Nelson Pass: Simple Sounds Better Page 3

Pass: The delay is more associated with internal testing and documentation than supply lead time in and of itself. One of the reasons for that is that, in almost all cases, considerably before the design has been frozen and gets documented, you know what the longest-lead-time parts are going to be, and you've settled on them: heatsinks, power-supply capacitors, transformers, semiconductors. You can pick up five to ten weeks. The time for us between the finish of a design and its actual release is about four months.

Norton: Tell us a little bit about your new Forté amplifiers, the 4 and 5.

Pass: They use a new output semiconductor called an IGBT—which stands for Insulated Gate Bipolar Transistor—which is a couple of years old as an n-channel switch device. Toshiba recently developed very linear complementary devices, specifically for audio. When we ran across those they looked extremely interesting; we acquired some and began working with them. I should point out too that I don't take the design credit on that particular product. The front end and some other pieces are mine, but the amplifiers are primarily the work of a Swedish engineer, Michael Bladelius—we brought him over a year ago.

The IGBT is a very interesting device. It's a hybrid, functionally between a MOSFET and a bipolar. It has more or less the input characteristics of a MOSFET and the output characteristics of a bipolar. Its input is characterized by transconductance—its transconductance is higher than you find normally with a MOSFET, but it's still a transconductance characteristic. It has a low output impedance, as is found in a bipolar, but wider bandwidth than you see with a bipolar. It's also quite linear, in this case sufficiently linear that the performance that we get on the Forté amplifiers is achieved without a feedback loop, much as in the Stasis amplifiers, but with a different approach.

Norton: So you've been able to reduce the feedback considerably compared with the earlier Forté amplifiers?

Pass: Well the Forté 1a and 3 had feedback loops around the amplifier; that is, the output of the amplifier went back to the input and was mixed with incoming signal. In these new amplifiers there is isolation between the input stages and the output stages, and each has far less influence on what the other might be doing. So they're independent. There's feedback around the front end itself, which sets the gain for the system, so there isn't any variation there. And there's a proprietary bias circuit which we have a patent on. Then just simply the IGBTs used as followers. We run them in parallel sets.

Norton: MOSFETS have been touted as having both tube-like characteristics and self-correcting thermal characteristics with no need for real thermal protection. Do these new devices fall into those categories?

Pass: They do have some secondary-breakdown characteristics so they're not completely as rugged as a MOSFET. There's a bipolar element to the power, and you do see some minor secondary-breakdown characteristics at high voltage levels, but it's nothing we have to worry about in devices in this power range—which is up to 200Wpc. They're quite rugged.

The thermal characteristics are quite similar to that of a MOSFET. That is to say, they're thermally stable. It's interesting to note that, contrary to what people believe, a MOSFET itself starts out at low-current with a positive temperature coefficient. It's only at the higher current levels that it starts exhibiting a temperature characteristic where the current decreases with temperature. So if you're biasing-up a MOSFET amplifier in class-AB, you actually find yourself with a wandering bias in much the same way you would with a bipolar. When you get them up to class-A or near class-A conditions, then you start seeing the real effect of that negative temperature coefficient.

With a hybrid device, one would expect, and one gets, hybrid performance. They have some of the [same] kind of tube characteristics or MOSFET characteristics that have been associated with devices which have transconductance characteristics. But they also exhibit more dynamic range, and greater punch than I've come to expect of MOSFET designs.

[I asked Nelson if he'd like to delve into other areas...]

Pass: The notion of simplicity ties in a bit to short- or long-lived trends—I don't want to call them fads—where the emphasis in electronics has been in one area over another. When Threshold first started out, back in the Phase Linear days, high power was the item that was going to deliver the performance that everybody really wanted. The emphasis was on high power. Then, high power and very little distortion, so that static figures—getting those double-0 distortion specs—became very important to everybody. But people still weren't satisfied with that. "How come tubes still sound better?" was a very common refrain.

The focus then shifted to slew rate and TIM—low amounts of feedback and high-speed circuitry—the idea being that high-speed signals would somehow confuse an amplifier. People began building fast amplifiers and, lo and behold, quite a few of those fast amplifiers sounded significantly better. The interesting thing was that in order to achieve that higher speed, they had to make the circuits simpler. I don't think it was actually a cause-and-effect relationship. I think that for the most part the higher-speed circuits sounded better because it took simpler circuits to get high speed with stability.

However, there were some examples of very high-speed circuits out of some companies—who shall remain nameless—where they were doing 1000V/µs but everybody thought they sounded pretty bad. And tube circuits aren't so fast, especially after they get through the output transformer. The notion of measuring slew rate on a tube power amplifier doesn't make a lot of sense because you don't ever get to observe slew as such.

After that, though, the focus shifted to class-A and high current. High current seems to have been the real thing that a lot of people began buying by. In other words, they stopped worrying about what the distortion spec and the slew rate were and simply wanted to know how many amps it would put out. And we, along with other high-end manufacturers involved in the race, have been able to demonstrate how much high current we can deliver. I believe we are the champions of that, and I say that only because we've publicly demonstrated output currents on amplifiers in excess of 200 amps. I haven't seen a similar demonstration elsewhere. But in fact, we do get output currents for brief periods of time that will go up to those levels, with fairly low distortion. I've got some examples of amplifiers where the output current is ±100A sinewave into, say, a 0.1 ohm load, at about 1% distortion. Not bad at all. And there really isn't any limiting built into the amplifiers. In our latest brochure we published the curves of all of our amplifiers at 8, 4, 2, and 1 ohm, and they actually hold up pretty nicely.

There is some reason to do this. Some loads that are out there do go below an ohm. Quite a few of the electrostatic designs have been observed to hit points below an ohm. Some of the woofers from major manufacturers go down below an ohm. And when you start looking at that you can see that there's some merit to having that much current. You can also view it from the standpoint of the engineer's bias. The engineer always says, "Well, if an amplifier is going to be called upon to do say, 10 amperes, then let's just put a 10:1 margin on it." This is a very common thing. I've heard John Curl say it, I've heard other designers say it...Well, 10:1 is pretty good. That sort of thinking was being used also when people were dealing in slew rate. A 1000V/µs slew rate was, at one time, obviously state-of-the-art, but it was also being touted as "you need this!"