- 20 dB from its resting centre position. A digital VM is useless. The
needle movement of the meter should be damped to give easy viewing of the
level being measured which moves around the centre point.

A switchable bandpass filter with constant Q of 12 is used to filter out 34
frequencies between 20Hz and 20kHz
from the amplified microphone signal. Testing should be done when things are
quiet.
The levels are plotted in dB on a suitably drawn up page with logathithmic
scales for F and level
The readings for each 4 positions are summed in Db and then averaged, so
readings of
say +3, +1, -4, -2 in dBV will give an average reading of the sum of those
number divided by 4, = -2 / 4 = -0.5dBV

PC programs used with a sound card will do this better and faster but I built my
own analog gear
before I got a PC in 2000. The graphed response can be printed out. I am not
aware of programs available
which allow four responses to be automatically be summed.
But the old fashioned method I use allows all sorts of variations to crossover
components and response deviations
to be checked. I find that when I have a response that is flat within +/- 2dB at
the chair some 3 metres from each speaker I am testing, then the response is
about as good as can ever be had because to get things flatter the crossover has
to become way too complex, and there are limitations to what one can achieve
with L,C & R.
I say this after testing a large mumber of speakers including those I built.
Many so called flat response commercial speakers are anything but flat often
with a presence peak between 2kHz and 9kHz, and bass peak around 80Hz to give
them more wham and bam in the showroom, ie, to make them sell.
But the best speakers really do have a flat response and are the least tiring
and most revealing to listen to
regardless of whether ESL or not, and are OK with any type of music from Heavy
Metal to Mozart.

But using fixed sine waves and a sound level meter that is not guranteed to give
a flat response
with a flat source of sound is a completely useless way to measure any speaker
anywhere.
Sine waves will be cancelled or reinforced due to room resonances and the graph
of response
will be +/- 12 dB with what look like random peaks and troughs, maybe there are
20
peaks and troughs along such a graph.
Pink noise is noise containing all frequencies at all times with varying
amplitude and phase
so the effect of room resonances is minimised.

Patrick Turner.