Maximum Capacitance before Pffffft?

Excessive capacitance doesn't come up often with regular magnetic coil loudspeakers, but electrostatic panels have quite a bit of it. And in the case of the amp / loudspeaker set I'm hoping to build here, It can be quite a problem. So the question I'm asking is, how do you tell how much capacitance an amplifier design can drive, without building it and going the 'lets try and see' approach?

The amp design I'm looking at is basically a scaled-up version of Kevin Gilmore's ES headphone amplifier design at Headwise . The panels are going to be full-range and LARGE; as large as i can make them. It's direct-drive, so there's no transformer in there to load; it's (almost) the same as wiring a capacitor across the output of an amp and letting it rip. Of course, the panels' capacitance is defined by its area and stator spacing, so in order to get as large a panel as possible, I need to know how to determine an amp's maximum capacitance tolerance.

So how do you calculate the limit capacitance? What mechanism is at work here that eventually drive the amp to explode? As I understand it, it deals with driving an amp into UHF oscillation. Any experience of pointers on this topic would be great, and of course criticism and commentary are welcome. Thanks.
- Jonathan
 
You have to calculate the impedance at high frequencies. The larger the capacitance, the smaller impendance the amp sees at high frequencies. Z = 1/(C*2*pi*f)
C : Capacitance
pi : 2,1314...
f : frequency

That´s not the only factor that is important. Even more important is if the amp is stable in the firat place. A capacitance in the output can cause many older design amps to oscilate. This is not at all easy to calculate.
 
play with a prototype, adjust compensation

When I did my power-amp design, I wanted it to be stable with any capacitive load, it spite of the fact that it drives reasonably benign T+A T160E conventional speakers.

I used current limiting on the supply and fed a square wave band-limited to 2 MHz to the input and observed the output with a high-bandwidth scope. Then I loaded the output with various resistive and/or reactive loads and looked for the onset of oscillations. Then I played with the overall miller compensation as well as with various compenstion/decompensation nodes within the amp. I also adjusted the Zobel damper, a combination of a 470 nF WIMA MKS stacked film and 5 R (doing the tests with a square wave requires a 2 W resistor!) to ground. The output insulating coil was maintainted as small as possible to maintain damping factor. It consists of 5 air turns of 1.5 mm copper wire, 1 cm in diameter (stabilized with expoxy glue) and a 1 R resistor in parallel. It proved helpful to add 2,5 mm ferrite beads to the bases of the power transistors. Also, adding a 10 R resistance between the input filter capacitor and the base of the non-inverting input transistors was a great advancement.

The compensation tweaking process was very tedious but in the end I could maintain 50 MHz unity gain bandwidth (with the input filter removed), maintain a damping factor of 500 at 10 kHz into 4 R and still drive any capacitive load connected to the output, even if it was stacked film without a series resistor. With the 2 MHz band limiting, there was some ringing at the edges with loads above 50 nF but with a more realistic 80 kHz input filter this was gone completely.

This goes to say that with a good design and fast bipolar output transistors, you can drive any capacitive load. The other way to go would be to add a large output coil but then you would lose damping factor. I am not sure if electrostats require a low dynamic impendance drive.

Greetings,

Eric