[Patrick Turner]

For those left wondering what the heck the facts are about sensitivity,
here they are.

But first I must describe the conditions of the test.

A reliable source of pink noise was used.

To ascertain what signal voltages are involved,
an osciliscope was set up so that 10Vrms of a 1kHz sine wave from 300
watt SS amp
registered 1/4 of the screen vertical room, ie, 40Vrms will just fill
the screen completely.
This is 56 peak ac volts.

Pink noise was fed to the amp and adjusted for level until
the peaks in the noise were equal to the height of the screen, ie, the
maximum
peak signals were 56Vpk.

No amplifier clipping was visible.

The maximum Vrms within the noise could be up to 40Vrms.

Using a variety of DMM meters in their Vac ranges all gave 12.1Vac when
in fact there
was a maximum of 40Vrms present.
So the ratio of maximum Vrms within the noise was 40 / 12.1 = 3.3:1

The pink noise was adjusted down to measure 0.85 Vac on my Fluke, ( 2.80
Vrms max )

This was fed to the ER Audio speaker which give a very quiet output and
the
SPL metering level was set at -12dB.

The speaker leads were then changed to a pair of average sensitivity
speakers i built,
the Sublimes, shown and described at
http://www.turneraudio.com.au/loudspeakers-new.html

Now the sensitivity stated at my site for these speakers is 90dB/W/M,
but
at present I have a series resistor in front of the midrange/treble unit
which drops the sensitivity to 88dB with the same applied voltage.
Impedance is 5.6 ohms average.

The level attained with my speakers was 0.0dB at 1kHz with the 0.85 vac
of pink noise.

The SPL meter registered +/- 2dB level changes along the band for both
ERA and my own,
so there was no need to do complete response tests to confirm what was
so
plain to my ears which is that the ERA ESLs produce an SPL of 12dB
lower than my own for the same input voltage.

So to power my speakers to healthy levels of listening I would use most
days an amp capable of putting 5.6Vrms into 5.6 ohms would work, which
is 5.6 watts maximum peak power.
In practice, a 5.6 watt amp will begin to sound weak so a 25 watter is
the lowest power I can get by well on.
50 watts is a luxury, 300 watts an excess.
When measuring busy music, the Vac in my meters is rarely more than
0.85Vac, and allowing for
Vac meter reading to maximum Vrms ratio of 1:4, I get maybe 3.4Vrms, so
allowing to go to 5.6Vrms is
being generous to the music.

With clasical orchestral music, a far higher ratio between the solo
passages and all musicians occurs,
and I need the 25 watts some days for peaks.

So with 0.85Vrms applied to my 5.6 ohms speakers I have 2.8Vrms max =
1.4 Watts max.

With the ERA ESLs, I need to apply 4 times the dynamic speaker signal
voltage
to get the same SPL level, ie, 2.8 x 4 = 11.2Vrms.

The average speaker Z = approx 14 ohms, so power max = 8.96 watts.

So ERA speakers need 8.96/1.4 = 6.4 times the power to produce the same
SPL meter reading.
The SPL meter merely measures the amplified microphone voltage.

So if my dynamic speakers produce 88dB for 1 watt @ 1M,
then the ERA need 6.4 watts for the same job, or if you like,

1 WATT FED INTO AN ER AUDIO ESL-IIIB makes an SPL = 80dB at 1M.

At 3 metres away, I measured a 6 dB drop in SPL levels for both speakers
in the room,
so 1 watt max doesn't produce a huge SPL volume in the ESLs.

This is what I consider to be the official maximum possible safe working
sensitivity
attainable when they are engineered for a flat response and to ensure
absence of arcing problems,
by setting the EHT at the maximum safe level of -2,700V.

The only way to improve the sensitivity is to use much less series
resistance
between the step up transformer secondaries and the bass panels, and
then the
response at 100Hz and 10kHz will barely change, but the middle of the
band will
be as arched as you want, and the sound will become just midrange with
nothing else,
ie, sound like crap.


Now most amplifiers are set up to run best with 8 ohms.

The Musical Fidelity A3 which I tried with the ERA ESLs is good for
120W into 8 ohms, and 240W into 4.
But this implies 31Vrms is available and will make only 68 watts into 14
ohms.


But the Z of ESL panels varies greatly, and in the case of ERA ESL
I have here with the necessary input filtering to prevent
the worst of membrane flapping and banging,
Z at 50Hz = 20ohms, falling to 10ohms at 100hz, then to a peak at 20ohms
at 300Hz and
then falling to 3 ohms at 20kHz.

So where most of the audio energy is located between 50Hz and 1 kHz,
the 120watt amp is barely enough.

Now much music doesn't have a convenient nice ratio between average
comfortable levels and peaks
and I have a good recording here where some laid back jazz
has a guy on a double bass going a bit beserko, in a nice exciting kind
of way.
As you all know modern bass levels in music are often very high,
and when i played this music to a friend last saturday using an amp
capable of 22Vrms into any load above 8 ohms,
the ERA speakers just made horrible noises when the bass player got
keen,
although the rest of the music seemed at a very relaxed level, so
the bass signals present went probably way over 22Vrms, max, and the amp
clipped.
Later in the day when my customer came for a listen I tried the MF A3,
and the same thing happened, but this time
it was a combination of speakers and amps clipping, and finally
the 300 watter gave the best performance but was occasionally triggering
its clipping indicators.

My customer wasn't impressed, because he knows his 25 watt SEUL will not
be able to adequately drive these ESL,
even if I altered the OPT windings to give a match to 13.4 ohms instead
of to
8 ohms, and thus lift the available voltage possible.

So although the sensitivity testing I have conducted tonight describes
what sensitivity is possible,
the real maximum signal capability at frequency has yet to be graphed.
It would appear the ESL just don't like a large bass signal voltage.

If I changed to a higher ratio for the step up transformer, say from
the existing 1:90 to 1:180, thus allowing a drive voltage = 1/2 what is
now needed,
the impedance of the speaker would be reduced from 14 ohms average to
about 2.8 ohms average.
At HF the speaker would indeed become horrible to drive!!!

So the maximum practical step up ratio = about 1:140.

Tonight I also phoned my colleague in Sydney who also does audio repairs
and
builds the occasional new speakers and amps and he wants to give me a
complete unassembled ERA kit
bought some years ago by a prominent audiophile in the
Audiophile Society of NSW. He was given it by a Mr L
because Mr L had bought it while a friend was constructing such a kit.
The friend's efforts were found to be very unsatisfactory, and Mr L
traded away
his unbuilt kit for chassis and other amplifier parts supplied by my
colleague, Mr S.

So despite what idiots like Basset are saying, I didn't have to ask
around
much amoung the few people I know before I found someone who wasted his
money with ERA.
I also have a second contact in Sydney who has been doing his own
experiments with ESL panels now
for 20 years, and he is the one man in the ASON club who knows a lot
about ESL,
and has acted as a partner to another Sydney man who have restored many
Quad ESL57.

