Chris, Mal,
This is interesting....... Let me give a few of the design considerations here. I don't want to rain on anyone's parade, not at all, but some might be interested in the design brief on precisely this aspect of the GK-1.
There are three principle signal caps in the GK-1 - this number being a reflection of the tube circuitry used. They are the input cap to the preamp, C1 (MMK, Evox-Rifa), the input cap to the tube grid, C19 (Suflex, polystyrene), and the output coupler C21 (polypropylene/ps composite, 1uF Auricap).
These three caps were the subject of high anxiety discussion (HAD - now you know all about Cary) between Darl and me.
One of these caps has essentially no DC bias across it, C1. It has about 150-200mV, that's all. With no DC bias across a cap to stress the dielectric, it's tough going to extract good sonics, because the extreme dielectric plasticity at low voltages is rather non-linear, giving higher distortion (read 'mud'). With strain on the dielectric due to appreciable DC bias, the mechanical behaviour of the plastic film is more linear, leading to less distortion, and most of it at a lower order, particularly at low AC levels. Generally speaking, an AC signal of 1Vrms passing through a cap with 0.1VDC across it will not sound as good as the same AC signal through the same cap with 10VDC across it. With music, of course, H2 and H3 are tolerable because they are relatively euphonic, and accord well with our intrinsic notions of musical harmony. H5, H6 and higher sound pretty non-musical, with the higher odd-orders sounding appalling, and at low levels below conscious perception, giving rise to marked listener fatigue.
Thus C1 had to perform well at low bias voltages. We tested a large number of caps and found that only the RTX polystyrene composite caps performed better than the Evox-Rifa in this role. Given the cost of these beasts, we decided to opt for the vastly less expensive Evox-Rifa. We also tried the Auricaps, but for some reason, with almost no DC bias across them, found they sounded dull and lifeless, a situation which did not prevail for C21, the output coupler role. Now, YMMV, and I'm not going to disagree with anyone because when it comes down to it this is a highly subjective opinion. Remember, I'm the good guy navigator, not the maniac fighter pilot!
C19 was different. Here there is about 45V across the cap, and thus the constraints are relaxed. However, this location calls for a very small capacitor, because the input impedance of the tube is very high, more than 1M. Knowing my dielectric theory, I opted for an industrial polystyrene, hard to source, but very, very good sonically. The cap need only be 10nF, with maximum size not exceeding 100nF, but polystyrene is probably the best dielectric shy of silver mica so it was our first choice, and remains so. I don't recall if Darl tried many others here, but my own experience over years tells me it's very difficult to best a tubular polystyrene sonically - just don't overheat them while soldering!
The output cap, a large value of 1uF, was a whole new ball game. With almost 50 volts across it, we needed something pretty damn good. Once again, the RTX tested very good but considering it's cost at this value we settled on the Auricap, which was significantly cheaper, available in 200V rating, and sounded very good indeed.
Knowing that the Auricap is not as good as the RTX, however, it does not surprise me that Chris found a little ps bypassing improved it. The GK-1 is redolent with bypass caps; C6, C7 and C18 all carry bypasses for precisely the reasons Ginger has explained. Not all caps are bypassed; there are always practical limits, and diminishing benefits, but the careful design of this preamp according to the conservative rules of engineering has definitely paid dividends. There is no magic in this design; just great care in component choice, operating points and implementation, and the wonderful sounds it makes are a result of plain vanilla engineering done right, a bit like a Chevy V8.
Chris, I despatched your AKSonics yesterday; you should have them within two weeks at the most.
Guys, thanks for all the interest and discussion. And a big thanks to Chris, Peter, Mal and Ginger for their valued input. Long may it continue!
Cheers,
Hugh
