Markus,
This is what I wrote about it:
I spent three hours today looking over the shoulder of a skillful repairman as he dissected a Quad 2905. It takes about 20 minutes to get the speaker stripped down to the point where you can see and test the panels. There are six separate panels, two bass at the bottom, two mid/high range in the middle, and two more bass panels at the top. For those who don't know, each panel is a sandwich, with a plastic grid of 2cm squares on the outside front and back. Inside this is a copper panel perforated by a grid of small circular holes, resting against a phenolic board like a PCB (again, front and rear). Between the phenolic boards is a Mylar membrane, glued at the edge to the rear plastic grid's outer edge. There is a 2-3mm gap between the Mylar and the PCB in front of, and behind it. The Mylar has a graphite coating all over except for the very edges. High voltage is fed into the copper stator on each side of the Mylar, and the graphite responds by by being drawn to or repelled from the stators. The stators are solid single pieces of copper for the bass panels as low sounds are non-directional. The mid range/treble panels have the classic Quad arrangements of concentric circles of the copper mesh, each part circle connected separately through delay lines so that the two treble panels act like one large concave/convex panel, simulating a point source.
All the delay lines must be unsoldered to remove panels, and reconnected afterwards.
My faulty panel, this time, was the lowest bass panel, and the phenolic board at the front of the sandwich with the copper stator on it was detached from the plastic frame at one side. Originally, it was glued in place. Careful probing with an insulated tool makes the loose part of the stator arc noisily as it is pushed towards the Mylar. So that panel was removed and replaced with a new one ordered from Mo-Fi some weeks ago.
However, the four original panels in that speaker (two now being replaced with new ones) all have the same problem developing, and all on the same side of the front face. They have come loose from the frame, but are not yet loose enough to cause arcing. Comparing old and new panels one can see some differences. The glue on the originals is slopped around a bit, but is extremely sparse or absent where the phenolic board is loose. The new panels have a visible bead of glue connecting the stator to the plastic grid.
So we scratched our heads for a bit, and discussed various glues. We don't want water based glue, and we don't want it to penetrate through the phenolic board. We decided to place a blob of industrial hot glue at each cross of the grid, and then used a dental pick to pull the copper stator back up against the glue as it set. We used one of the old panels as a test first to ensure the Mylar would not be hurt by the heat of the glue - it wasn't. Then I remembered the roasting bags for poultry, which I think are made of Mylar, so it should resist heat OK. At the end, all the stators on the front are well-affixed to their frame. None of the stators on the back of the panels were loose at all. Looks like a poor manufacturing process led to insufficient glue, and I'm guessing Quad became aware of it since the new panels have lots more glue on them. Then the speaker was reassembled and we set off home.
It works just fine now. So what was the point of the hot glue? As the end of the stator starts to flap in the breeze it gradually detaches for more of its length, and at some point it can move enough to touch the Mylar and it arcs. After a short (pun intended) while there is a hole burned right through the Mylar. By fixing the stators firmly with glue I hope we have reset the clock on those panels, or at least at that end of them, as the other end had no looseness, and nor did the rear stator as I said above. Time will tell.
As for doing it myself if needed, I think I can. There are airtight dust screens that look like Saran wrap on the front and rear of the speaker, to keep dust out (!) - because an accumulation of dust would increase the risk of arcing. The PCB with the delay coils and circuitry on it is coated in wax, probably beeswax, as it is outside of the sealed dust screens, and again any dust gathering between the terminals that the delay lines are soldered onto will encourage the terminals to arc. So after re-soldering the delay lines we dripped a beeswax candle onto the terminals. Not as pretty as the factory finish, but it should work.
Now the other speaker hasn't had a problem at all, so far. I could take it apart now and go at it with the glue gun if needed, but I think I'll wait until it starts to cause trouble, replace the panel that goes, and apply glue to all the others if showing signs of loosening. Fingers crossed, it won't happen...