Nelson Pass: Circuit Topology and the End of Science

[Mr Spock]

Nelson Pass is a consummate engineer, but he got his start in physics, earning a bachelor's degree from UC Davis. As he worked on his degree, he was already an audio designer, focusing on loudspeakers—great training for a designer of audio amplifiers. Soon, in 1974, he cofounded Threshold Audio with René Besne, of audio and folk-dancing fame; their goal was to build electronics, partly because the field is less competitive—it's harder than building speakers. As he told Thomas J. Norton in "Simple Sounds Better," an interview in the November 1991 issue of Stereophile, Pass created one of the first high-power class-A amplifiers: the Threshold 800A. He'd dreamed up its key technical approaches in the back of a bus, on his way to visit a cave. Also in 1991 he founded Pass Labs, which focuses on high-power transistor designs, and in 2004 he started a second company, First Watt, to produce and market some of his more interesting experiments, mostly in low-power amplification, using an unusually wide variety of solid-state devices.
I wanted to know more about Pass's approach to amplifier design, so we had a chat by e-mail.

Jim Austin: Do you own any tube amps?

Nelson Pass: I have a couple of commercial EL34 SET [single-ended-triode] and P-P pentode amplifiers—nothing special—and a pair of point-to-point pentodes that Wayne [Colburn] made for me (footnote 1). The SET is interesting because it sounds a lot like the [First Watt] SIT-1 and SIT-2 [power amplifiers], but has more noise and lacks the bandwidth. It's nice to take a listen to them now and then. I enjoy listening to differences.

Austin: You've chosen to work with semiconductor devices, not tubes. Yet you've expressed appreciation for tubes, and in some of your designs you seem to be working to mimic the sound of tubes. Please explain.

Pass: Fundamentally, what interests me most about amplifiers are the differences in sound created by different topologies and the characteristics of the active gain devices. There are few things I enjoy so much as to contemplate the specific (and complex) characteristics of the many transistors (or tubes) and how they might fit into an amplifier to deliver a sound which has a particular signature. Toward that end, I like simple circuits, partly because well-designed, simple amplifiers tend to sound better, and also because they bring the part's personality into sharper relief.

This relates nicely to tubes, in that the economics of tube circuits enforces a discipline of simplicity, where it is not convenient to achieve quality by throwing lots [of] parts and feedback at the problem. You will not see 50 tubes and 80dB of negative feedback under the hood of any tube amplifier.

However, most of the best approaches to tube design are well mapped already. Not so much with solid-state, where the art of simple has been somewhat neglected and where there is a vast array [of] new and interesting parts. That's why I choose solid-state.

Also important, it's relatively easy to take ideas in simple solid-state, and build them up and play with them in short order. It's much quicker to try lots of variations and refine the design. It's much easier to get some built, and have other people listen to them as well.

For example, here's a solid-state circuit you will not have seen, which I showed at the Burning Amp Festival [BAF] last month (fig.1):



image: https://www.stereophile.com/images/917pass.schem.jpg

Fig.1 Nelson Pass's self-biased, class-A, push-pull output stage using a depletion-mode, N-channel power JFET coupled to an enhancement-mode P-channel MOSFET.
It's a self-biased, class-A, push-pull output stage that uses a depletion-mode, N-channel power JFET coupled to an enhancement-mode P-channel MOSFET. No bias circuit, doesn't need degeneration, has built-in temperature compensations. Works great. How's that for simple? I'll make a product out of it (or a kit for DIYers) next year.

So I don't have any problems with tubes. It's just that, with the limited number of years I have to work with, I need to have some measure of focus.

Tube design has plenty to say to solid-state, although I don't see most designers paying that much attention. On numerous of my projects, I have revived old ideas and added some of my own. My favorite transistors are JFETs, MOSFETs, and SITs, all of which are analogs of tubes. The JFETs and MOSFETs are a lot like pentodes, and the SIT, a special form of JFET, is best modeled as a triode (footnote 2).

I haven't designed a bipolar transistor amplifier in 27 years—[my work during] the previous 22 years was adequate, I think—so you could fairly say that I replicate tubes at least some of the time.

The other fundamental thing—number 2—is that I am centrally aware that all this is just entertainment, mine and yours. The objective needs of amplifier users are largely solved on a practical level, and as [Marshall] McLuhan perceptively noted, when that happens, we turn our technology into art. For me, the art lies in making simple, unusual amplifiers that sound great and measure fairly well. They aren't for everyone, but if they appeal to even a narrow segment of audiophiles, I'm perfectly happy. I'm equally happy if they are reliable.

Austin: You say that your favorite transistors are JFETs, MOSFETs, and SITs. Would you say that each of these classes of transistor has its own sound, which you can extract with appropriate implementation?

