Backstabbing Capacitor

[kevmurray]

I found an interesting problem today in a three way crossover. It was donated to me so I was doing sweeps to determine it's response (@8ohms). The load was purely resistive through the audio band so everything should have gone farely smoothly, but I noticed at around mid-band a ringing could be heard coming from the test setup. Like any good nerd I immediately assumed a coil was ringing, so I potted them by injecting low viscosity glue into the windings. Next day same thing, ringing! To find the problem I found the peak frequency (~3kHz) and turned up the voltage for a nice loud ring. Then I used my sound pressure meter to find the offending coil. Turns out the coils were silent, it was a cap that was singing! Little bastard. At only a few volts input it was puting out 94dB @1". I've attached the response plot and a photo of the pcb. The offending cap is the one in the top left (orange).

It's a cheapo unit of course but I've never heard of this happening at audio frequencies before. Anyone else have similar experiences?

[joecarrow]

That's interesting- what type of capacitor is it?

Electrostatic speakers are essentially large, flat capacitors that rely on the force between two charged bodies. The dielectric in a regular capacitor should keep this to a minimum- have you tried switching out the capacitor? It could be bad.

[kevmurray]

That's interesting- what type of capacitor is it?

Electrostatic speakers are essentially large, flat capacitors that rely on the force between two charged bodies. The dielectric in a regular capacitor should keep this to a minimum- have you tried switching out the capacitor? It could be bad.

It's a standard MKT type found in power supplies, 4.7uF. I doubt I have a replacement but here is a sweep after removing the cap. It was in the high pass circuit so the midrange still works. The high pass circuit was fed from the mid which explains it's affect on the mid response. I'll find a replacement and post my results.

[JonP]

Interesting... probably the cap has a loose area in it's structure that can vibrate, and maybe it's a one off fault. That notch in the response is wierd though, I wonder if there's some resonant circuit in the crossover (that isn't supposed to be there) and a lot of energy is pumping back and forth between an inductor and that cap. A bad design?

Have you reverse engineered the schematic of that crossover?

Probably more likely a bum cap.

[kevmurray]

schematic

After looking at the plot again I think you may be onto something Jon. There are two peaks suggesting a critically coupled resonant pair, the mid circuit and another. It's not a pretty plot of one but I think that's what this might be. I'll double check my load resistor value on the HP and make sure all solder joints are sound. Here is the schematic.

[fjhuerta]

I never thought I'd find a damaged cap until yesterday... I was running sweeps, and indeed, the impedance plot looked completely resistive. It turned out the first capacitor on the tweeter circuit had more than double its real value. I replaced it and everything was fine.

It was brand new, too. A Dayton poly.

Moral of this story: measure twice, then build once.

[kevmurray]

...Moral of this story: measure twice, then build once.

Words to live by. I found the problem and it's my fault (as it so often is :Z). I was using two 4.3ohm resistors in series to make the load for the HP circuit and found that a bunch of 430k had found their way into the wrong bin. When replaced with an 8.5 (measured!) the circuit worked properly with the original cap reinstalled. I still find it amazing how loud that cap rang.
Basically everyone was right. The HP section was resonating with the BP section since it was essentially unloaded, the cap was probably acting like an electrostatic, and I had the wrong component values in circuit. ops:

I've attached a new plot of the response which includes each section and the reference level. Note the insertion loss in the LP section, presumably due to the resistance of the inductor. The BP section has less but it's still noticeable. Good support for active crossovers, at least for the lower frequencies where inductors are large.