So before writing off the ERA ESL, its fair that I research a little
more
and measure the levels of applied voltages which get the ESLs into
trouble,
which are at the bass frequencies.
As I said a month ago, it may be quite pointless to expect to get decent
bass
headroom and sound quality, and to build bass speakers in boxes under
the panels like Martin Logan.
Both guys who listened to the ESL and my dynamics last saturday
concluded the
bass from my speakers was far better sounding, and overall distortion
was lower.


Patrick Turner.

[Don Pearce]

On Mon, 28 May 2007 17:17:57 GMT, Patrick Turner
wrote:

For those left wondering what the heck the facts are about sensitivity,
here they are.

But first I must describe the conditions of the test.

A reliable source of pink noise was used.

To ascertain what signal voltages are involved,
an osciliscope was set up so that 10Vrms of a 1kHz sine wave from 300
watt SS amp
registered 1/4 of the screen vertical room, ie, 40Vrms will just fill
the screen completely.
This is 56 peak ac volts.

Pink noise was fed to the amp and adjusted for level until
the peaks in the noise were equal to the height of the screen, ie, the
maximum
peak signals were 56Vpk.

No amplifier clipping was visible.

The maximum Vrms within the noise could be up to 40Vrms.

Using a variety of DMM meters in their Vac ranges all gave 12.1Vac when
in fact there
was a maximum of 40Vrms present.
So the ratio of maximum Vrms within the noise was 40 / 12.1 = 3.3:1

The crest factor of pink noise is about 11dB, so no, you won't get
40Vrms from 56V peak of pink noise.15Vrms would be more like it. Your
technique of basing this on filling the screen is not a good one with
noise, either. It is very hard to catch the odd spike that defines the
peak. Just go with the RMS meter every time.

I'm afraid the rest was just too much of a jumble to make anything of
it. ANy chance of a precis?

d

--
Pearce Consulting
http://www.pearce.uk.com

[TT]

"Patrick Turner" wrote in message
...
For those left wondering what the heck the facts are about sensitivity,
here they are.

snipped technical stuff

As I said a month ago, it may be quite pointless to expect to get decent
bass
headroom and sound quality, and to build bass speakers in boxes under
the panels like Martin Logan.
Both guys who listened to the ESL and my dynamics last saturday
concluded the
bass from my speakers was far better sounding, and overall distortion
was lower.


Patrick Turner.

Thank you Patrick for the very detailed and lengthy accounts of your
technical dramas with these speakers. I have found it more than interesting
and have been made to feel very pleased that I never ended up with these
kits.

What you have done for me is *prove* that my initial listening evaluations
of these were quite correct i.e that they didn't go very loud, they lacked
bottom end and the bass quite frankly sucked (flapped?).

As you have also proved it is also very hard to beat conventual cone
speakers for overall performance.

One question though, where to now? Your customer has obviously been very
upset that he either throws his amps away or these speakers. Also I would
hate to see how much your bill is by now, *IF* you charged for every hour at
your going repair rates ;-)

Cheers TT

[Phil Allison]

"Patrick Turner"
For those left wondering what the heck the facts are about sensitivity,
here they are.

But first I must describe the conditions of the test.

A reliable source of pink noise was used.

To ascertain what signal voltages are involved,
an osciliscope was set up so that 10Vrms of a 1kHz sine wave from 300
watt SS amp
registered 1/4 of the screen vertical room, ie, 40Vrms will just fill
the screen completely.
This is 56 peak ac volts.

Pink noise was fed to the amp and adjusted for level until
the peaks in the noise were equal to the height of the screen, ie, the
maximum
peak signals were 56Vpk.

No amplifier clipping was visible.

The maximum Vrms within the noise could be up to 40Vrms.

Using a variety of DMM meters in their Vac ranges all gave 12.1Vac when
in fact there
was a maximum of 40Vrms present.
So the ratio of maximum Vrms within the noise was 40 / 12.1 = 3.3:1


** Shame you got nearly everything wrong.

1. You need a " true rms " volt meter, with wider than audio band
response, to determine the rms value of a pink noise signal. No way out of
this exists using scopes or average responding meters.

2. For speaker sensitivity testing, the pink noise MUST be first be
filtered to remove out of band frequencies - supersonic and sub sonic
signals are not being reproduced so having them present at the speaker
terminals gives a false low reading for the " dB/watt/metre " figure.

3. One tests the sensitivity of a speaker or driver ( woofer, mid or
tweeter) in its *pass band* - so the pink noise must be filtered to match
the expected pass band. Pointless and wrong to test a sub woofer with pink
noise that extends out to beyond 20 kHz. When the dB/W/m figure is found,
the procedure is to also quote the bandwidth of the test signal that was *
actually applied * to the speaker or driver.

4. Nominal watt figures are used for sensitivity testing - ie the true
rms noise voltage is adjusted to 2.82 volts for a nominal 8 ohm speaker to
give 1 watt. This is then deemed to be 1 watt input, whether it really is or
not.

Failure to take account of the above leads to *gross errors* in the result.




........ Phil

[Patrick Turner]

Don Pearce wrote:

On Mon, 28 May 2007 17:17:57 GMT, Patrick Turner
wrote:

For those left wondering what the heck the facts are about sensitivity,
here they are.

But first I must describe the conditions of the test.

A reliable source of pink noise was used.

To ascertain what signal voltages are involved,
an osciliscope was set up so that 10Vrms of a 1kHz sine wave from 300
watt SS amp
registered 1/4 of the screen vertical room, ie, 40Vrms will just fill
the screen completely.
This is 56 peak ac volts.

Pink noise was fed to the amp and adjusted for level until
the peaks in the noise were equal to the height of the screen, ie, the
maximum
peak signals were 56Vpk.

No amplifier clipping was visible.

The maximum Vrms within the noise could be up to 40Vrms.

Using a variety of DMM meters in their Vac ranges all gave 12.1Vac when
in fact there
was a maximum of 40Vrms present.
So the ratio of maximum Vrms within the noise was 40 / 12.1 = 3.3:1

The crest factor of pink noise is about 11dB, so no, you won't get
40Vrms from 56V peak of pink noise.15Vrms would be more like it. Your
technique of basing this on filling the screen is not a good one with
noise, either. It is very hard to catch the odd spike that defines the
peak. Just go with the RMS meter every time.

I'm afraid the rest was just too much of a jumble to make anything of
it. ANy chance of a precis?

Blind Freddy can see that much more than 0.15Vrms of any given F is
present
at any instant in a pink noise signal displayed on a CRO.

But lest there be any confusion in ppl's minds I described how and what
I measured.
Pink noise is very much like a busy music signal.

But fine, work on the idea that my 300 watt amp can process 56peak volts
without THD beyond 0.005%.

I will do slightly more because the rails are +/-65V, and the driver
rails are +/- 83V,
enabling the output mosfets to turn on with a low Vds.

A peak and hold meter would reveal that the 56peak volts is about right.

At present I don't have a commercially made P&H voltage measurement
facility.

The point of the exercise was to establish the voltage sensitivity
and power sensitivity of these ESL speakers.

It takes 4 times the drive voltage to make the same SPL as my dynamics
need.

My dynamics create 88dB/W/M from one watt of applied power.

The ESL need 6.4 watts to produce the same level allowing for the
impedance difference between
the two types of speaker.