Pass: The JFETs and MOSFETs have a lot in common, generally a pentode characteristic, and with a few exceptions, JFETs are relegated to input-stage and line-level buffer types of use. Those exceptions would be The Beast with a Thousand JFETs, and of course the power JFETs I have in Silicon Carbide (SiC) from USiC [United Silicon Carbide Inc.] and the now-defunct SemiSouth (footnote 3).

All of these have their specialty, but you can't characterize the sound generally, as each has strengths and weaknesses, which become important in different modes of usage. FETs [field-effect transistors] as a group are often described as more tube-like than bipolar transistors, and that is a reasonable sort of generalization, but too broad to be actually useful. Bipolars have their own character, but the same sort of thing applies about use. For the past 27 years I have dealt in FETs as a matter of personal preference, but bipolars still make it into some of the circuits, usually as cascode transistors.

The SITs are a special breed of JFET that have a triode-like characteristic, not pentode like the other FETs. This gives rise to some interesting performances, the most entertaining being simulations of SET amplifiers without the output transformers, and really good, larger, push-pull power amps.

Austin: You say you haven't designed a new bipolar transistor amp in 27 years. Is that because you don't like the sound of bipolars, or is it because you felt you'd done all you could with them?

Pass: Mostly because I was done, but I also wanted to head in a different direction with devices and the sound.

Austin: Apart from the choice of devices and circuit topology, what else matters in amplifier design? Layout? Boutique parts?

Pass: People may not appreciate how many potential options a designer has. Think of it this way: There are three ways to use a three-pin device. For a FET, this would be Common Drain (a voltage follower), Common Gate (voltage gain with unity current gain), and Common Source (voltage and current gain). All transistor and tube devices have the same list, but the pin names are different.

For a thumbnail calculation, [if] we go to the Digi-Key catalog and look under power MOSFETs powerful enough to use in an output stage, we see about 3000 easily obtained parts. Under bipolar devices we see about 600 such parts. So the permutations available for a single stage (in this case, output stage) start at about 10,000 for a single-device, single-ended circuit. This sort of number becomes squared for a two-stage amplifier, and cubed for three-stage amplifier, and so here we are at maybe a trillion possibilities. And we still haven't decided which stages are biased by how much (a continuum of values), or how much each device is allowed to contribute to gain and by what technique. And so on.

 

[Mr Spock]

image: https://www.stereophile.com/images/917pass.einstein.jpg

Rather than becoming overwhelmed by the permutations, the author, artist, or designer—because writing and painting, for example, offer the same endless range of possibilities—can borrow from the work of predecessors, refining technique and hopefully eventually transcending those influences and "finding their own voice." I refer you to Harold Bloom's The Anxiety of Influence: A Theory of Poetry (footnote 4)—or if you just want a great book to read, John Horgan's The End of Science: Facing the Limits of Knowledge in the Twilight of the Scientific Age is back in print again after 20 years (footnote 5).




Footnote 1: Wayne Colburn is a designer and a member of the core Pass Labs team. He first joined Nelson Pass in 1994, at Pass Labs, after answering an ad in Audio Amateur seeking a "wizard's assistant," as John Atkinson wrote in his review of the Pass Labs XP-30 preamplifier in the April 2013 issue. Colburn is largely responsible for the design of Pass Labs' preamplifiers and integrated amplifiers.
Footnote 2: Because of its vertical construction, the static induction transistor (SIT) was known in the late 1970s as the vertical field effect transistor (VFET).

Footnote 3: See, for example, the First Watt J2, which was reviewed by Herb Reichert in October 2016.

Footnote 4: In this book, literary critic Harold Bloom argues that a poet's influences can cause her or his work to be derivative, and outlines ways poets can avoid the resulting "anxiety" and create more original work.

Footnote 5: Similar to Pass's comment that "The objective needs of amplifier users are largely solved on a practical level," The End of Science, first published in 1996, posits that most of science's big ideas—evolution by natural selection, the big bang, relativity, quantum mechanics—are in place, and that in the future there will be far fewer scientific "revolutions or revelations." The book was widely criticized, including by prominent scientists, but in an essay in Scientific American, Horgan said, "hell, no," he hasn't changed his mind.






Read more at https://www.stereophile.com/content/nelson-pass-circuit-topology-and-end-science#Bel6y0yfX0jzuBtv.99

 

[Mr Spock]

We have gain devices; basic topologies; values of the many passive components, like resistors, capacitors and inductors; classes of operation (bias points and load lines); and a host of other details—but it's more about process. Because I personally do not have anyone dictating the desired results—lucky me—I often start with a consideration of the gain devices themselves, as their characteristics suggest the best or most interesting circuits that might play to their strengths.

I lean toward class-A because it makes the process smoother—again, no boss—and I work on individual gain stages with an eye toward an amplifier comprising three or fewer of them.

Of course, a lot of the work is just the day-to-day routine of layout, testing, refining, revising, documenting. Depending on the design, boutique parts are either first or last. Some of my transistors are boutique.

