300b lovers

By way of comparison, neither the Raven nor the Blackbird use any form of feedback, either local (around the tube) or global (around the entire amplifier). The incoming audio signal only flows forward, with no secondary paths around the circuit. Further isolation is imposed by isolated B+ supplies for input+driver and output sections, so there is no secondary path for B+ power supply intermodulation (clipping in the output section has no effect on the preceding circuits).

This means distortion is entirely the result of device linearity in the specified circuit. The gains are scaled so each preceding stage has 3 to 6 dB of headroom compared to the following stage, so in practice clipping only happens in the 300B power section. The B+ regulator for the output section has a peak output of 200 watts, so the only limiting factor is the peak current capability of the output tubes. The performance of the 300B pair sets the performance of the entire amplifier.

By contrast, in a feedback amplifier (of any kind, solid-state or tube), clipping and/or slewing creates large error transients at the feedback summing node. This can saturate the input stage, which means the entire amplifier is now clipping, and can lengthen the recovery time from clipping.

I should note my description of feedback circuits is a grossly oversimplified, non-mathematical overview of a complex subject. For the curious, read about how op-amps are stabilized, and the concepts of loop gain, excess gain, dominant-pole compensation, and phase margin. Once you get a reasonably firm grasp of how it works, then read about slewing distortion and settling time. I tend to use settling time as a figure-of-merit when looking at op-amps, or more complex discrete circuits.

It all comes together at the summing node, which is simply an analog comparator between input and output. In an op-amp, which has extremely high forward gain, the high gain of the op-amp forces the differences between the two nodes to zero. This is fine until the op-amps clips or slews, which creates very large error voltages at the comparator input. The large error voltage can force the comparator itself into nonlinearity, and feedback theory relies on a distortionless comparator.

In addition, if the comparator is saturated, or if the power supply sags or is discharged, then recovery time can be quite long (tens or hundreds of milliseconds), much longer than the original clipping or slewing event.

During this settling time, amplifier distortion can be quite high, since feedback is only partially effective. This will not appear in FFT harmonic distortion or multitone IM distortion measurements, which are taken over several seconds and then averaged.

This is the gap in existing measurement techniques. Harmonic and IM distortion are averaged over several seconds, and do not sense events happening in microseconds or milliseconds. High-speed scope measurements are insensitive to distortion unless it is very high, such as 10% or more, where it becomes visible. Transient distortions, in the microsecond to millisecond range, are not seen.

The key principle of non-feedback amplifiers is they are insensitive to transient upsets or interactions with the load. Steady-state distortion is higher, but there are no issues with phase margin or settling time.

Areas for improvement: The 5U4G rectifier is not ideal. I’d use HEXFREDs, high-voltage Schottky rectifiers (Don’s choice), or damper diodes for the 300B plate supply. Any of the three will have more dynamics and more vivid tone colors. The improvement should be immediately audible, two weeks will not be needed, you should hear it right away.

RC coupling will sound more dull and compressed compared to dynamic loads, LC coupling, or IT coupling. What RC has going for it is resistor coloration is less potentially noticeable than the other three methods, which each demand very careful component selection.

The final capacitor coloration of the filter sections will be audible, although less so than the cathode bypass caps, which are extremely sensitive to cap coloration.

I wouldn't try mixing and matching Ralph's approach with ours. Ralph has his way of doing things, and his own unique taste in sonics, and we have ours. Most designers in this biz have a distinct "house sound" that they aim for, which results from design approaches and parts selection.

I should mention SE tuning is not the same as balanced-amp tuning. The dominant coloration with SE are the tubes themselves, and it requires artful selection to avoid heavy additive coloration. The fad for 2-stage SET amps makes this worse, since high-transconductance tubes are not designed for audio, and distortion can be all over the place using tubes designed for RF use. Selecting designed-for-audio tubes in a 3-stage amp makes things simpler and more manageable.

Unfortunately, common design practice in Golden Age push-pull pentode amps is not helpful in designing a balanced non-feedback amp. All of the 1950’s and 1960’s Golden Age amplifiers use feedback as a required part of the design, and the "balanced" part of the circuit (the output section) is typically running in Class AB. You have to reach back to the 1930’s to find useful non-feedback Class A designs.

Once pentodes and beam tetrodes took over, feedback came along with them, and that changed the overall emphasis of the contemporary designs. The search back then was for more power, more efficiency with B+ supplies of 500 volts or less, low measured distortion, and cost and weight reduction.

That search reached an end when high-powered transistor amps replaced tube amps around 1966~1968. Transistors dominated the broad consumer market with the exception of guitar amps, which kept the tube factories going. The decades-long Japanese fascination with triodes finally came West in the early Nineties, where it created a niche market in the high-end sector. (Helped along by Joe Robert’s "Sound Practices" magazine.)

It really helps that tubes are now so popular in the true high-end sector of the market. Transistors ruled the market in the Seventies and Eighties (with the exception of Japan), and the Nineties were an era of transition and growing acceptance of tube electronics. The home theater and 500-watt crowd are still all-transistor because they need the efficiency, and Class D will give them even more efficiency.

Let’s talk technical about DHT filament power supplies. My friend John Atwood built a low-RF filament supply a few years back (100 kHz or so), but he made the discovery that the filament inductance of DHT tubes was all over the place, making it difficult to assess how much power was actually going into the filament. This is a big deal because DHT tubes require very tight control of filament power, preferably 5% of specification, or better. In practice, the RF supply had to be individually tuned for each tube ... and it didn’t sound any better.

