Rich May of Sumo: An Audio Dynasty Page 4

Atkinson: You mean an IC's characteristics change as the output current modulates its temperature.

May: Yeah. And this will show up at low frequencies. this is a thing which can get you into trouble playing a phonograph record with the inherent 6-cycle warp component. That can come right back and haunt you.

An excellent article by National Semiconductor points out that, hey, thermal limitations are the real problem in the amount of gain in IC op-amps. Not electrical problems. To give an example, the open-loop gain of some of their earlier ICs is considerably less than the gain of the identical configuration built with discrete components.

Atkinson: Despite being on a chip where you have much better opportunity to do some very careful matching?

May: You get better chance of match, you also have much better thermal coupling. And there are areas where you want the thermal coupling, other areas where you want to avoid it as much as possible. The Athena has a feedforward error-correction circuit used in the input stage. Very low noise, very low distortion. And without feedback and out in free air, you can watch the distortion modulated by air currents in the air conditioning. You can wave your hand and watch the distortion go up and down as the air currents change the temperatures of the input devices.

Atkinson: If you regard your circuits as always being the perfect circuits you work with in textbooks, you wouldn't dream of looking for these things. Do you think analogs of such problems exist in all components, including supposedly passive components such as resistors and capacitors? Do components tend to change what they are according to the level of the signal that you put into them?

May: Capacitors change, resistors change. It's not widely known that resistors have a thing called a voltage coefficient, where their value changes as a function of the applied voltage. In fact they're varistors. And they can introduce large amounts of third-harmonic distortion.

Atkinson: Again, if you try to measure these things by just using pure sinewaves, everything will appear to be really quite good. But when you process a music signal, that just isn't the case?

May: We're still at somewhat of a loss as to how the ear works. But maybe if we put in as many different sinewaves as we have in the music signal, we could find measurable differences. Right now, however, what are the most difficult test signals we use? Two-tone, either CCIF or SMPTE IM. And I don't think two tones are enough. Califone did some work for the cable TV industry some years ago, and it's really interesting how, when you start piling a number of signals into a given spectrum, the distortion components add up astronomically. There are amplifiers that measure very well with two tones that are just totally unusable with 30 or 40 different frequencies going into them, the IM products are so severe. Now we've got some tests we want to do; it's a matter of getting a few new things out in the marketplace, then going back and doing a little more R&D effort.

I would really like to have a better understanding of all the components that go into this whole chain. Some of the things we've done have an absolute correlation with measured observable phenomenon. Other things are instinctively right. You can't put it any other way.

Atkinson: Both the Sumo Andromeda and the class-A Nine Plus amplifiers are basically balanced designs, the load being hung between two output stages in push-pull. Are you looking at the possibility of fitting a preamplifier with differential outputs, like those on the Delilah active crossover, so that you can run the entire signal chain in balanced mode? What are the advantages of doing that?

May: One of the advantages, if the amplifier design is well carried out, is to make the amplifier quite immune to any noise picked up by the audio cable between the preamplifier and the power amplifier. And while we haven't achieved it, my Utopian goal is to make the link between the preamp and the power amplifier a true transmission line, so that all its characteristics fall out. It ceases to be a capacitive cable, it looks like a resistor at both ends.

Maintaining the amplifier differential all the way through does have some advantages. One of the problems with a conventional, single-ended, class-B amplifier is the garbage thrown off by the power supply wire. The interconnects between the filter capacitors and the output stage carry . . .

Atkinson: . . . halfwave-rectified audio signals!

May: Absolutely. It radiates all over the place. And how many amplifiers do you know that you can put on something as mundane as a bench and tweak the output or the power-supply wiring around and watch the THD move all over the place?