So where a 25 watt amp gives me sufficient listening levels for my
dynamics,
I need about 150 watts to get the same ceiling.
A "300 watt" amp that can make 350watts into 3 ohms max will make about
235 W into 8 ohms,
and 132 W into 14 ohms.

My 300 watter can only produce a maximum 44Vrms for 14 ohms and this
voltage
is only 4 times the maxima used for the tests I did last night.

But I still have to produce a graph for maximum allowable voltages
for bass above 40Hz before distortions caused by excessive membrane
movement occurs.
I suspect a quite low threshold exists for bass frequency voltages.

I hope this summation clarifies what I said last night.

Patrick Turner.









d

--
Pearce Consulting
http://www.pearce.uk.com

[Patrick Turner]

TT wrote:

"Patrick Turner" wrote in message
...
For those left wondering what the heck the facts are about sensitivity,
here they are.

snipped technical stuff

As I said a month ago, it may be quite pointless to expect to get decent
bass
headroom and sound quality, and to build bass speakers in boxes under
the panels like Martin Logan.
Both guys who listened to the ESL and my dynamics last saturday
concluded the
bass from my speakers was far better sounding, and overall distortion
was lower.


Patrick Turner.

Thank you Patrick for the very detailed and lengthy accounts of your
technical dramas with these speakers. I have found it more than interesting
and have been made to feel very pleased that I never ended up with these
kits.

What you have done for me is *prove* that my initial listening evaluations
of these were quite correct i.e that they didn't go very loud, they lacked
bottom end and the bass quite frankly sucked (flapped?).

As you have also proved it is also very hard to beat conventual cone
speakers for overall performance.

One question though, where to now? Your customer has obviously been very
upset that he either throws his amps away or these speakers. Also I would
hate to see how much your bill is by now, *IF* you charged for every hour at
your going repair rates ;-)

Cheers TT

Its the first time I have built a pair of ESL.

I am thinking that I may have to use dynamic bass units.
I may also have to wind better step up trannies.

Then we may have something good.

The other limiting factor for sensitivity is the
amount of treble available.

The ESL-IIIB has two identical 1.5M x 0.18M bass/midrange panels
mounted each side of a centre treble panel 1.5M x 0.04M.

The membrane distance to the treble stators is approximately the same as
for the bass/mids.

The treble panel stators are driven directly from the secondary of the
SUT
but a crossover RC network exists from each sec end to give pole at
1kHz consisting of 0.0033uF at each end of the sec with 50k to CT.
This keeps bass F out of the treble panel, but from 1khz, the treble is
effectively
directly driven off the SUT.

The sensitivity of the treble panel is about the same as the bass panels
at 100Hz when there
is 450k between each end of the SUT sec ans the bass stators.

The equivalent circuit I have for the bass panels is 900k in series with
800pF.
The 800pF has a reactance of 2 megohms at 100Hz, so the series 900k
makes little difference to
100Hz output; one could have 90k series R and have only a small increase
in SPL at 100Hz.


It would be possible to reduce the series R to bass and lift the
sensitivity of bass panels, but then the response with low bass would be
much poorer, because the
slope of the response tilts upwards from 100Hz and with R = 90k, you'd
have a horrible
12db+ peak at 1kHz.

If the sensitivity of treble panels could be lifted to match the lifted
midrange then a panel
with good midrange and reble is the result, but without bass, and a
woofer in a box becomes imperative.

But the treble sensitivity is difficult to change predictably.
The gap between membrane and stators would need to be about halved, and
it is doable
since the membrane movement above 300hz becomes so tiny.

I am presently locked into using the panels as I have built them
according to
instructions, but with my minor mods, and removing treble stators and
placing in 1mm plastic packing strips to
reduce the membrane distance is very difficult, and means i will have to
rebuild the panel I have got working a sixth time, and I will run out of
supplied materials.

I should not have to do any of this, and all the science and facts and
figures I am presenting here
should have been established by the kit maker, but he has been able to
sell kits without
the carefully published R&D.

The kit maker has never really adressed the issues which discerning ppl
will raise about his speakers.

I've made no money so far while battling to get somewhere, because I
have to proove
I am woth paying, ie, my customer has to be happy with what I have done
in listening tests.
He isn't happy so far, and nor am I, and I have to absorb the loss.

I may have spent 30 hours so far, worth about $1,500, but I will never
get it.

I still have to build the second panel and make the timber surround, and
at least build
enclosures for the transformer and EHT supply, and perhaps make them
large enough
at about 50L to have a 200mm woofer reflex speaker capable of the 20Hz
to 300Hz
of my dynamic drivers.

The transformers used for the ESL are pretty damn ordinary, and nothing
like the quality which
is used for Quad speakers. The guy I had in mind who could wind better
ones to my spec
has retired to the coast after 50 years in the trade.

All the old guys who could really do things either retire, expire, or
die.

Patrick Turner.

[Chris Hornbeck]

On Tue, 29 May 2007 10:41:55 +1000, "Phil Allison"
wrote:

1. You need a " true rms " volt meter, with wider than audio band
response, to determine the rms value of a pink noise signal. No way out of
this exists using scopes or average responding meters.

This is true for a non-linear system, much less so as
the system is considered more linear. For loudspeakers
linearity must either be proven or assumed for any
(magnitude) response measurement.

2. For speaker sensitivity testing, the pink noise MUST be first be
filtered to remove out of band frequencies - supersonic and sub sonic
signals are not being reproduced so having them present at the speaker
terminals gives a false low reading for the " dB/watt/metre " figure.

This is incorrect. Removing out-of-band signal can help
to increase measurement sensitivity, but doesn't effect
in-band response in a linear system.

3. One tests the sensitivity of a speaker or driver ( woofer, mid or
tweeter) in its *pass band* - so the pink noise must be filtered to match
the expected pass band. Pointless and wrong to test a sub woofer with pink
noise that extends out to beyond 20 kHz. When the dB/W/m figure is found,
the procedure is to also quote the bandwidth of the test signal that was *
actually applied * to the speaker or driver.

4. Nominal watt figures are used for sensitivity testing - ie the true
rms noise voltage is adjusted to 2.82 volts for a nominal 8 ohm speaker to
give 1 watt. This is then deemed to be 1 watt input, whether it really is or
not.

Good points. Of even greater concern are any measurements
made in-room. Removing the room from measurements is only
(comparatively recently) possible with time-windowed measurement.

All good fortune,

Chris Hornbeck

[Patrick Turner]

Phil Allison wrote:

"Patrick Turner"
For those left wondering what the heck the facts are about sensitivity,
here they are.

But first I must describe the conditions of the test.

A reliable source of pink noise was used.

To ascertain what signal voltages are involved,
an osciliscope was set up so that 10Vrms of a 1kHz sine wave from 300
watt SS amp
registered 1/4 of the screen vertical room, ie, 40Vrms will just fill
the screen completely.
This is 56 peak ac volts.

Pink noise was fed to the amp and adjusted for level until
the peaks in the noise were equal to the height of the screen, ie, the
maximum
peak signals were 56Vpk.

No amplifier clipping was visible.

The maximum Vrms within the noise could be up to 40Vrms.