An aspect of even very clean sinewave AC heating are "hum sidebands" ... not hum per se, but IM distortion harmonics that are displaced by 100/120 Hz on each side of the fundamental and about 60 dB down. It is clearly visible on a good spectrum analyzer, and is caused by small temperature fluctuations on the filament modulating emission, which in turn modulates the forward gain of the tube ... not much, but it is measurable. (The discovery of hum sidebands led to the experiments with RF heating.)

An aspect of DC heating is where the virtual center-tap appears. If the virtual center-tap is in the middle of the filament, the inherent balance of the filament can give a free bonus of 30 dB or more of additional noise rejection. Considering how difficult an additional 10 dB of noise rejection can be, 30 dB is not to be sneered at. And the whole DC supply has to float, relative to ground, while appearing symmetric from the viewpoint of the filament (mimicking an AC supply in that respect).

Although I don’t enjoy "tuning", it is an unpleasant necessity for speakers and power amps. For reasons that are not clear, various brands of metallized polypropylene capacitors sound quite different from each other, and there is little correlation with DA and DF parameters. Based on measurements, they should all sound the same.

On a system with moderately high resolution, subjective differences appear that can mimic crossover balance shifts and driver swaps. I found during development of the Ariel, back in 1993, that cap substitutions required 0.5 to 1 dB crossover adjustments to subjectively offset the colorations ... and this was with pink-noise test stimulus, not music.

You can really get into the swamp comparing silver vs copper wire. This should not be audible at all, and I have heard of no convincing argument why any differences are audible. You can go out on a limb and compare silver oxide vs copper oxide, and various weird sources of corrosion, but it’s all very speculative, and again, no useful measurements to be had.

On the other side of the objective/subjective fence, I have heard of well-known speaker designers who never audition their new speakers ... they do it all by numbers, then walk away. I frankly didn’t believe it when I first heard that about fifteen years ago, but other folks confirmed it, so I guess it happens. So it is possible to ignore "tuning" and let the product sound like whatever.

But in the speaker world, it is widely recognized that a "perfect" zero-coloration speaker is impossible at the current state of the art, so it comes down to choosing which set of parameters are most important. Speakers are still very imperfect, compared to any other audio component.

In principle, it should be possible to design and build a zero-coloration amplifier. I started my career in audio design in 1973, and haven’t heard a "perfect" amplifier yet. They all have a sound, and a little bit worse, topologies tend to have distinctive sounds. But that’s my personal experience, not necessarily the experience of others.

If a customer, or reviewer, is in the fortunate position of finding that all well-engineered amps sound alike, that’s great! You can sure save a bunch of money, skip over tubes entirely, and just buy the latest Shenzen-made confection for a few hundred dollars.

My first experience of a 300B amplifier was back when I was writing reviews for Positive Feedback magazine back in 1993. It had grown from a 4-page mimeographed club magazine for the Oregon Triode Society to a fat 100-page periodical with cartoons, editorials, and a staff of reviewers.

I was one of them, after submitting a series of construction articles for the Ariel twin transmission line loudspeaker. I was listening to a whole string of amplifiers when David Robinson, the magazine editor, dropped off not just the Audio Note Ongaku (which cost three times the price of my car) and the Reichert Silver 300B’s. So now I’m in the reviewing business, too. Oh well, not one to look a gift horse in the mouth, away I go, with a month’s free use of entirely new amplifiers on my brand-new speakers.

Which was quite a revelation. When you design a speaker, and you’ve been doing it for decades, you get to know them pretty well. What they can and can’t do, and what the overall character of the speaker sounds like. Nothing new to me, and the Ariel was my eighth speaker design, after my time at Audionics during the Seventies.

One of the last stages of a near-commercial design is auditioning the speaker on many different amplifiers, keeping in mind the impact of output impedance (damping factor) on the bass alignment and the crossover. The Ariels are designed to have low sensitivity to output impedance: transmission line bass with no impedance peaks, low-Q 2nd-order crossovers, very flat drivers that don’t need equalization, and a tweeter that is running flat out, with no attenuation, since the paired midbass drivers have matching sensitivity. So all amplifiers see an equal playing field, with a minimum of power disappearing in resistors.

The shock with both DHT amplifiers was a radically different sound than Class A transistor, Class AB transistor (with high slew rate), and push-pull pentode. Much higher transparency and much more vivid tone color ... out of a speaker that I knew very well, and was my own brainchild. This is what led me to design my own DHT amplifier, but not following the path of either the 211-based Ongaku or the Reichert Silver 300B. But definitely using the 300B for sure; no 2A3 came close, and 211’s and 845’s (with plates running at 1 kV) are extremely difficult to use.

Don has brought me up to speed on modern power supply design, things I didn’t know about in 2003, when the Karna was designed. That takes the Karna to a new level, and reduced it from a ridiculous four chassis setup with high-voltage Amphenol connectors to a much more sensible pair of monoblock chassis. One pleasure of working with Don is he will chase down every possible variant, build it, audition it, and let me know how it measures and compares to all the rest. Fortunately, we have gone full circle and have arrived at a Karna Mark II with far superior transformers and power supplies.

Don did a marvelous job here. It’s actually quite functional, with a clean and direct front-to-back signal path, and with all wires in the balanced circuit equal length (yes, he went to that much trouble). You can infer that from the top plate.

The power supply side is equally tidy, with two independent B+ regulators, one in front of the other, and the three low-voltage regulators on the other side of the power supply section, all supplied by the custom Monolith power transformer. The soft-start circuit in the back of the amp keeps incoming AC power away from the front-panel power switch, as well as protecting the tubes and regulators from AC line transients.

The amp is considerably simpler to build, because signal flow is obvious, and color-coded wiring is used to keep track of polarity. It is also the same 18"/457mm width as the matching Raven preamp.

