To begin with: It is well known that the step response of an ESL63 is not perfect.
It's only near perfect. Unfortunately, there are some and also publised anomalities of the decay after some 0.4ms or so. Would there be a way to smooth these very early jaggies?
And then, there are these ugly four vertical parallel aluminium profile bars, each one with it's own resonance. You may eventually try to damp this by applying a felt interlayer, touching both the individual bar and the outer, protective metal grid. I already have made such a mod on my ESL63, but that seemed not very efficient after all. As the grid itself is not acoustically clean and may rattle, only slightly damped by the sock.
So I experimented to reduce these two annoyances at the same time. Using bits of PET 24mm thick sound isolating panel material (e.g. Echoboard by echojazz.com), I squeezed and glued small bits between these bars. This material is like a very, very stiff felt and is mechanically stable. The size and the location of these bits have been determinded by a rather long iterative procedure.
Don't be upset while looking at the pictures because of the white colour of the dust protections you see. It's because I fitted my ESL63 with a Vliesseline Stickvlies (https://www.vlieseline.com/Produkte/Kreativsortiment/Stickvlies) as a dust protection instead of the usual BOPET. This thin and dense textile material showed a bit less reflections than BOPET in the wavelet analysis and also has some absorbtive/damping properties. The tradeoff is that this Vliesseline is not sealing, so that microdust is still potentially penetrating to the diaphragm. Therefore, this mod is only good for non-smokers.
On the frontside, from top to bottom, 20mm - 45mm - 55mm - 45mm - 20mm bits of Echoboard. The 3 central bits fit centrally over the panel's boarders.
On the backside, from top to bottom, 20mm - 70mm - 70mm - 20mm bits. The 2 central bits fit the center part of the two segmented panels.
The two native ESL63 showed a step of the SPL towards higher frequencies (the green curve labeled 0-0). The central and frontal 55mm Echopanel does nicely tame this step. A 70mm wide one did damp too much (the brown curve labeled 1_large-0), 45mm a wee bit not enough. Then came all the other pads, on the frontside, and on the backside, more or less positively influencing on the speaker response. And with all the pads squeezed between them, these metal rods were no more singing theirs own tune.
The final fittings and results were quite pleasing.
First of all, the nasty >>>ZOJINNNNGGG<<< metal rods now got absolutely mute. And the frequency response of the naked speakers got nicely smoothed, and finally approx. fitted into a +-2.5dB tolerance band above 1.5kHz (curves xy_AA). Refitting then the metal grid and the sock, things got worse again (curves xy_GS). The two naked (AA like Air&Air) and grid/socked (GS for Grid&Sock) AB-sibling's curves are artificially grouped together. In reality, the SPL curves of both speakers differ by approx. 1dB. By the way: a wavelet analyses shows that the grids are the meanest spoilers in terms of reflections. Then comes the dust cover(s), then the sock. And as the frequency domain correlates with the time domain, also the step response got a bit smoother, more uniform. It still presents the 0.4ms anomaly, however. I did not manage to get it smoother than that. But anyway, a bit of a progress is still a bit of a progess ...
A further, tour-de-force-option to improve on the step response might be to selectively obturate some parts of the different concentric stator aeras. I will not do that.
It's only near perfect. Unfortunately, there are some and also publised anomalities of the decay after some 0.4ms or so. Would there be a way to smooth these very early jaggies?
And then, there are these ugly four vertical parallel aluminium profile bars, each one with it's own resonance. You may eventually try to damp this by applying a felt interlayer, touching both the individual bar and the outer, protective metal grid. I already have made such a mod on my ESL63, but that seemed not very efficient after all. As the grid itself is not acoustically clean and may rattle, only slightly damped by the sock.
So I experimented to reduce these two annoyances at the same time. Using bits of PET 24mm thick sound isolating panel material (e.g. Echoboard by echojazz.com), I squeezed and glued small bits between these bars. This material is like a very, very stiff felt and is mechanically stable. The size and the location of these bits have been determinded by a rather long iterative procedure.
Don't be upset while looking at the pictures because of the white colour of the dust protections you see. It's because I fitted my ESL63 with a Vliesseline Stickvlies (https://www.vlieseline.com/Produkte/Kreativsortiment/Stickvlies) as a dust protection instead of the usual BOPET. This thin and dense textile material showed a bit less reflections than BOPET in the wavelet analysis and also has some absorbtive/damping properties. The tradeoff is that this Vliesseline is not sealing, so that microdust is still potentially penetrating to the diaphragm. Therefore, this mod is only good for non-smokers.
On the frontside, from top to bottom, 20mm - 45mm - 55mm - 45mm - 20mm bits of Echoboard. The 3 central bits fit centrally over the panel's boarders.
On the backside, from top to bottom, 20mm - 70mm - 70mm - 20mm bits. The 2 central bits fit the center part of the two segmented panels.