Using a variety of DMM meters in their Vac ranges all gave 12.1Vac when
in fact there
was a maximum of 40Vrms present.
So the ratio of maximum Vrms within the noise was 40 / 12.1 = 3.3:1

** Shame you got nearly everything wrong.

Probably you are right because you know more about such things than I
do.

However, the basic issue for me is the relative VOLTAGE sensitivity.

1. You need a " true rms " volt meter, with wider than audio band
response, to determine the rms value of a pink noise signal. No way out of
this exists using scopes or average responding meters.

Well, the CRO seemed to tell me what maximum peak voltages I need to
produce a given
sound level, and what an amp must produce.


2. For speaker sensitivity testing, the pink noise MUST be first be
filtered to remove out of band frequencies - supersonic and sub sonic
signals are not being reproduced so having them present at the speaker
terminals gives a false low reading for the " dB/watt/metre " figure.

The pink noise source I built has such filtering with -3dB at 20Hz and
20kHz.


3. One tests the sensitivity of a speaker or driver ( woofer, mid or
tweeter) in its *pass band* - so the pink noise must be filtered to match
the expected pass band. Pointless and wrong to test a sub woofer with pink
noise that extends out to beyond 20 kHz. When the dB/W/m figure is found,
the procedure is to also quote the bandwidth of the test signal that was *
actually applied * to the speaker or driver.

The ESL speakers have been constructed to as near full range as
possible,
in this case its 40Hz to 20kHz.

4. Nominal watt figures are used for sensitivity testing - ie the true
rms noise voltage is adjusted to 2.82 volts for a nominal 8 ohm speaker to
give 1 watt. This is then deemed to be 1 watt input, whether it really is or
not.



Failure to take account of the above leads to *gross errors* in the result.

Whatever my errors and equipment limitations may be, the fact remains
that
the ESL speakers i am testing require 4 times voltage level required for
my dynamic speakers with much the same response flatness and BW.

The Digital Fluke meter has limited ac BW from about 20Hz to 2kHz, and
when 20Hz to 20kHz of pink noise with peak value of close to 56V is
present,
it reads about 12Vac, with fluctuations up and down, but averaging about
this reading.
The Fluke has an LCD bar meter below the FCD figure display, so some
indication can be
gained of what reading can be taken.
Vrms values for any given frequency would be hard to measures since the
noise
contains randomly varying F, phase, and amplitude.

With busy rock and roll the figures I measured are about the same as
pink noise,
but although the way I have described it all is not perfect, its a
guide,
and if I measured 0.85Vac with busy music with my Fluke, its about
where average listening levels would be for music.

The ESL I am building require 4 times the voltage applied for the same
measured and perceived sound levels compared to the dynamics.

To allow for enough SPL ceiling for all my ordinary modest needs,
my dynamics need an amp capable of at least 12Vrms max,
which is 25.7 watts into 5.6 ohms.

If I wanted the same ceiling with the ESL, I need 48Vrms max,
and this is 164 watts into 14 ohms.

The Musical Fidelity A3 capable of 120W into 8 ohms is barely enough to
drive these ESL,
because its capable of only 31Vrms, which is only 68 watts into 14 ohms.

My problem with these ESL is not only the low sensitivity
but a low ceiling for bass F.

R&D is continuing.

Patrick Turner.



















....... Phil

[Chris Hornbeck]

On Tue, 29 May 2007 04:33:36 GMT, Patrick Turner
wrote:

My problem with these ESL is not only the low sensitivity
but a low ceiling for bass F.

R&D is continuing.

You're observing results more than 10dB out-of-bounds;
that's not a measurement artifact, it's a real life
problem.

I could quibble about this, that, or the other detail
of loudspeaker measurement, but 10dB (plus) is 10dB (plus).

Ain't no talking a way around a number like that.

Much thanks, as always,

Chris Hornbeck

[Don Pearce]

On Tue, 29 May 2007 03:19:48 GMT, Patrick Turner
wrote:



Don Pearce wrote:

On Mon, 28 May 2007 17:17:57 GMT, Patrick Turner
wrote:

For those left wondering what the heck the facts are about sensitivity,
here they are.

But first I must describe the conditions of the test.

A reliable source of pink noise was used.

To ascertain what signal voltages are involved,
an osciliscope was set up so that 10Vrms of a 1kHz sine wave from 300
watt SS amp
registered 1/4 of the screen vertical room, ie, 40Vrms will just fill
the screen completely.
This is 56 peak ac volts.

Pink noise was fed to the amp and adjusted for level until
the peaks in the noise were equal to the height of the screen, ie, the
maximum
peak signals were 56Vpk.

No amplifier clipping was visible.

The maximum Vrms within the noise could be up to 40Vrms.

Using a variety of DMM meters in their Vac ranges all gave 12.1Vac when
in fact there
was a maximum of 40Vrms present.
So the ratio of maximum Vrms within the noise was 40 / 12.1 = 3.3:1

The crest factor of pink noise is about 11dB, so no, you won't get
40Vrms from 56V peak of pink noise.15Vrms would be more like it. Your
technique of basing this on filling the screen is not a good one with
noise, either. It is very hard to catch the odd spike that defines the
peak. Just go with the RMS meter every time.

I'm afraid the rest was just too much of a jumble to make anything of
it. ANy chance of a precis?

Blind Freddy can see that much more than 0.15Vrms of any given F is
present
at any instant in a pink noise signal displayed on a CRO.

But lest there be any confusion in ppl's minds I described how and what
I measured.
Pink noise is very much like a busy music signal.

But fine, work on the idea that my 300 watt amp can process 56peak volts
without THD beyond 0.005%.

I will do slightly more because the rails are +/-65V, and the driver
rails are +/- 83V,
enabling the output mosfets to turn on with a low Vds.

A peak and hold meter would reveal that the 56peak volts is about right.

At present I don't have a commercially made P&H voltage measurement
facility.

The point of the exercise was to establish the voltage sensitivity
and power sensitivity of these ESL speakers.

It takes 4 times the drive voltage to make the same SPL as my dynamics
need.

My dynamics create 88dB/W/M from one watt of applied power.

The ESL need 6.4 watts to produce the same level allowing for the
impedance difference between
the two types of speaker.

So where a 25 watt amp gives me sufficient listening levels for my
dynamics,
I need about 150 watts to get the same ceiling.
A "300 watt" amp that can make 350watts into 3 ohms max will make about
235 W into 8 ohms,
and 132 W into 14 ohms.

My 300 watter can only produce a maximum 44Vrms for 14 ohms and this
voltage
is only 4 times the maxima used for the tests I did last night.

But I still have to produce a graph for maximum allowable voltages
for bass above 40Hz before distortions caused by excessive membrane
movement occurs.
I suspect a quite low threshold exists for bass frequency voltages.

I hope this summation clarifies what I said last night.

Patrick Turner.

No, it is still a jumble. Try to organize your thoughts a bit please.

d

--
Pearce Consulting
http://www.pearce.uk.com

[Patrick Turner]

I hope this summation clarifies what I said last night.

Patrick Turner.

No, it is still a jumble. Try to organize your thoughts a bit please.

Well let me break down what i said a litle further, just for you Don:-


My own dynamic speakers need 2.36vrms applied at all F to make an SPL =
88dB at 1M.

ERA ESL-IIIB speakers need 9.47Vrms applied at all F at produce an SPL =
88dB at 1M.

Therefore the sensitivity
with regard to applied voltage only is 12dB worse for the ESL compared
to my dynamics.

Or 8dB worse with regard to db SPL per watt per meter, because of the
impedance differences.

Is that clear enough for you? Or would you like it even more simple?

Patrick Turner.




d

--
Pearce Consulting
http://www.pearce.uk.com

[Don Pearce]

On Tue, 29 May 2007 07:02:51 GMT, Patrick Turner
wrote:


I hope this summation clarifies what I said last night.

Patrick Turner.

No, it is still a jumble. Try to organize your thoughts a bit please.

Well let me break down what i said a litle further, just for you Don:-


My own dynamic speakers need 2.36vrms applied at all F to make an SPL =
88dB at 1M.

ERA ESL-IIIB speakers need 9.47Vrms applied at all F at produce an SPL =
88dB at 1M.

Therefore the sensitivity
with regard to applied voltage only is 12dB worse for the ESL compared
to my dynamics.

Or 8dB worse with regard to db SPL per watt per meter, because of the
impedance differences.

Is that clear enough for you? Or would you like it even more simple?

Patrick Turner.


No - I got all that.

Just a few points that may help you.

Ignore manufacturers specs about dB SPL vs watts. Always consider it
vs volts, because that is really how speakers work. The fact that
impedance is frequency-dependent makes the power comparison difficult.

If you are comparing speakers of different impedance, ignore the power
rating of your amplifier. Assuming that it is capable of delivering
enough current, it is the voltage it can produce that matters; that is
what is set by the volume control. So measuring the output voltage is
the best way of doing the sums.

If you are comparing loudness with various signals, ignore peak values
and go for RMS - this is the simplest, and probably the best way of
doing things. Use peak values for nothing more than making sure you
aren't clipping.

Most cheap multimeters don't measure RMS voltage. They measure peak
voltage with a rectifier, then scale the meter reading to give you the
RMS equivalent for a sine wave. If your signal is something other than
a sine wave (eg pink noise) you can expect your multimeter to read
incorrectly. Spend a bit more for one that does the maths properly and
will give you true RMS measurements at many different crest factors.

Finally, who really cares about speaker sensitivity? These days it is
simple to get whatever amplifier power you need, so just bolt together
whatever gives you the sound level you need and don't worry about it.

d

--
Pearce Consulting
http://www.pearce.uk.com

[jasee]

Don Pearce wrote:
Finally, who really cares about speaker sensitivity? These days it is
simple to get whatever amplifier power you need, so just bolt together
whatever gives you the sound level you need and don't worry about it.

I suppose cost might come into most peoples' heads at this point?

[Don Pearce]

On Tue, 29 May 2007 09:07:20 +0100, "jasee"
wrote:

Don Pearce wrote:
Finally, who really cares about speaker sensitivity? These days it is
simple to get whatever amplifier power you need, so just bolt together
whatever gives you the sound level you need and don't worry about it.

I suppose cost might come into most peoples' heads at this point?


Sure, but you only have to buy it once - so just save up.

d
--
Pearce Consulting
http://www.pearce.uk.com

[Patrick Turner]

Don Pearce wrote:

On Tue, 29 May 2007 07:02:51 GMT, Patrick Turner
wrote:


I hope this summation clarifies what I said last night.

Patrick Turner.

No, it is still a jumble. Try to organize your thoughts a bit please.

Well let me break down what i said a litle further, just for you Don:-


My own dynamic speakers need 2.36vrms applied at all F to make an SPL =
88dB at 1M.

ERA ESL-IIIB speakers need 9.47Vrms applied at all F at produce an SPL =
88dB at 1M.

Therefore the sensitivity
with regard to applied voltage only is 12dB worse for the ESL compared
to my dynamics.

Or 8dB worse with regard to db SPL per watt per meter, because of the
impedance differences.

Is that clear enough for you? Or would you like it even more simple?

Patrick Turner.


No - I got all that.

Just a few points that may help you.

Ignore manufacturers specs about dB SPL vs watts. Always consider it
vs volts, because that is really how speakers work. The fact that
impedance is frequency-dependent makes the power comparison difficult.

I agree, much confusion occurs with regard to watts if we take into
account the
impedance, but it should be mentioned though.



If you are comparing speakers of different impedance, ignore the power
rating of your amplifier. Assuming that it is capable of delivering
enough current, it is the voltage it can produce that matters; that is
what is set by the volume control. So measuring the output voltage is
the best way of doing the sums.

But one needs to know about what loads will have to be powered to
calculate currents.

The typical ESL has high Z at LF with lots of voltage needed with low
current,
and low Z at HF which means that if the signal level is the same 10kHz
as it might be at 100Hz,
then an impossibly large power ability for an amp has to be present.

But Quad II amps with pretty poor current ability manage to work fine
with ESL57
because there is rarely if ever a large voltage signal to be applied to
the
2ohms Z at 18kHz.
The small amount of declining signal above 7kHz means very little
current needs to be fed
to the low Z at HF, and a 20 watts amp is enough, and there is no need
to have 312
watts to give 25vrms across 2 ohms at 18kHz.



If you are comparing loudness with various signals, ignore peak values
and go for RMS - this is the simplest, and probably the best way of
doing things. Use peak values for nothing more than making sure you
aren't clipping.

Yes, but music has voltages and F all over the place at different F
like pink noise, and music has less F than noise, and the the F of music
is
dominated by strong bass tones, drums, with harmonically related tones
riding along on the crests of the bass tones.

So we do need to keep an eye on peak voltage abilities of all gear used
in a line up.



Most cheap multimeters don't measure RMS voltage. They measure peak
voltage with a rectifier, then scale the meter reading to give you the
RMS equivalent for a sine wave.

Yes, 0.707 x peak voltage = Vrms.

If your signal is something other than
a sine wave (eg pink noise) you can expect your multimeter to read
incorrectly.

With my Fluke, and another cheaper DVM, when max peak noise was deemed
to be 56V,
I got a hovering 12Vac measurement on the DVMs.

Spend a bit more for one that does the maths properly and
will give you true RMS measurements at many different crest factors.

I will search for one, or build one.



Finally, who really cares about speaker sensitivity? These days it is
simple to get whatever amplifier power you need, so just bolt together
whatever gives you the sound level you need and don't worry about it.

Hang on a minute, these ESL need a maximum of 11.2vrms to give very tame
levels of music.

To give any sort of performance these ESL require an amp to be able to
make 44Vrms max
to give a comfortable ceiling. amps are usually rated for 8ohms, and
44Vrms into 8 = 242 watts,
so a "300 watt amp" would be ideal.

If an amp is capable of 44Vrms for the maximum Z of 20 ohms of these ESL
as I have set them up,
then its 100watts, but that's only 28V into 8 ohms, so most 100 watt
amps won't be good enough.
A solution is to play with impedance matching to by using an impedance
matching tranny
to reduce the ESL average Z from say 14 ohms to 7 ohms,
thus allowing better use of the 100 watt amps power band.