This is very close to the production version. It is simpler and more straightforward to build than the shoebox format shown at the PAF show ... more spacious, more direct layout, and the power supplies are confined to their own section of the amp, on the right side of the chassis. The vent holes for the twin B+ regulators are visible on the right side of the amp, next to the VR tubes.

The audio-only circuit is on the left side, with very short signal paths from 6SN7 -> interstage 1 -> matched balanced 6V6 -> interstage 2 -> matched balanced 300B -> Monolith output transformer -> speaker jacks. The input selector switch bypasses the input transformer when XLR is selected. Compared to the show amps, there are actually fewer parts in the production version, with a very simple signal path from input to output. From input to output, there are only wires, transformers, and triodes, in a fully balanced circuit. No coupling caps, no plate-load resistors, no plate inductors, and no dynamic loads.

One subtle difference is each single grid is driven by a pair of balanced plates, so distortion and noise are minimized in every stage of the amplifier. In the show amps, each 6V6 grid was driven by the corresponding 6SN7 plate. In the production amp shown here, each 6V6 grid is driven by a balanced pair of 6SN7 plates, thanks to interstage 1. I was doubtful a good interstage could be made for the 6SN7, but our transformer designer came through with performance from 18 Hz to 32 kHz. Close collaboration with modern transformer design is what made this possible.

I met Thomas Meyer at the 2004 European Triode Festival (I was the invited keynote speaker). He’s a lot of fun, and super knowledgeable about tube history. He, too, is a transformer enthusiast, particularly with modern transformers.

I’m super happy that he made the transition from hard-core hobbyist to the ultra-high-end commercial world. He’s set an example for all of us. It helps that Europe has a fine tradition of artisan-built audio, with wealthy patrons who appreciate the arts.

When I was Switzerland as the guest of Christian Rintelen (host of the 2004 ETF), I  visited the museum in Zurich, and astonished to see wooden clocks that were a thousand years old ... and still in working order. The traditions of technology in Switzerland and Germany are ancient, and a deep part of the culture.

Yes, I would not mind a matched quad of Elrog 300B's with their thoriated-tungsten filaments. That would be something quite wonderful. The Mark I Karna's thrived on a matched quad of Emission Labs 320B-XLS, but that was also a serious investment. The European super tubes are something else.

Considering that each Blackbird B+ regulator can crank out 500 mA peaks, and 400 mA steady-state, there’s plenty of current the super tubes can use. The only thing limiting the Blackbird is peak emission from the 300B filaments ... nothing else.

400 mA might not sound like much until you realize it’s at 480 volts, and the output transformer multiplies the current 28.7 times (on the 8-ohm tap).

(The 4-ohm tap has 40.2 times current multiplication, for those with current-hungry electrostatic speakers.)

The operating points are conventional, well within standard 300B specs, with the plate at about 75% of max rating (which typically gives maximum life). But if you really want to slam the amplifier, well, you can. There’s nothing stopping you. The amplifier gradually turns into a limiter, with the output current saturating on peaks. If you treat it like a guitar amp, running into heavy distortion for hours at a time, the output tubes will wear out more quickly (the same as a guitar amp). Momentary overload, as in music playback, is benign.

If peak power, and particularly, peak current, is important to you, the enhanced-rating European tubes are the best choice. Instead of 40-watt plates of the classic 300B, they have 65-watt plates, and peak current emission is 50% higher, or even more. Although they are related to 300B’s and bias the same, they are in the KT120 class, not EL34 class. Unlike the 845, which relies on voltage to deliver more power, the super tubes provide more current thanks to higher emission, and plates are more generously sized, keeping them cooler.

(User note: Not recommended or guaranteed for guitar amp use. 300B’s are physically fragile and can be damaged by high vibration, excess heat, or operation with a tilted chassis. If you own an electric guitar, get the right amp for the job.)

What separates the Euro super tubes is massively higher electron emission. For example, what separates the EML 320B XLS from the EML 300B XLS is 50% greater heater current, with correspondingly higher emission. But even the EML 300B XLS, although consuming the same heater current as a standard 300B, has significantly higher emission.

Same story for the thoriated-tungsten Elrog tubes. Thoriated-tungsten is used in high-power transmitter tubes, not "radio tubes" with the typical coated filaments. The peak emission characteristic will sound more like the Eimac giant transmitter tubes than a 2A3 or 300B.

What’s fun about these is they are plug and circuit compatible with classic 300B’s, but they are not replicas. Not at all. They are modern high-power designs, optimized for both linearity and peak emission. And they sound like it ... powerful and blazing fast.

But ... the amp circuit has to match the peak power and speed, or you never hear what they can really do. They sound like just another 300B with generic RC-coupled drivers, for example, and the potential is mostly wasted. Give them a driver that is a powerful Class A amplifier that is transformer coupled, and whoa, stand back. It’s not a meek little flea-power amp any more; put on Mahler or Mastodon and frighten the neighbors.

One unique feature with transformer coupling is 97~98% of the plate power is presented to the following grid. The plate power doesn’t disappear into a resistor, choke load, or MOSFET transistor current source, or jump through a plastic-film dielectric in a capacitor. It’s right there at the grid, with only tiny losses in the transformer.

This is audible as vivid and tactile tonality, speed, and power. Basically, less electronic cruft in the sound, which is what you’d expect. Don and Whitestix will attest to that.

I suspect any electrostat is fine. The weird capacitive/inductive load will not bother a zero-feedback amplifier.

What will not work is something like Wilson Audio or B&W speakers, with low efficiency, a band split into three or more drivers, complex crossovers, and big woofer arrays. Or MBL. They really do need 200 to 500 watts with a high damping factor (lots of feedback). Transistor amps, in other words.

