QUAD ESL63 parallel bars gymnastics

Never say never again.

Time ago Liberty Instrument IMP worked with a single impulse stimulus, with the option to average multiple runs in order to improve the S/N ratio by averaging the ambient noise. Sometimes one of these "averaging" pulses completely messed up the measurement, because this one single pulse had been overloaded by the neighbour's dog barking, or the neighbour himself slamming a door. Then came IMP/M with a MLS (minimum lenght sequence) hardware generator with a much higher stimulus energy and consequently with more robust and better S/N ratio results. And then came the sinesweep method along with capable soundcards, which made MLS/M obsolete.
I remember this quite well, not to mention when an airliner jet roared in the area. Still have a PC with Libinst installed, running under Win 98. Oh dear I really feel old...
 
  • Like
Reactions: jan.didden
All nice theory and simulations, but my practical questions remain:
What can I read in the graph aside from one looks better than the other? Is more specific info possible? Maybe not.
What is there to improve and how does that show? My only conclusion so far is the smoother the line, the better. Correct?
See my graphs in post #15 and #17
It looks like I have to look into Acourate as a replacement of REW. 😎
 
My only conclusion so far is the smoother the line, the better. Correct?
See my graphs in post #15 and #17

Yes. A smooth step response does correlate with a smooth frequency response. Therefore, working towards a smooth line in the step response is always a good option. Your Red curve shows both a smoother step response and a smoother frequency response at frequencies > 1kHz. But it's to take with caution: Two measurements might differ a bit one from the other, which also could account for some of the differences.

Your graphs basically show some anomalies around 2ms ... 3ms after the main step. Therefore in the physical construction of the panels you might track down any structures which could reflect or emit a second, overlaying sound artefact with a path related delay of lenght 2ms ... 3mm * 344m/s = 70cm ... 100cm. And there, there is a shift between your "standard" and your "improved" panel. Maybe you could disclose what exactly the constructive change was.

Instead, the quasi "ESL63-typical" unevenness at 0.4ms (lambda approx 17cm) might well stem from the metal sandwiched construction including severaly steps of radiation impedance changes from 2.5mm (membrane-stator distance) ... to a max. 14cm, the max. spacing of the protective metal grids.

While tweaking for the smoothest response, keep in mind that any jaggie, dip, ripple occuring at <5ms after the main step/pulse will not be perceived as an echo, but rather might change the tonal perception of the signal because of it's analogy in the frequency domain.
 
It looks like I have to look into Acourate as a replacement of REW.

Don't expect Acourate to be a replacement of REW. Because of both have their unique features and might be complementary depending on your intentions. Inversely, both lack nice features found in the other option. And both have theirs very specific awkwardnesses. REW is not better than Acourate, and Acourate is not better than RES. I use both.
 
  • Like
Reactions: jan.didden
I used to do impulse responses with MLS signals (both with old PC software called Loudspeaker Lab and also my own LabView measurement software), but I've switched to log swept sines. This little white paper from SRS talks about the advantages, but it comes down to lower crest factor:

https://thinksrs.com/downloads/pdfs/applicationnotes/SR1_SweptSine.pdf

I'll try to get you folks better measurements when I'm done with my next set of Quad 63's. I typically take my measurements at 48KHz so I can match legacy measurements. But I'll up it to 96khz and try to get you naked, and with dust cover, and also with dust cover, grilles, and grille cloth. I typically take measurements at 1 meter away.


Sheldon
 
  • Like
Reactions: jan.didden
I'll up it to 96khz and try to get you naked, and with dust cover, and also with dust cover, grilles, and grille cloth.

Thank you for your commitment, but jokingly, your announcement made me think of this mock advertisement for coffee. More seriously, such isolated measurements have been performed before several times. Therefore, they will not basically contribute to new findings, not even at higher sampling rates. There already is a consent that the mentionned structures may and also do alter the impulse response. So you may spare your time (and e.g. drink coffee instead).

As for the dust protecion and by the time of ESL63 development, a thin mylar-or-so film might have been the best compromize to protect against dust. Today, there are more choices. IMO it would be interesting to investigate on pet felt. Pet felt might protect at the same time against dust and would also provide some basic electric protection. In this logic and as mentionned before in this thread, I used sheets of Vliesseline as dust protectors which measured quite pleasingly.