A friend here has a pair of such Z matching toroidal trannies which I
placed into wooden boxes
which gave him 20:20, 20:10, 20:8, 20:4, etc, and
BW I measured was 4 Hz to 1MHz, and virtually no THD.
he finds it useful to transform a modern speaker's common low Z to a
high Z
more suited to the tube amps he's using, therefore reducing the
effective Rout of the amp, improving damping,
and reducing THD/IMD by a factor of 0.25.
I forget where he bought them in the US somewhere, but they did the
business really well.

Patrick Turner.







--
Pearce Consulting
http://www.pearce.uk.com

[Don Pearce]

On Tue, 29 May 2007 09:44:52 GMT, Patrick Turner
wrote:



Don Pearce wrote:

On Tue, 29 May 2007 07:02:51 GMT, Patrick Turner
wrote:


I hope this summation clarifies what I said last night.

Patrick Turner.

No, it is still a jumble. Try to organize your thoughts a bit please.

Well let me break down what i said a litle further, just for you Don:-


My own dynamic speakers need 2.36vrms applied at all F to make an SPL =
88dB at 1M.

ERA ESL-IIIB speakers need 9.47Vrms applied at all F at produce an SPL =
88dB at 1M.

Therefore the sensitivity
with regard to applied voltage only is 12dB worse for the ESL compared
to my dynamics.

Or 8dB worse with regard to db SPL per watt per meter, because of the
impedance differences.

Is that clear enough for you? Or would you like it even more simple?

Patrick Turner.


No - I got all that.

Just a few points that may help you.

Ignore manufacturers specs about dB SPL vs watts. Always consider it
vs volts, because that is really how speakers work. The fact that
impedance is frequency-dependent makes the power comparison difficult.

I agree, much confusion occurs with regard to watts if we take into
account the
impedance, but it should be mentioned though.



If you are comparing speakers of different impedance, ignore the power
rating of your amplifier. Assuming that it is capable of delivering
enough current, it is the voltage it can produce that matters; that is
what is set by the volume control. So measuring the output voltage is
the best way of doing the sums.

But one needs to know about what loads will have to be powered to
calculate currents.

Sure but only to the extent of "have I got enough current available to
power this load?"

The typical ESL has high Z at LF with lots of voltage needed with low
current,
and low Z at HF which means that if the signal level is the same 10kHz
as it might be at 100Hz,
then an impossibly large power ability for an amp has to be present.

But Quad II amps with pretty poor current ability manage to work fine
with ESL57
because there is rarely if ever a large voltage signal to be applied to
the
2ohms Z at 18kHz.
The small amount of declining signal above 7kHz means very little
current needs to be fed
to the low Z at HF, and a 20 watts amp is enough, and there is no need
to have 312
watts to give 25vrms across 2 ohms at 18kHz.

Possibly true, but it does depend on music type. I suspect Quad made
their assumptions on the basis of vinyl source, and no
hypercompression. Things are no longer that way.



If you are comparing loudness with various signals, ignore peak values
and go for RMS - this is the simplest, and probably the best way of
doing things. Use peak values for nothing more than making sure you
aren't clipping.

Yes, but music has voltages and F all over the place at different F
like pink noise, and music has less F than noise, and the the F of music
is
dominated by strong bass tones, drums, with harmonically related tones
riding along on the crests of the bass tones.

So we do need to keep an eye on peak voltage abilities of all gear used
in a line up.


I don't know what F is here - do you mean frequencies? If so, I don't
understand what you mean when you say music has less frequencies than
noise?



Most cheap multimeters don't measure RMS voltage. They measure peak
voltage with a rectifier, then scale the meter reading to give you the
RMS equivalent for a sine wave.

Yes, 0.707 x peak voltage = Vrms.

For sine waves only, of course.

If your signal is something other than
a sine wave (eg pink noise) you can expect your multimeter to read
incorrectly.

With my Fluke, and another cheaper DVM, when max peak noise was deemed
to be 56V,
I got a hovering 12Vac measurement on the DVMs.

Spend a bit more for one that does the maths properly and
will give you true RMS measurements at many different crest factors.

I will search for one, or build one.


If you have a Fluke, it probably already does a good job in this
regard.



Finally, who really cares about speaker sensitivity? These days it is
simple to get whatever amplifier power you need, so just bolt together
whatever gives you the sound level you need and don't worry about it.

Hang on a minute, these ESL need a maximum of 11.2vrms to give very tame
levels of music.

I have no idea what that means.

To give any sort of performance these ESL require an amp to be able to
make 44Vrms max
to give a comfortable ceiling. amps are usually rated for 8ohms, and
44Vrms into 8 = 242 watts,
so a "300 watt amp" would be ideal.


OK...

If an amp is capable of 44Vrms for the maximum Z of 20 ohms of these ESL
as I have set them up,
then its 100watts, but that's only 28V into 8 ohms, so most 100 watt
amps won't be good enough.

It is as I said - it is the volts that matter. Amplifiers supply
volts, speakers consume watts. If your speakers are typically of a
high impedance, then they may indeed need quite a few volts to produce
those watts.

A solution is to play with impedance matching to by using an impedance
matching tranny
to reduce the ESL average Z from say 14 ohms to 7 ohms,
thus allowing better use of the 100 watt amps power band.

A friend here has a pair of such Z matching toroidal trannies which I
placed into wooden boxes

Toroidal trannies wouldn't be my first choice for audio power. They
have no air gap, and so tend to be quite non-linear.

which gave him 20:20, 20:10, 20:8, 20:4, etc, and
BW I measured was 4 Hz to 1MHz, and virtually no THD.

How did you measure this? This is indeed a most remarkable tranny if
it can do this.

he finds it useful to transform a modern speaker's common low Z to a
high Z
more suited to the tube amps he's using, therefore reducing the
effective Rout of the amp, improving damping,
and reducing THD/IMD by a factor of 0.25.
I forget where he bought them in the US somewhere, but they did the
business really well.

Patrick Turner.

If you use this on your amp to make a better impedance match, it will
reduce the damping factor, not improve it. As for THD - what it does
to that is anybody's guess, but I'm pretty certain it won't improve
it.

d

--
Pearce Consulting
http://www.pearce.uk.com

[Patrick Turner]

........
The small amount of declining signal above 7kHz means very little
current needs to be fed
to the low Z at HF, and a 20 watts amp is enough, and there is no need
to have 312
watts to give 25vrms across 2 ohms at 18kHz.

Possibly true, but it does depend on music type. I suspect Quad made
their assumptions on the basis of vinyl source, and no
hypercompression. Things are no longer that way.

Actually, correct, because when I measure busy modern music
with my 33 band band pass filter there is considerable energies
at 10Khz, as well as at 1kHz, and 100Hz.



If you are comparing loudness with various signals, ignore peak values
and go for RMS - this is the simplest, and probably the best way of
doing things. Use peak values for nothing more than making sure you
aren't clipping.

Yes, but music has voltages and F all over the place at different F
like pink noise, and music has less F than noise, and the the F of music
is
dominated by strong bass tones, drums, with harmonically related tones
riding along on the crests of the bass tones.

So we do need to keep an eye on peak voltage abilities of all gear used
in a line up.