Reflecting on a recent phone conversation with Don (he’s in BC Canada and I’m in Colorado, a bit north of Denver), I suggested that SET amps are kind of like a paint-box, and a much more fun way of tuning a system than messing with cables. If you are DIY’ing, there are many ways of changing the tone color ... which coupling caps, what kind of passive power supply, which rectifiers ... the options are endless. And a lot of fun if the amp is on a breadboard and you can solder in new parts in a few minutes.

The characteristic SET sound works in your favor, giving a lot of leeway with parts selection. And the amp is fundamentally simple and easy to understand, a godsend when you are tuning with many variables. Like I said, a paintbox. When you learn painting, you learn color harmony and the art of mixing. Nothing teaches faster what XYZ cap sounds like than heating up the soldering iron and swapping parts.

A balanced amp is a harsher taskmaster. Yes, more transparent, potentially by 20 to 30 dB, but not nearly as forgiving. Colorations can sound pretty ugly if the wrong part is in the wrong place. And there is no feedback to tidy up the mess. Maybe more like working with an airbrush, or transparent watercolors, instead of pigments. There’s still balancing to be done, but the high level of transparency, and lack of feedback, exposes everything. I found this out the hard way with the original Amity amplifier back in the Nineties.

And the colorations from different part selections are not the same as SET. This makes sense when you reflect on it ... the balanced circuit is cancelling most, but not all, colorations, and the residue left over can be unwelcome and surprising. The SET experience can be a very rough guide telling you which parts sound really awful, but it will not tell you which sound the best.

This mirrors working with speakers. As transparency goes up, tolerance for coloration goes down. In the absolute sense, this is wonderful, because now you’re really hearing the music. In a way, I’m not surprised the simplest topology won ... less to go wrong, and with the most efficient plate-to-grid coupling.

Hi, Alex!

The original Karna, designed by me, and built in four-chassis format by Gary Pimm (of Portland, Oregon) in 2003, used a 5687 class of input tube. (The 6900, 7044, and 7119 all have the same pinout and similar operating points. The current production JJ ECC99 is similar but has a different pinout.)

I selected that tube for the Amity, back in 1997, and also for the Karna because it had a low plate impedance ... around 2K ... and pretty decent linearity, much better than a 12AU7, which is quite poor and not really suited to driver duty. But I was never entirely happy with the 5687 or the other similar types. I tried just about all of them ... I have quite a stash of 5687, 6900, 7044, and 7119 tubes ... but there was always a bit of glassy, hard quality, nowhere as bad as a 6DJ8, but still there.

There’s nothing wrong with them, again, far better than any 6DJ8, but these are commercial tubes never intended for audio use, and never used in any Golden Age amplifiers, tuners, or TV sets. They were designed for analog computers, commercial radio relay use, and aerospace ... high-end commercial and military applications, at high prices, and not sold in consumer retail channels.

These days they come from military surplus stocks, and only produced in consumer format by JJ as the ECC99. So supplies are getting a little dodgy, twenty years on. Not really suitable for consumer use unless you already have a substantial stash of them, in the hundreds, and all tested and matched, of course.

The 6SN7, and its single-triode predecessors, like the 6J5, 6C5, etc. etc. are famous for their linearity, and they were designed for radio applications in the audio sections of the receiver and power amplifier. Millions were made, in varying quality, but all of them were more linear than the 12AU7 successor, or the quite different 6DJ8 (which was an RF tube never intended for audio). So there’s nothing rare or exotic about the 6SN7, unlike the 5687 family.

I mention "designed for audio" as if it is something special. Well no, not really. But if a tube was originally designed as an RF amplifier, it would never be checked in production for linearity, since RF circuits don’t care about linearity. Nowadays, of course, 6DJ8’s are never used for RF circuits, and only for audio, mostly high-end audio, not guitar amps.

This has the practical effect that vintage (NOS) stocks of authentic 6DJ8’s can be all over the place for in terms of linearity, since that’s not a controlled manufacturing parameter and would have no effect on its performance plugged in to a 1965 RCA color TV set or FM tuner, the task for which it was designed.

In practice, using Gary Pimm’s custom-designed spectrum analyzer with 140 dB resolution, we found that upper-harmonic (5th on up) spectral shapes mostly reflected a given manufacturer, and was surprisingly consistent from year to year. Gary Pimm and I have both worked in manufacturing for big and small companies, and we surmised that consistency reflected the special jigs that aligned the grids, and different manufacturers used slightly different techniques to align the inner structure.

Although tube models are intellectually useful in a design phase, they model ideal tubes that are only available as beautiful Platonic Ideals in a store somewhere in Heaven. Sadly, we humans on Earth have no access to that store. No Platonic Ideals for us.

The tubes we can actually buy were, and are, hand-made by skilled human beings, not robots. The grid pitch is not perfectly uniform, the grids are all tilted just a little bit, electrons escape out each end of the structure, the list of imperfections (and departures from ideal models) goes on and on. These tiny imperfections result in high-order harmonics that can be seen in a high-resolution spectrum analyzer, and heard in a good audio system.

Surprisingly, these departures from perfection are consistent with the manufacturer. That’s why Gary and I surmised it came down to small variations in assembly technique, or even the individual assembler. Again, tubes were never assembled by robots, and still aren’t today. The assembly was, and is, semi-automatic at best.

Frame-grid tubes, like the 6DJ8 or more exotic WE417A, are even more difficult to make consistently, and it doesn’t matter in a high-gain RF circuit anyway. Using them in an audio circuit is a roll of the dice, especially if there is no feedback to tidy up the mess. Harmless in a preamp at millivolt levels, not so good in a power amp.