1. Take e.g. a 2mm thick sheet of dense pet felt. It will be a bit reflective, but much more absorbtive. Therefore, you might expect some loss of SPL. I expect this absorbtion to be low, but evenly spread in frequency domain, easily compensated by some filtering of the input signal. Big plus of felt is the lack of resonances. Furthermore, if the felt would be attached immediately onto the panel's outside, then it also might partly dampen the resonances occuring inside of the panels assembly. Without any damping, the crate structure may act as an array of multiple open pipes.

2. Then instead, take a "traditional", tensioned, some micrometer thin film of mylar/pet. This film might be increasingly reflective at higher frequencies, causing resonances between the two parallel films. It might show a basic resonance when tensionned, and also might break up into submodes of resonances if driven slightly unevenly by the sound pressure. This might well be the case because of the stator delay structures of the ESL63. And it might produce some noise at low frequencies. Therefore, such a sheet might potentially add it's own erratic imprint to the signal.

3. Then omit all protections. You will get no additional reflections, nor damping. No damping of the inner panels resonances either. And you will have two electrical chairs in your living room.

The good, the bad and the ugly. Make your choice.

One question then remains: How good is the good, and how good will it remain also in long term? Maybe after another cup of coffe, I might set up a test frame to have some sine-sweeping fun for a while. If a material then emerges as a clear winner, comparative measurements against the "standard" solution would be most welcome.
 
Last edited:
Maybe you could disclose what exactly the constructive change was.
First, there are more and new clips to fit the 2 frame sides together. This makes the frames stiffer and hence reduces unwanted noise with (louder) audio signals and low frequencies that put strain on the frames. We put 12 new spring steel clips on every panel, 6 on every long side. In most cases there are very few clips and most have lost strength over time or are even broken.
Second, there are stiffeners in place. Think of changes made by SME owner Alastair Robertson-Aikman many years ago. Same idea, but only not obstructive like his very bulky and ugly version. With our stiffeners everything (dust panels, grills) still fits like with the original.
Third the frame is made more stiff. Again without being obstructive for original parts.
So on the outside everything looks the same like a standard ESL-63, on the inside not so much. Very important! This third mod is not done by attaching the aluminium L shape that holds the panels to the aluminium outside rails. That was the very first construction by Quad that was abandoned in favour of the floating frames on the 2 L strip in rubbers for less resonance.

sme.jpg

I made a simple video to show the difference in stiffness of the panels. Less movement means less distortion among other advantages.
 

Attachments

  • Rigid frame 480.mp4
    9.6 MB
Last edited:
Thank you for sharing and for your precisions. It's interesting to see that obviously for both of us the same flaws called for some rework.

Original:

Array_Harlequin_Front.jpg


Modified/while modified:

Interconnections_4.JPG



Interconnections_3.jpg


You may notice the stonger and multiplied clips and the reinforcing GFK rods for every panel. Another option is to damp the panel-to-panel coiled drive voltage connection, which in the original state goes <<<ZOINNNGGGGG>>> when mecanically/vibrationally stimulated. And there is a damping, fixing (electrically isolating) structure squeezed into the interspace of two panels (seen as the white band). This stiffens and damps the whole assembly at the same time, and also prevents some minor and eventually turbulent dipole sound pressure shunting between front and back which may occur in the originally free, three interpanel spaces of the completed assembly. Other mods I applied was to stiffen the frame with massive CFK rods and to attach the panels to the frame on a layer of felt. I also made some other, probably pactically less effective mods.

Now, when hitting/knocking onto the finished assembly, the ESL63 no more shreeks like an afraid piggy, but responds with a brfief <<<BOC>>> instead.
 
Last edited:
I'll try to get you folks better measurements when I'm done with my next set of Quad 63's. I typically take my measurements at 48KHz so I can match legacy measurements. But I'll up it to 96khz and try to get you naked, and with dust cover, and also with dust cover, grilles, and grille cloth. I typically take measurements at 1 meter away.

Thank you for your commitment, but jokingly, your announcement made me think of this mock advertisement for coffee. More seriously, such isolated measurements have been performed before several times. Therefore, they will not basically contribute to new findings, not even at higher sampling rates. There already is a consent that the mentionned structures may and also do alter the impulse response. So you may spare your time (and e.g. drink coffee instead).