I don't know what F is here - do you mean frequencies? If so, I don't
understand what you mean when you say music has less frequencies than
noise?

Noise is a conglomeration of many randon F changeing all the time,
but with most music, the tones of dominant instruments or voices
appear as sine waves with many other harmonics, and obviously music is
more
selective of its tones because of the harmonically related F.
The exception is heavy Metal, or Dark Metal,
where the music resembles a continual sound of 747 planes crashing on
your house,
and young gits screaming thie lungs out buried in this crap.




Most cheap multimeters don't measure RMS voltage. They measure peak
voltage with a rectifier, then scale the meter reading to give you the
RMS equivalent for a sine wave.

Yes, 0.707 x peak voltage = Vrms.

For sine waves only, of course.

If your signal is something other than
a sine wave (eg pink noise) you can expect your multimeter to read
incorrectly.

With my Fluke, and another cheaper DVM, when max peak noise was deemed
to be 56V,
I got a hovering 12Vac measurement on the DVMs.

Spend a bit more for one that does the maths properly and
will give you true RMS measurements at many different crest factors.

I will search for one, or build one.


If you have a Fluke, it probably already does a good job in this
regard.

Well maybe it does. But BW is 20 to 2kHz only.

The other very cheap meter also reads similarly.




Finally, who really cares about speaker sensitivity? These days it is
simple to get whatever amplifier power you need, so just bolt together
whatever gives you the sound level you need and don't worry about it.

Hang on a minute, these ESL need a maximum of 11.2vrms to give very tame
levels of music.

I have no idea what that means.

My own dynamic speakers sound fine with busy pop music
when my Fluke says about 0.9Vac, so peaks are much higher.
The ESL need 3.6Vac to give the same SPL, measured the same way.
Many ppl would find the levels a but tame.

0.9Vrms into 5.6 ohms is only 0.144 watts, enough for me for
average levels with two speakers. But other folks would want
1.44 watts, and 2.88Vrms is needed, and the ESLs would require
11.52Vrms average and peaks will be lot higher....



To give any sort of performance these ESL require an amp to be able to
make 44Vrms max
to give a comfortable ceiling. amps are usually rated for 8ohms, and
44Vrms into 8 = 242 watts,
so a "300 watt amp" would be ideal.


OK...

If an amp is capable of 44Vrms for the maximum Z of 20 ohms of these ESL
as I have set them up,
then its 100watts, but that's only 28V into 8 ohms, so most 100 watt
amps won't be good enough.

It is as I said - it is the volts that matter. Amplifiers supply
volts, speakers consume watts. If your speakers are typically of a
high impedance, then they may indeed need quite a few volts to produce
those watts.

Tube amps allow you to match the load somewhat; Quad-II
has 16 and 8 ohms matching to get the best voltage x current outcomes.

The trouble with ESL is that below 1kHz they need lots of V and low I,
and above 1khz,
one needs little V and lots of I, so an amp set up
to give lots of voltage at low current into
16 ohms must be able to make much higher currents for where Z is low.
But as i said the load I change at HF is not overwhelming because the HF
energy is low compared to
the LF, and the amp has reserves; ie, it normally runs average power
which is quite low.


A solution is to play with impedance matching to by using an impedance
matching tranny
to reduce the ESL average Z from say 14 ohms to 7 ohms,
thus allowing better use of the 100 watt amps power band.

A friend here has a pair of such Z matching toroidal trannies which I
placed into wooden boxes

Toroidal trannies wouldn't be my first choice for audio power. They
have no air gap, and so tend to be quite non-linear.

which gave him 20:20, 20:10, 20:8, 20:4, etc, and
BW I measured was 4 Hz to 1MHz, and virtually no THD.

How did you measure this? This is indeed a most remarkable tranny if
it can do this.

I forget the exact brand and URL for them but there were a considerable
number of taps
to choose from.

Anyone could easily wind such a tranny.
You just need a 500VA rated mains toroidal core, use about 1.5mm wire,
and wind about
70 turns around to get one layer of wire.
about 4 such layers will fit on, with suitable taps.
The tranny can be set up any way you like including as an autoo
transfromer
which is the most efficient and when 1:2 ratio is chosen, its 1:4 Z
ratio, so 4:16,
or 8:32.

The formulas for more exact calculations of the required turn count to
avoid saturation
is all at my website.



he finds it useful to transform a modern speaker's common low Z to a
high Z
more suited to the tube amps he's using, therefore reducing the
effective Rout of the amp, improving damping,
and reducing THD/IMD by a factor of 0.25.
I forget where he bought them in the US somewhere, but they did the
business really well.

Patrick Turner.

If you use this on your amp to make a better impedance match, it will
reduce the damping factor, not improve it. As for THD - what it does
to that is anybody's guess, but I'm pretty certain it won't improve
it.

A step down tranny after any amp provides the amp with a higher ohm load
if the same load is transfered to the secondary.
A 2:1 tranny makes 4 ohms feel like 16 ohms to the amp.

So if the Rout was 2ohms at the amp, it becomes 0.5 ohms at the sec.
if the load is 4 ohms at the sec, the DF is much better.
Of course max power into 4 ohms is reduced to what 16 ohms would get at
the amp
without the tranny.

Amps working heavily into mainly class B see a higher load ohms so their
gain rises
and hence the amount of NFB rises, and THD/IMD falls.

Class AB tube amps with THD of say 1% at Xwatts in AB will thus produce
maybe
0.25% for the same power because they get to see a better load for
lineararity,
and NFB is increased and the amp moves more into pure class A, so the
add on tranny
isn't a curse, but a profound blessing to overcome the errors of
designers who
tend to always use too low a load for the tubes.

But with SS, you may be able to bias a 300 watt class B amp with high
bias current,
if the heatsink permits, and
connect an output step down tranny, and you can get 20 watts of class A
easily,
but maybe only 50 watts max.
Those first 20 watts can be virtually distortion free.
The amp load can be 40 ohms.
44Vrms into 40 ohms = 48 watts.
if the load is 4 ohms, then ZR = 10:1 so TR = 3.2:1.

The damping factor is determined by the tranny winding resistances,
because SS amps with all that NFB have Rout ty[ically 0.1 ohms
at the amp output terminals after the LR zobel network.

Patrick Turner.






d

--
Pearce Consulting
http://www.pearce.uk.com

[Don Pearce]

On Tue, 29 May 2007 16:22:18 GMT, Patrick Turner
wrote:



Noise is a conglomeration of many randon F changeing all the time,
but with most music, the tones of dominant instruments or voices
appear as sine waves with many other harmonics, and obviously music is
more
selective of its tones because of the harmonically related F.
The exception is heavy Metal, or Dark Metal,
where the music resembles a continual sound of 747 planes crashing on
your house,
and young gits screaming thie lungs out buried in this crap.


No, I wanted to know what F is. I only know it as Farads - the unit of
capacitance. If you are making up your own abbreviations, I get a bit
lost.




Most cheap multimeters don't measure RMS voltage. They measure peak
voltage with a rectifier, then scale the meter reading to give you the
RMS equivalent for a sine wave.

Yes, 0.707 x peak voltage = Vrms.

For sine waves only, of course.