For all these reasons, Don and I decided to move away from the 5687 family. (Neither Don nor I are fans of miniature 9-pin tubes anyway.) True, the 5687 family greatly simplifies the interstage transformer design, since the plate impedance is about three times lower than 6SN7, but that low plate impedance is the result of high transconductance and more difficult assembly procedures. Part of the reason that direct-heated triodes have a much cleaner spectra is they are big and easy to assemble ... as dumb as that. We’re talking late Twenties to late Thirties technology here ... precision assembly was very difficult back then, especially on a production basis.

Effectively, Don and I took the ultra precision out of the tube and put it into the transformer designer and assembler. That’s where the 21st Century tech comes in. These transformers could not have built in 1939, when the 6SN7 first came on the market (replacing single triodes). The 5687 family dates from the mid-Fifties, with transformer design still in the build-and-try phase, like the loudspeakers of the day. Computer modeling was still decades in the future.

To sum up, we have a circuit with big, simple tubes designed no later than 1939, combined with 21st-Century transformers and power supplies. In that sense, it is a hybrid amplifier, spanning 84 years of time and technology.

(If you want the Blackbird to fly even higher, look to the Emission Labs 320B-XLS or the ELROG 300B with thoriated-tungsten filaments. Those are 21st-Century 300B’s.)

Alex, the Shishido 811 circuit is basically uncopyable, since it relies on DC flowing through the secondary of the custom interstage transformer that goes into the 811 grid. Unlike nearly all other audio circuits, this circuit operates the 811 power tube ONLY in the positive grid region ... from zero volts to a substantial positive voltage.

When I met Shishido at the CES back in the Nineties, as technical editor of Glass Audio, I pressed him on this point. In Shishido’s "Inverted Interstage Transformer" designs, the grid voltage swings from zero volts to a higher voltage. It never passes through the zero-bias region (according to Shishido).

This requires DC current to steadily flow into the grid, while the grid is an extremely nonlinear load for the driver stage. There’s only two ways to pull this off: a powerful MOSFET driver with a paralleled current source (MOSFET likewise), or a very special interstage transformer that can tolerate a lot of DC going through the secondary, while current goes through in the opposite direction in the primary. If you did it with MOSFETs the chances of a spectacular explosion would be pretty good. You don’t mess around with transmitting tubes.

Brilliant but the weirdest thing I’ve ever seen. A (very) custom interstage with bidirectional DC current flow. Zany doesn’t begin to describe it. My worry would be matching the current flows to the exact values. Tubes love to drift ... they are not well suited to DC circuits. Tektronix scope designers went to insane lengths to DC-stabilize their vacuum tube scopes, and this amplifier would also require a complex DC-stabilized supply.

How did it sound? I preferred its big brother, the monster 833 amplifier, which was the top-of-the-line Wavac IIT amplifier. That used a hand-selected vintage KT88 from WAVAC private stock as the driver. When you bought the WAVAC 833, they set aside several vintage KT88’s (real British Genelex) just for replacement purposes. Shishido told me that, and I believed him.

I also loved the stunning solid aluminum NC-milled chassis and custom safety glass enclosure for the insanely hot (and very dangerous) 833 transmitting tube with the top cap at many kilovolts. That probably doubled the price, but man, it looked really cool and high-tech.

Transmitting tubes are in the "look but don’t touch" category. In real transmitters, they are behind thick safety glass, with interlocked steel cabinet doors. If they blow up, it’s no joke. The steel doors and safety glass are there for a reason.

Investigated in some depth in this 1997 Glass Audio article by Matt Kamna (designer of the Whammerdyne 2A3 amplifier) and myself:

Hidden Harmonics

We found that transformer coupling had the most favorable distribution of harmonics ... by that, the smoothest and fastest drop off. Other forms had more harmonics, with more uneven distribution. Test conditions: 6SN7, single section, 50V rms out, with several different circuits, with and without cathode bypass capacitors. Noise floor with this setup was -118 dB, and harmonics out to the 11th were investigated.

To my knowledge, this was the most thorough examination of vacuum tube harmonic generation at the time, using direct measurement instead of reliance on tube models. Standard assumptions about local feedback from cathode degeneration, and SRPP distortion cancellation, were proven wrong. RC coupling, in particular, was shown to have quite high distortion, while transformer coupling was the lowest.

Direct coupling would have no effect on tube loading, which is responsible for the spectra shown. If we were to re-do the article, we’d try a MOSFET cascode current source load, as well as transformer coupling, SRPP, and RC coupling.

We were surprised that cathode degeneration doesn’t work, and creates some nasty high-order terms instead. Separating the data into even and odd-order terms was essential to unscrambling the chaotic results of the spreadsheet ... a legitimate way of looking at the data, since the underlying transfer curves of odd-order (S-shaped) and even-order (C-shaped) distortion are fundamentally different.

Nowadays, we have the computer power to discover the actual shape of the input/output transfer curve, and exaggerate it enough to be visible. The regrettable drawback of FFT spectral information is that phase is usually discarded, so the underlying transfer shape cannot be found (although it can be inferred).

(What I mean by this is the phase of the distortion harmonics is important. For example, a square wave and a triangle wave look exactly the same on an FFT spectral display; the only difference is the phase of the harmonics. The magnitudes are the same. In real circuits, square waves and triangle waves are created by entirely different mechanisms, so this is important data.)

Which returned the Blackbird to the original Karna topology, with far superior power supplies, and the luxury of interstage transformers specifically designed for the Blackbird amplifier.

I give Don full credit for doggedly trying every possible form of coupling, optimizing each circuit with the most favorable operating point, and giving it a serious, I’d even say exhaustive, evaluation. While I sat back with original 20-year-old Karna circuit and criticized from afar. I’m sure I annoyed the hell out of Don more than once.