The two native ESL63 showed a step of the SPL towards higher frequencies (the green curve labeled 0-0). The central and frontal 55mm Echopanel does nicely tame this step. A 70mm wide one did damp too much (the brown curve labeled 1_large-0), 45mm a wee bit not enough. Then came all the other pads, on the frontside, and on the backside, more or less positively influencing on the speaker response. And with all the pads squeezed between them, these metal rods were no more singing theirs own tune.
The final fittings and results were quite pleasing.
First of all, the nasty >>>ZOJINNNNGGG<<< metal rods now got absolutely mute. And the frequency response of the naked speakers got nicely smoothed, and finally approx. fitted into a +-2.5dB tolerance band above 1.5kHz (curves xy_AA). Refitting then the metal grid and the sock, things got worse again (curves xy_GS). The two naked (AA like Air&Air) and grid/socked (GS for Grid&Sock) AB-sibling's curves are artificially grouped together. In reality, the SPL curves of both speakers differ by approx. 1dB. By the way: a wavelet analyses shows that the grids are the meanest spoilers in terms of reflections. Then comes the dust cover(s), then the sock. And as the frequency domain correlates with the time domain, also the step response got a bit smoother, more uniform. It still presents the 0.4ms anomaly, however. I did not manage to get it smoother than that. But anyway, a bit of a progress is still a bit of a progess ...
A further, tour-de-force-option to improve on the step response might be to selectively obturate some parts of the different concentric stator aeras. I will not do that.
Last edited:
The bars have a very low density polyfill on them that rests against the dust cover. There is one at the front and one at the back. If the front one is at the top, the back one is at the bottom and visa versa. It seemed to be a pretty short lived thing, but I come across them from time to time in my rebuilds.
Sheldon
Sheldon
Damping the dust cover?
On the picture it looks like the bars are folded to the other side, towards the panels, not towards the dust cover!
But a picture can be tricky.
Never seen them fitted in a set.
They are folded toward the panels, those vertical bars are outside the dust covers. I have a couple of them, I'll fit them on a quad while I'm reassembling and take a couple pictures.
Sheldon
Sheldon
The bars on the 63 are seldom seen. I might even say next to none.
The difference in sound between 63, 988 and 2805 is more in the type off grill used.
The 63 has a Venitian shades aluminium grill, where the 988, 2805 and 2812 have flat perforated steel grill.
On top of the perforated steel metal grill, the 2805 and 2812 also have a steel reinforcement bar directly in front of the center of the rings. Not too smart a construction if you ask me. The 988 construction was so flimsy that they tried to correct that with the steel bar.
I made a set of these "dust cover damping bars" with the 3D-printer.
Think they must have looked something like this.
Think they must have looked something like this.
Here's what they look like up close. I'm holding the part against where it would mount in a speaker frame with no panels or dust cover. On speakers with these, there are slits cut in the vertical support bars for these damping pad assemblies to slot into. This particular pair of quads is older than these dampers and does not have the slits.
Sheldon
Sheldon
+1, had a pair with these brackets on them - they weren't fixed like in the photos above, just located in the grooves and against the membrane tension. I removed them. I think the poles just add a small amount of stability - I have had some hex grilles made up ( 82% open) so use some thin edpm foam to stop rattles.
Joe
Joe
Removing the metal grids really improved my step response. I have also removed dust cover... and since I do my own refurbish with new membranes and coatings I took the chance. It´s almost 20 years ago and they still play... so no need for dustcovers i think. Depends on dustlevel i guess. But the cloth seems to keep the dust away. Maybe we can try to dip the cloth in antistatic fluid and insulate the cloth, charging it up to 10kV and let the cloth be the dust absorber?
Would love to see a before and after graph of the measurement results and see the improvements.
Here are some before and after differences on the step of the Quad ESL-63 with some mechanical changes to the panels made.
No damping material added, just the standard mesh glued to the rear stator. These measurements are done with dust covers, grilles and speaker cloth in place.
In the overlay the yellow line is before, the red line is after. The line smoothens out quite a lot after the changes.
Indeed the step is never perfect for any speaker, but changes can be made. Are they improvements?
Also there is a difference in the frequency response. Again yellow before change, red after.
Aligned the frequency response @ 1 Khz where the measurement volume is set @ 75 dB @ 1 meter on axis.
No damping material added, just the standard mesh glued to the rear stator. These measurements are done with dust covers, grilles and speaker cloth in place.
In the overlay the yellow line is before, the red line is after. The line smoothens out quite a lot after the changes.
Indeed the step is never perfect for any speaker, but changes can be made. Are they improvements?
Also there is a difference in the frequency response. Again yellow before change, red after.
Aligned the frequency response @ 1 Khz where the measurement volume is set @ 75 dB @ 1 meter on axis.
Last edited: