wtm@neoucom.UUCP (Bill Mayhew) (08/08/88)
For some curious reason capacitors seem to be a very hot religious issue among the audiophile community. Many series resistance and inductance issues that are described in teh audiophile trad with semi mystical verbiage can be analyzed quite resonably. In many cases, cause and effect can be related analytically. The biggest problem are electrlytic capacitors wich have insulating material which exhibits nonlinear frequency and voltage dependency. Obviously, most people would say to simply avoid the use of electrolytic capacitors. Fancy polypropylene and teflon dielectric material units would appear to be a solution. There is a bug that the required surface area of the plates in the capacitor made of the latter material is increased for a given capacitance. The surface area is increased beause the dielectric material is thicker, thus increasing the effective distance between the plates. Capacitance is proportional to surface area devided by distance. Unfortunately, the increased plate area increases the inductance of the unit, which is perhaps even less desirable than the nonlinearities of the electrolytic device. Last of all polypropylene capacitors tend to be viciously expensive. One intermediate soltuion is to place a small high performance capacitor in parallel with the larger non-ideal unit. The smaller capacitor fills in to work properly at the high frequencies where the electrolytic device does not function as desired. This still leavs the low voltage, low freqency non-idealness (is that a word?) of the electrolyic device unaddressed. It is interesting to read audiophile magazines that say that paralleling capacitors is a recently discovered audio panacea. In reality ham radio people learned the trick at least as long ago as the 1930s. One should make the logical jump that it is a good idea to eliminate capacitors from a circuit altogether. One of the more important applications (outside tone control circuits) for capacitors in audio amplifiers is in miller-effect feedback bypass circuits. Miller effect feedback involves placing a resistor in the emitter lead of a transistor to establish a certain DC operating characteristic effectively placing the transistor in a linear amplifying modality. The capacitor is placed in parallel with the resistor to supply correct AC coupling of the transistor. The problem is that the required AC and DC characteristics are often different, and thus the capacitor really is necessary. It possible to build a so-called direct coupled circuit that does not use any capacitors at all, but such a device tends to suffer markedly from temperature instibility, gain drift, etc. One really neat solution that I have seen is in a Fujitech A520 audio preamplifier that I assmebled (Fujitech is the kit division of Luxman, a japanese audio company). The Fjuitech amplifier used field effect transistors wired in a direct coupled fashion. Such a cuircuit would usually be unstable and/or have poor characteristics. What Luxman did was to use a servo circuit to replace the customary miller effect components. They use an operational amplifier to monitor and adjust the feedback. The operational amplifier is a two-pole / one-zero (if memory serves me right) that is a low-pass filter that cuts off at about 3 Hz. The servo circuit has very high gain at low frequencies and can very effectively control the bias point of the FET devices. The advantage here is that the operational amplifier in the servo circuit acts as a capacitance multiplier. This allows them to use a very modest physical size high quality capacitor and have it act as though it were a quite large uF value unit. I unscientifically auditioned my Fujitech kit A520 against a guy that owns a very exensive Audio Research SP-11. Both the AR owner and myself felt that the Fujitech kit was a fucntional equal of the SP-11. It definitely proves that in some cases, science can be the equal of magic. Incidentally, the Fujitech manual goes into great detail explaining how the "Duo Beta" servo works, including Bode plots and an S domain block diagram of the circuit. Unfortuately, it is all in japanese characters. about the only thing recoginzable in the text was "(1/1+Ba)". --Bill Bill Mayhew Electrical Engineering Northeastern Ohio Universities' College of Medicine Rootstown, OH 44272 (216) 325-2511