The selection of an IT for the output section is obvious. The driver, which has to swing a lot of volts at very low distortion, gets to transfer all of its power to the DHT grids. If the DHT grid swings into Class A2 and starts drawing current, no big deal. The power is there, and there are no caps to charge or discharge. Recovery time is instantaneous, unlike RC or LC coupling, and there no risk of DC-coupled failure propagating from driver to output, as there is in solid-state equipment. It really is ideal.

The input tube was another question. In principle, at the lower working voltages, there shouldn’t be much difference between any of the methods, with RC coupling as the obvious and cheapest method. Unfortunately, that’s exactly what it sounds like.

The more serious auditioning over the last year was between current-source + cap coupling, inductor loading + cap coupling (LC), and straight transformer coupling, with no coupling caps or grid resistors involved. And that sounded the best.

Also the simplest. Six parts ... two custom inductors, two good-sized and quite expensive caps, and two grid resistors ... are replaced by one reasonably compact, purpose-designed transformer. The folks who own the "shoebox" format amps, as demonstrated at the show, can be upgraded to the new circuit, which actually opens up space under the chassis. All new amps will have the new circuit, of course.

Charles1dad, thanks for the compliment ... much appreciated. Don and I put a lot of work into these seemingly simple amplifiers.

DHT’s had a rather short reign in audio (much longer in transmitting tubes). It was only from the early Twenties ... the dawn of radio ... to the late Thirties. Once the 6L6 and 6V6 came out (they were designed by the same team), that wiped out the 45, 50, 2A3, 300B, 211, and 845. Even Western Electric abandoned the 300B by 1940 when they designed their new generation of amplifiers around push-pull 6L6’s (WE350). Since the 300B first came out in 1935, it wasn’t in favor all that long.

300B’s have now been in production longer than they were in the Thirties and Forties, rather odd when you think about it. It was the vogue in Japan, Europe, and finally the USA in the Nineties that created the continuing demand for the type and DHT’s in general. It’s been thirty years now, so I think it’s safe to say they are here to stay, along with their pentode cousins.

Class D GaN amplifiers will continue to erode Class AB transistor amplifiers, but I think vacuum-tube amps have an enduring appeal that continues to grow. They now dominate high-end audio, which was not true thirty years ago. I remember going to some CES shows with hardly any tube amps at all, never mind DHT’s, and now they are everywhere.

Now you see quality record players, and tube amps, in movies as a marker of good taste. The movie viewer gets a little buzz when the tonearm descends into the groove, making that distinctive vinyl "click" sound, then you see a tube amp quietly glowing in the background, and wonderful music comes out. The camera pulls back, and you see the protagonist, looking contemplative, and out-of-focus city lights in the background. That alone sets a mood.

I’m really pleased about this. In an era of superb all-digital, all-solid-state 4K HDR video, tube amps continue to make new friends because they sound so good, on all types of music.

Back when I was playing around with 300B’s, I found that each brand had its "sweet spot". The authentic 300B’s, and the exact reproductions, seemed happiest between 65 and 75 mA. The European super tubes, between 72 and 85 mA. The monsters with giant 65-watt plates are probably happy between 80 and 95 mA. Power supplies need to deliver at least twice the quiescent current, preferably more.

If the current is a little low, they are rich-sounding, but also murky and dull, and if too high, super detailed but also wiry and hard-sounding. Very much like setting VTA in a phono cartridge ... there is a correct setting, and you know it when you hear it.

There is also a tiny range of allowable filament voltage, from 4.85 to 5.05 volts. This can change the entire character of the tube, from very dull but rich sounding to wiry and hard. For longest life, it should be exactly 5 volts, and leave the subjective tuning to setting the quiescent current.

The B+ voltage only had a minor effect compared to the other two parameters ... well, three, counting tube swapping. The other indirect-heated tubes aren’t as temperamental ... put them in the usual range, and they sound fine.

One of the maddening things about the original Karna amplifier was sonic variability. Some days, it would be a glimpse of Heaven, and other days, nothing special. It was always extremely transparent ... that’s the nature of the circuit ... but the tuning came and went. Most of the time it was quite good, but every now and then, it was extraordinary.

By contrast, SET amps are usually consistent, any time of day, due to massive 2nd harmonic masking all the high-order harmonics. Tuning a SET is straightforward ... which parts complement the dominant 2nd harmonic most gracefully.

One of my experiments was bringing out the filament circuits on a separate power socket. I bought the Audio-GD AC regenerator, and powered the filaments with that. Sure enough, the variability went away, and I discovered quite small variations in filament power had a big effect on the sound, overwhelming any other tuning decision, including tube swaps. But the Audio-GD liked to run its fans, despite the easy load.

So I was open to Don’s approach of regulating everything, using the proprietary regulators he’d been using so successfully on his 6L6 and KT88 amps. Sure, why not? I’d already been splitting the high voltage B+ supplies between input+driver and output section (to prevent crossmodulation between sections), and using high-quality regulators for a balanced drive for the filaments and heaters made a lot of sense.

(I should mention regulating filamentary tubes is not trivial, and the usual 3-pin low-voltage regulators introduce unacceptable colorations. Filament and cathode nodes are extremely sensitive to coloration, and balanced discrete circuits are required.)

Sure enough, variability gone, vanished, like the morning mist. No change in sonics depending on the time of day. No annoying and objectionable regulator coloration, which is the bane of high-end audio ... that obnoxious grainy transistor sound, coming out of an otherwise good tube amplifier. Not a trace of that, thankfully. Reliable, too, which goes with good regulator design.

This alone justified re-naming the amplifier. I suggested Blackbird (because Red-Winged Blackbirds are a common sight in Colorado), and to my surprise, the name appeared in the Pacific Audio Festival show guide. So Blackbird it is.

