Towards Perfection - Building a Better Capacitor
With improvements in speaker drivers, triodes, and audio transformers, the strongest remaining coloration in an artisan-quality high-fidelity system can be the capacitors. I know from experience that most of the prejudice against passive crossovers in loudspeakers is actually the sound of mediocre caps in a tweeter circuit - upgrade to a top-grade part, and the sound of the crossover mostly disappears (assuming it's correctly designed in the first place). The problem with coupling caps in vacuum-tube amps is worse, since cap-coloration affects the entire bandwidth, and may be level-dependent (some caps are preferred for low-level circuits, some for high-level, and some for speaker crossovers).
The "why" of capacitor sonics is no longer the guessing game it was thirty years ago. A prime offender is Dielectric Absorption (DA), which is complex and nonlinear electrochemical/electrostatic energy storage mechanism. (Visit A Chemist's View of Dielectric Absorption on the Audience Auricap page for an interesting discussion.)
Another serious problem is microphonics - caps act as low-grade, very resonant condenser microphones that also self-excite due to magnetostriction and electrostatic effects. Matt Kamna measured the "microphone response" of caps with spectrum analyzer and a loudspeaker three feet away and it wasn't a pretty picture. The response curve is about as smooth as a 2" computer speaker, with many peaks in the midrange.
OK then, what would a "perfect cap" look like? How about a film-and-foil Teflon cap encased in a brick of transformer wax - with the wax carefully spaced away from the lead-in wires. As mentioned on the Auricap home page, Teflon has the lowest Dielectric Absorption of any solid material - and DA is clearly audible in a high-quality system.
The surrounding wax brick would provide a mechanical absorber that also has favorable dielectric qualities (although not quite as good as Teflon, which is why you'd want to air-space the lead-in wires). For convenience in chassis mounting, the four corners of the brick could be pre-drilled to accept Teflon bolts and stand-offs for chassis mounting - no point in degrading the performance of the super cap by wrapping a metal strap around it.
Not that a cap like this is for sale, of course, especially in the range of 1 to 10uF, which would be useful for speaker crossovers as well as vacuum-tube coupling caps. I can only guess how big a brick would be for a 630V cap - maybe as big as a real brick - and a lot more expensive.
But thinking a little more deeply about how film and foil works, there are additional problems that aren't immediately apparent. Let's take a mental ride on an electron-microscope, and zoom in on the metal-to-film interface. This is the region where "the rubber hits the road", right at the working interface of the dielectric.
No matter how tight the cap is wound, you can see the metal is not really touching the Teflon, except in a very few spots. At the molecular level, what separates the metal atoms from the Teflon molecules is air - mostly O2 and N2. We don't need to worry about the N2, but the O2 is another story, since it will corrode the exposed metal surface.
Metallic oxides create serious electrical problems. Aluminum is notoriously easy to oxidize, acquiring a film minutes after it's removed from an acid-etch cleaning bath. Copper oxidizes more slowly, but just as surely. Silver also oxidizes slowly.
I'm not an expert, but I know that aluminum oxide is considered a near-perfect insulator, which is why it is the effective dielectric in electrolytic capacitors (the wet electrolyte transports the charge electrochemically). Copper oxide is a semiconductor diode, and was actually used in (inefficient) copper-oxide power rectifiers in the Twenties. Silver oxide I know less about, but it is classed as a conductor, not a semiconductor.
What's awkward about these metallic oxides is that they lie between the intended dielectric - Teflon - and the mostly pure metal carrying the charge. For a little while, a few areas of the metal foil may be free of oxidization, but it won't take long for oxygen to creep between the windings of the capacitor and gradually oxidize all of the metal foil surface area. To get the process started, it only takes an oxide layer one molecule thick.
Not only that, the voltage breakdown of the metal-oxide layer is going to be different in different areas of the metal foil, depending on how thick the oxidization is. Voltage breakdown is something we care about, since the cap may or may not be used in a dry circuit (not enough current to "punch-through" an oxide layer).
If punch-through does occur, it certainly won't be all over the surface area of the metallic foil. It'll only be in pinpoint arc-over regions, where the oxidization is thinnest and current flow the heaviest. Since the oxide layer is extremely thin, the voltage breakdown potential will be low as well. (This isn't the same as voltage breakdown of the intended dielectric, which happens at much higher voltages. This is breakdown of the corrosion layer.)
Although surface corrosion on wires and transformer windings is troublesome, it can be dealt with by adding an enamel or urethane coating on the wire as soon as it emerges from the wire-drawing machine or an acid bath. (You can see how surface corrosion is an intractable problem on bare-copper stranded wire. This is the great advantage of Litz wire, which isolates each strand in its own insulated coating.)
This approach doesn't work as well for metal foils intended for capacitors. An additional protective surface has dielectric properties of its own, and certain to be less favorable than polypropylene or Teflon.
One possible technique - not easy to visualize - is to acid-clean the metal foil, keep it in nitrogen gas, wind it into a cap in a pure-nitrogen atmosphere, and hermetically seal the capacitor, taking special care to avoid oxygen incursion along the lead-in wires. (Glass or ceramic encapsulation in a nitrogen atmosphere?) This sounds pretty exotic, although it might be routine for aerospace-grade parts, where outgassing and chemical migrations over time are a serious problem.
So in addition to the known problems with coloration from Dielectric Absorption in the bulk dielectric, there are problems with thin-layer corrosion on the metal-foil surface. Most plastics are nowhere close to gas-tight, especially along the vulnerable lead-in area. I'm pretty sure corrosion begins within minutes of the cap being assembled - if the metal isn't already corroded by the time it's used to assemble the capacitor.
I'm getting suspicious of subjective "break-in" that take days or weeks to stabilize. Anything that takes this long strongly suggests electrochemical reactions, although relaxation and re-orientation of the grain structure might be possible as well. For transformers, the cause is fairly obvious: the stress of winding the wire fractures the grain structure of the wire at each bend around the winding former, and it takes weeks or months for the winding stresses to be relieved. This is where cryogenic treatment might be useful, making the tedious process of break-in unnecessary.
There could be stresses created by the high-tension foil-wrapping process in a capacitor, and if this is present, cryogenic treatment can be useful. But it won't do much for corrosion on the surface of the metal.
Maybe - just maybe - the superior, more transparent sonics of silver come down to nothing more than a more benign corrosion layer than copper. Something to think about.
© Lynn Olson 2002. All Rights Reserved.