Meanwhile I myself did drink several coffees and thus I had time to realize that my jokingly and critically meant reply might have had an unintended repellent, pejorative and/or even aggressive character, dependent on the way one reads and interpretes it. I would be sorry if it had offended Sheldon or any other reader. Around a table while drinking coffee together, exchanging directly and in real-time, this could have been cleared up immediately.

There is an imporant addendum I completely missed to point to: I think it would be very precious if anyone interested in the subject would be able to rely on a set of handily available, robust and standard, well done measurements, and eventually had access to the files also. This would allow further transformations for advanced analysis which possibly could lead to unexpected insights. And therefore I would welcome such a set of measurements, of course.

@ Sheldon: I really hope I did not upset you.
 
Last edited:
Thank you for sharing and for your precisions. It's interesting to see that obviously for both of us the same flaws called for some rework.

Original:

View attachment 1464070

Modified/while modified:

View attachment 1464073


View attachment 1464072

You may notice the stonger and multiplied clips and the reinforcing GFK rods for every panel. Another option is to damp the panel-to-panel coiled drive voltage connection, which in the original state goes <<<ZOINNNGGGGG>>> when mecanically/vibrationally stimulated. And there is a damping, fixing (electrically isolating) structure squeezed into the interspace of two panels (seen as the white band). This stiffens and damps the whole assembly at the same time, and also prevents some minor and eventually turbulent dipole sound pressure shunting between front and back which may occur in the originally free, three interpanel spaces of the completed assembly. Other mods I applied was to stiffen the frame with massive CFK rods and to attach the panels to the frame on a layer of felt. I also made some other, probably pactically less effective mods.

Now, when hitting/knocking onto the finished assembly, the ESL63 no more shreeks like an afraid piggy, but responds with a brfief <<<BOC>>> instead.
In the lower pic I see some wire ends stick out from the solder pad. Those are potential corona sources.
I always make a point of cutting them off as much as possible, and placing a nice round solder blob on them.

Jan
 
3. Then omit all protections. You will get no additional reflections, nor damping. No damping of the inner panels resonances either. And you will have two electrical chairs in your living room.
But you can agree that dozens of sockets in the living room will be "electric chairs" if you stick two fingers into them, here the question arises - is it necessary to do this, maybe there is another, more worthy use for our fingers and our hands?
A simple question - what should be done with the Kvod 63 speakers without protective panels so that you are killed by voltage?
Judging by the photos posted here, you need to take two pieces of bare wire with bent ends, stick them in simultaneously from the front and front sides so that from the inside they touch the copper foil of the fiberglass, turn on the signal to the speakers and ...
But this is much, much more difficult than just taking two pieces of wire and sticking them into the nearest socket? Maybe I missed something or misunderstood, it’s just that in the topics about electrostatics they constantly and persistently talk about “electric chairs”, but there are also electrostatic headphones, they generally need to be put on your head, and voluntarily, without an assistant.
 
Maybe I missed something

Yes, you are right. You missed at least two things:
  • Your point about the power sockets is flawed. If these are considered to be a risk, then you may easily eliminate this risk. Minimizing any risk is a good approach to more safety. Instead, introducing another extra and unneeded risk into one's living sphere is plain stupid.
  • The two lower picures show a state while work was in progress. As shown, the upper panel has no connection yet to the audio HV signal. Of course the final result is interconnected correctly.
 
Last edited:
In the lower pic I see some wire ends stick out from the solder pad. Those are potential corona sources.

Thank you for this one! This exactly is why this forum is so precious for me. Expose myself while expecting constructive critics. I will fix this flaw at the next opportunity, since I always had the impression that there was some faint, but distinctively eerie glow around me which I never was able to track down.

StElm.jpg


Apropos corona: You minimize the risk by rounding the geometry of the connection, which is a good approach. Could then eventually also an additional drop of epoxy be further helpful in terms of an eventual enhanced electrical isolation?
 
Last edited:
If you look at the electronics box at the bottom of an ESL63 you'll see that the boards are covered with what I think is wax.
Looks like they were immersed in a wax bath. That also cuts down on potential corona discharge.
Don't know what the best way is for a diy-er to apply that. Melt candles in your wife's fav cooking pan?

Jan
 
  • Like
Reactions: Daihedz