Austin: You're very open with your ideas—with what's often called "intellectual property." The DIY community is surely grateful. Can you explain your thinking about that?

Pass: I have more stuff than Pass Labs can possibly bring to market, much of it not appropriate to PL's audience anyway. So I created First Watt for "low volume" commercial product, and release most of the rest to DIYers.

The DIY thing has worked well since 1973, starting with The Audio Amateur, and currently mostly through www.diyaudio.com and the www.passdiy.com and First Watt websites. The most complete collection of articles is at www.firstwatt.com

It's still [my] intellectual property—and copyrighted—but it's available for noncommercial use. It has been argued that this favors our competition in some way, but I have not seen it.



image: https://www.stereophile.com/images/917pass.withcircuites.jpg

Austin: Some of your comments, in this interview and elsewhere, suggest that you take a nuanced view of the connection between measurements and the perception of music. Can you expand on this?

Pass: Measurements and listening go hand in hand. There is a correlation between objective and subjective, but they're not strictly causal relationships. Clearly, there are some amplifiers that measure great with "standard" measurements but don't sound so good, and there are examples of good-sounding/bad-measuring as well. The discrepancies are interesting because they point to either things that have not been measured—more likely, misinterpreted—or aspects of perception and taste that don't correlate to measured flaws. Or both.

In the end, the subjective experience is what our customer is looking for. Our taste in sound may not appeal to everyone, but it's what we have to work with, and we only need a small segment of the market to be successful. I don't neglect the measurements; I put them to work.

Austin: In the brochure for the Pass Labs Xs-series amplifiers, there's a discussion of the influence of your First Watt SIT experiments on the development of the Xs series, via the "concept amps" you were working on earlier this decade. A key to your design process, apparently, was the subjectively excellent and easily heard sound of the SIT amps. You were, as the brochure says, "able to objectively identify the qualities of the sound—information which remains proprietary." As a former scientist and current subjective reviewer, I find this very interesting. Is there more you can say about "objectively" identifying "qualities" of sound?

Pass: When I had a custom batch of SITs made by SemiSouth, the first thing I did was test the performance of the part in the simplest possible circuit, which was just the one transistor in single-ended, class-A, Common Source mode without feedback. The SIT being very much like a triode, it is easy to make a single bias adjustment which affects the second harmonic distortion of the device, ranging from a relatively large amount [of] positive phase second [harmonic] through a null point with no second [harmonic], to large, negative phase second-harmonic distortion.

The result was so interesting that I released it as the SIT-1, with a knob on the front panel and a meter showing the voltage/current operating point of the SIT, so that users could make the adjustments and decide what they liked best. We floated some prototypes around and found that most people preferred a particular setting. There was a consistent subjective observation that there was a difference not only with the level of second harmonic, but phase also. Negative-phase second harmonic tends to expand the perception of front-to-back space in the soundstage, separating instruments a bit. Positive phase does the opposite, putting things subjectively closer and "in your face." I have heard this sort of comment from people who were not in a position to have expectation bias, so I treat it seriously.

Of course, there was also such a thing as too much second harmonic. It is a trick best done in small doses.



image: https://www.stereophile.com/images/917pass.neilyoung.jpg

Armed with this, I went back to the still-under-development Pass Labs Xs amplifier, which has a massive, balanced, class-A push-pull output stage biased partially with single-ended high-current sources. Some work on the values of the bias currents gave us what we wanted, and finished products went to market. They were followed by a generation of .8-version X and XA amplifiers. They sound great, and they also measure well.

Both Pass Labs and First Watt have enjoyed success from the initial investment in the SITs. The big difference between them is that it's practical for First Watt to build anything without disrupting business at Pass Labs, and being all the same pockets, there is no conflict.

Austin: Earlier, you said, "I am centrally aware that all this is just entertainment, mine and yours. The objective needs of amplifier users are largely solved on a practical level, and as McLuhan perceptively noted, when that happens, we turn our technology into art." Are you saying that amplifiers already are—perhaps have long been—good enough, musically?

Pass: It depends on who the customer is. Except for the needs of a narrow segment of the population, the problems of amplifiers are largely solved. There remain those who want better yet, and devote themselves to that.

There are objectivists still working to improve the measured performance on the assumption that it will sound better—not a bad assumption, really. And there are people who are looking for a better subjective experience apart from specs. Maybe they like Kodachrome. I'm still exploring the territory, and when I find something interesting and different, I work with it and try to realize its potential. Apart from the commercial interest, there's a large audience in the form of do-it-yourselfers, so little gets wasted.

Austin: Thinking back on the Horgan book, in science—and probably other fields—there are explorers who seek out big, new ideas, and others who focus on getting the smaller things perfect. Would you say you're more the former or the latter?

Pass: Oh, I want perfection, all right, but I'm secure in the knowledge that I won't achieve it.


Read more at https://www.stereophile.com/content...gy-and-end-science-page-2#jYpyW7tmICr1mfui.99