The problem of DC coupling vacuum tubes in a balanced circuit is maintaining DC balance ... during warm-up, in steady-state operation over hours, and as the pair age over the life of the amplifier. A small DC imbalance error in the first stage becomes very large in the second stage, resulting in a massive current imbalance in the second stage.

This can be servoed out by a housekeeping circuit, using a bit of analog logic, but if that ten-cent opamp fails, it takes out the entire amplifier. I have seen that happen while I was sitting in the listening room of the editor of the magazine I write for, Positive Feedback. A cheapo servo circuit in the preamp took out the entire power amplifier and the bass driver. $50,000 worth of damage in a few seconds. I don’t care how it sounds, that’s just bad design.

DC coupling without a servo basically doesn’t work. Small drifts become big ones over time, and the circuit will have to be manually re-balanced by the user whenever tubes are replaced, which will happen many times over the life of the amplifier.

I think the horror of transformers has been taken much too far. There’s a reason they have been used so widely in studios for the last eighty years. They are problem solvers. Output transformers take the pint-size currents of output tubes and multiply them 28 times, or more. Input transformers reject common-mode noise and RFI, presenting a clean, quiet signal to the input grids. Interstage transformers sends the power of both driver plates, summed together, to whichever grid needs it the most (grids take turns going into Class A2).

Looking at the driver section, I don’t see the appeal of a cathode follower drive circuit. Adding an additional stage is not exactly direct coupling, and it requires another regulated power supply with oddball voltages for both plus and minus. It’s not a simplification, it’s considerable added complexity, and for what gain? There’s no improvement in slew rate, which is controlled by the current available to drive the Miller capacitance of the power tube grid. The Blackbird has 32 mA of current from each side of the driver, far more than the usual 8 mA of many other amplifiers. The driver can even enter Class AB for a half-second or so, so 32 mA is not the upper limit for grid drive.

The least necessary transformer is between the input tube and the driver tube. The driver grid is relatively easy to drive, and no clipping is seen in that part of the amplifier. The biggest annoyance is the slowly drifting DC imbalance of the input tubes.

In an AC coupled circuit, it doesn’t matter ... it’s only a few volts out of 150 or more. In a DC circuit, though, it controls the bias of the driver tubes, which is a big deal. You really don’t want one tube at be at 50% power while the other is at 90% power, and you don’t to burden the user with meter and knob twiddling on a regular basis. Most of all, you never want to give the user the power to destroy their own amplifier with a thoughtless knob twist.

The 6SN7 DC balance will drift ... not by much, but by a few volts. I do not want it controlling the 6V6 bias point, and most of all, I do not want a solid-state servo circuit to control the 6V6 bias point. That circuit will fail sooner or later.

One option that was considered was a center-tapped inductor for the 6SN7 plates, with direct coupling to the 6V6 grids. But the performance of the inductor, against expectation, was actually worse than the dedicated transformer, and the transformer completely eliminates DC imbalance at the 6V6 grids. Like all transformers, DC is not getting through.

Don and I tried all the more complex options that would give supposedly better operation. They were worse. Removing current-source coloration is non-trivial and difficult in a very transparent amplifier. Bipolar transistors and MOSFETs are audible, even as plate loads.

All of the coupling caps were colored, some much worse than others, but they were all colored sounding. (Once you hear capacitor coloration, you cannot unhear it. Just ask Don.) That was a source of great disappointment. The high-value load inductors had their own set of issues, mostly excess stray capacitance that could not be removed.

I did not expect the transformer to win, honestly. Don and I tried everything else, and the more complex options were always a step downward. In a zero-feedback circuit, you hear every single part. It’s a dumb truism in audio, but simpler usually does sound better. Not that I’m a fan of 2-stage amplifiers or full-range drivers ... there’s such a thing as too simple. Every designer has to find the balance point between simplicity and complexity.

The Kootenai is no slouch: a classic Mullard circuit (as used by Marantz and others) with 6SN7 input and drivers, PP KT88’s biased into near Class A, advanced B+ regulators, all on a compact stereo chassis. A clear step up from restored Marantz, McIntosh, and Citation II amplifiers, and the best Golden Age style amp I've heard.

Don and I are moving towards production of the Blackbird 300B amplifier. We’re now using triode-connected KT88 drivers running at twice the current (56~64 mA per KT88) and one-third the plate impedance (Rp = 700 ohms) of the previous drivers. Not only that, owners will be free to use matched pairs of their favorite 6L6, KT66, KT77, 6550, or KT88’s, with no bias adjustments needed.

The tube lineup is: 6SN7 input, KT88 drivers, 300B outputs, with VR105 shunt regulators. As before, fully balanced Class A from input to output. Special-design Cinemag input and interstage transformers, with custom Monolith output and power transformers. 18"/457mm full-width monoblock chassis with two isolated audio-circuit and power-supply sections, a slow-start circuit, surge protection, and a 12V trigger option. (No AC power goes to the front panel, just DC trigger voltage for the slow-start circuit.)

Don and I anticipate Spatial Audio (of Salt Lake City) will start manufacturing in November 2023.

Don and I were talking on the phone today, and I realized that, of the four to five key people involved, we might have two centuries of professional design experience in audio. I think Don and I might have a century; I filed my initial Document of Disclosure for Shadow Vector in 1973, and Don’s been doing vacuum tube amps since he was sixteen.

And the development arc has been interesting: the PP Karna 300B of 2003 has merged with Don’s PP KT88 Kootenai amplifier.