If your signal is something other than
a sine wave (eg pink noise) you can expect your multimeter to read
incorrectly.

With my Fluke, and another cheaper DVM, when max peak noise was deemed
to be 56V,
I got a hovering 12Vac measurement on the DVMs.

Spend a bit more for one that does the maths properly and
will give you true RMS measurements at many different crest factors.

I will search for one, or build one.


If you have a Fluke, it probably already does a good job in this
regard.

Well maybe it does. But BW is 20 to 2kHz only.


That shouldn't be too far in error with pink noise.

The other very cheap meter also reads similarly.




Finally, who really cares about speaker sensitivity? These days it is
simple to get whatever amplifier power you need, so just bolt together
whatever gives you the sound level you need and don't worry about it.

Hang on a minute, these ESL need a maximum of 11.2vrms to give very tame
levels of music.

I have no idea what that means.

My own dynamic speakers sound fine with busy pop music
when my Fluke says about 0.9Vac, so peaks are much higher.
The ESL need 3.6Vac to give the same SPL, measured the same way.
Many ppl would find the levels a but tame.

0.9Vrms into 5.6 ohms is only 0.144 watts, enough for me for
average levels with two speakers. But other folks would want
1.44 watts, and 2.88Vrms is needed, and the ESLs would require
11.52Vrms average and peaks will be lot higher....


You are giving some very precise values for what appear to be highly
approximate quantities, and you are thus still confusing me.


It is as I said - it is the volts that matter. Amplifiers supply
volts, speakers consume watts. If your speakers are typically of a
high impedance, then they may indeed need quite a few volts to produce
those watts.

Tube amps allow you to match the load somewhat; Quad-II
has 16 and 8 ohms matching to get the best voltage x current outcomes.


That is a failing of tube amps, not a quality. Speakers are designed
to be driven by a voltage source. If you are going to approach a
matched condition you can expect major frequency response errors as a
result.

The trouble with ESL is that below 1kHz they need lots of V and low I,
and above 1khz,
one needs little V and lots of I, so an amp set up
to give lots of voltage at low current into
16 ohms must be able to make much higher currents for where Z is low.
But as i said the load I change at HF is not overwhelming because the HF
energy is low compared to
the LF, and the amp has reserves; ie, it normally runs average power
which is quite low.

No, you have this wrong - or I hope you do. If your esl's are designed
properly they have exactly the same voltage requirements at low and
high frequency. The current requirements will differ depending on the
impedance.


A solution is to play with impedance matching to by using an impedance
matching tranny
to reduce the ESL average Z from say 14 ohms to 7 ohms,
thus allowing better use of the 100 watt amps power band.

A friend here has a pair of such Z matching toroidal trannies which I
placed into wooden boxes

Toroidal trannies wouldn't be my first choice for audio power. They
have no air gap, and so tend to be quite non-linear.

which gave him 20:20, 20:10, 20:8, 20:4, etc, and
BW I measured was 4 Hz to 1MHz, and virtually no THD.

How did you measure this? This is indeed a most remarkable tranny if
it can do this.

I forget the exact brand and URL for them but there were a considerable
number of taps
to choose from.

I find the frequency response highly improbable.

Anyone could easily wind such a tranny.
You just need a 500VA rated mains toroidal core, use about 1.5mm wire,
and wind about
70 turns around to get one layer of wire.
about 4 such layers will fit on, with suitable taps.
The tranny can be set up any way you like including as an autoo
transfromer
which is the most efficient and when 1:2 ratio is chosen, its 1:4 Z
ratio, so 4:16,
or 8:32.

The formulas for more exact calculations of the required turn count to
avoid saturation
is all at my website.



he finds it useful to transform a modern speaker's common low Z to a
high Z
more suited to the tube amps he's using, therefore reducing the
effective Rout of the amp, improving damping,
and reducing THD/IMD by a factor of 0.25.
I forget where he bought them in the US somewhere, but they did the
business really well.

Patrick Turner.

If you use this on your amp to make a better impedance match, it will
reduce the damping factor, not improve it. As for THD - what it does
to that is anybody's guess, but I'm pretty certain it won't improve
it.

A step down tranny after any amp provides the amp with a higher ohm load
if the same load is transfered to the secondary.
A 2:1 tranny makes 4 ohms feel like 16 ohms to the amp.

But you will be using it in step up mode to increase the voltage to
suit your high impedance speakers - thus damping factor is worsened.

So if the Rout was 2ohms at the amp, it becomes 0.5 ohms at the sec.
if the load is 4 ohms at the sec, the DF is much better.
Of course max power into 4 ohms is reduced to what 16 ohms would get at
the amp
without the tranny.

Amps working heavily into mainly class B see a higher load ohms so their
gain rises
and hence the amount of NFB rises, and THD/IMD falls.

No, the gain doesn't rise, although the voltage might - this is a
power stage, remember. And why would the NFB rise? That makes no
sense.

Class AB tube amps with THD of say 1% at Xwatts in AB will thus produce
maybe
0.25% for the same power because they get to see a better load for
lineararity,
and NFB is increased and the amp moves more into pure class A, so the
add on tranny
isn't a curse, but a profound blessing to overcome the errors of
designers who
tend to always use too low a load for the tubes.

Most tube amp designers will just buy an output tranny designed for
their choice of valve, and impedance of speaker. Why would you say
they go too low? And how does this move the amp into class A? This is
getting frankly bizarre.

But with SS, you may be able to bias a 300 watt class B amp with high
bias current,
if the heatsink permits, and
connect an output step down tranny, and you can get 20 watts of class A
easily,
but maybe only 50 watts max.
Those first 20 watts can be virtually distortion free.
The amp load can be 40 ohms.
44Vrms into 40 ohms = 48 watts.
if the load is 4 ohms, then ZR = 10:1 so TR = 3.2:1.

The damping factor is determined by the tranny winding resistances,
because SS amps with all that NFB have Rout ty[ically 0.1 ohms
at the amp output terminals after the LR zobel network.


Only if you have a particularly thin piece of wire in the tranny. Make
it with decent stuff, and the damping factor will be determined by the
amplifier output resistance and turns ratio. This is not a competent
transformer design you are describing here.

d

--
Pearce Consulting
http://www.pearce.uk.com

[Phil Allison]

"Don Pearce"


Most cheap multimeters don't measure RMS voltage. They measure peak
voltage with a rectifier, then scale the meter reading to give you the
RMS equivalent for a sine wave.


** Most analogue and DMMs measure the average rectified value of the voltage
and then scale that up by 1.11 times.

Gives the right answer with sine waves and can be corrected for most other
regular wave shapes.


Finally, who really cares about speaker sensitivity?


** Just as important as it ever was.


These days it is
simple to get whatever amplifier power you need,


** Shame if the speaker is an ESL and has strict input voltage limits.

Shame the Pearce fool did not read the damn heading !!



....... Phil

[Phil Allison]

"Don Pearce"


Toroidal trannies wouldn't be my first choice for audio power. They
have no air gap, and so tend to be quite non-linear.


** What absolute ******** !!!!

Toroidals ( and C-core ) transformer have miniscule magnetising current are
therefore the most linear kind for use in audio.

Don Pearce is an UTTER ASS !!!!




........ Phil