The self-bias cathode resistor lets you choose anything between 6L6, KT66, 6550, and KT88. So far, the KT88 is winning, but the KT66 is very good also. We're shipping them with a matched quad of KT88's, but the user can use any matched quad in that family ... including NOS unobtainium tubes if they choose, because the operating point allows a life in excess of 10,000 hours.

I was reflecting on this approach vs simply using them as output tubes. Thanks to the gain of the 300B's, they're only using 1/3 to 1/4 of their potential swing, and the 300B's completely isolate them from the speaker load. So the only "load" is the 60 pF Miller capacitance of the 300B grids, not a loudspeaker. This puts them in the most linear operating region since they are essentially unloaded triode-mode power tubes, very different than the heavily loaded Class AB ultralinear of classical Golden Age amplifiers.

If we chose, we could have a switch that would alter the quiescent current for the driver section, so 6V6’s could be accommodated as well. All of these tubes have the same pinout, but the optimal current for the 6V6 is in the 25~30 mA range, while the 6L6 and KT88 are happiest in the 50 to 65 mA range. (By comparison, the optimal range for a 6SN7 is 8 to 10 mA per plate.)

If the switch is in the wrong position, the 6L6 and KT88 would be under-biased and have increased distortion, but no harm done. If it is the wrong position with 6V6’s installed, though, they would burn up in minutes, and if one shorted, would take out the interstage transformer as well. So a visit back to the factory, all from one user mistake. You don’t want to give the user the power to damage the amplifier, just from a single switch setting.

A limited range switch to trim between 6L6 and KT88 would not harm either (in the wrong position), so that’s an option. But optimizing current for the KT88 still gives plenty of current for the 6L6, and we’ve found these types aren’t all that sensitive to current settings, just so long as there is enough.

A general rule-of-thumb in vacuum tube design is setting the plate dissipation somewhere between 50% and 75% of max rating, with 65% to 70% usually considered a good balance between tube life and overall performance. Unfortunately, many KT88 amps take the tubes right through their ratings, so new types like KT120’s have been created for these amps.

Last but not least, although four driver tubes are required, they don’t have to be matched quads. Matched pairs are fine. It doesn’t matter if the drivers in one amp are at 50 mA each and the drivers in the other amp are at 55 mA each. The difference can’t be measured and won’t be heard.

So if you have a stash of priceless MO-Valve KT66’s, which were manufactured in matched pair sets for the amps of the day, go ahead and use them! They will be operated very conservatively (far more so than most power amps).

There are a pair of pin jacks on the top panel to check if plate voltages match. Measure across the pin jacks with a basic DVM, set the DVM to measure DC, and if you see 3 volts or less, you’re good to go.

Let’s do a little quick math to see how hard the driver is working, compared to an output tube. The gain of a 300B is 3.9, according to this Western Electric data sheet. The interstage transformer has a moderate gain of 1:1.2, so the net voltage gain of the output section is 4.68. In power terms, that’s a ratio of 21.9, or 13.4 dB.

The driver is working at 0.2137 times the voltage swing of the output stage, into an open load that only has 60 pF of capacitance, and only draws current when the 300B grids pass into the A2 region at 80 volts swing. The driver can push the 300B grids at least 20 volts positive, and surprisingly, the 300B remains linear in this region, with no visible transition when it goes from A1 (negative bias) to A2 (positive bias).

There are no charge storage effects, unlike conventional RC coupling, so the transition between A1 and A2 is seamless, and recovery is immediate on departure from the A2 region. In addition, if there are any nonlinear grid impedances from the 300B as the drive voltage goes up and down, it is swamped by the orders-of-magnitude lower plate impedance (700 ohms) of the driver section.

There are further subtle benefits of transformer coupling. The 300B grids do not go into the positive-grid region at the same time: they take turns. This means both driver tubes are available to drive whichever grid needs current, not just one. And the paired drivers aren’t just paralleled, but in a Class A balanced circuit, so most distortion is cancelled.

This is important with an output tube with distortion as low as the 300B; the driver should be as clean as possible, yet capable of peak-to-peak 200 volt swings. At 30 kHz (which is a 38 V/uSec slew rate).

I went through that with the Amity amplifier in the late Nineties. Nice square waves vs sonics. The amp sounded best with NO grid resistor, and the overshoot was a non-factor. Your amp may be different, of course. Keep in mind that music sources will never excite the overshoot, since it’s all in the far ultrasonic, and most studio mikes are all gone by 25 kHz.

The tube doesn't need that 510K resistor. The DC impedance of the IT is a few kohms at most, and that stabilizes the tube at low frequencies. If the driver is really unstable, maybe the 510K resistor helps, but that's usually the function of a grid stopper resistor, something quite different.

Now’s a good time to discus the difference between overshoot in feedback vs non-feedback amplifiers. Despite similar appearance on the scope, they are caused by completely different mechanisms.

* Overshoot in a feedback amplifier is quite malign, since it indicates the onset of oscillation, something that can destroy the amplifier and the speaker it is connected to. It is caused by the amplifier running out of phase margin, possibly the result of a reactive load, but also the result of design oversights in the feedback loop.

* Overshoot in a non-feedback amplifier is quite different. Now, it might be the result of high-transconductance tubes self-oscillating, but this can prevented by grid-stopper resistors and good layout practices. Normally, though, it is merely transformer overshoot, the result of phase shift at the edge of the passband, and has nothing to do with stability. That is what we are seeing here.

Something to be considered about ultrasonic behavior in the time domain: if there is no spectral content in the frequency range of the overshoot, it will never be stimulated in the first place. It never happens.

This is the difference between overshoot in a feedback amplifier and a non-feedback amplifier: in a feedback amplifier, it is a warning sign, like the LOW OIL light in a car. You ignore it at your peril. In a non-feedback amplifier, it has nothing to do with stability, since there is no feedback loop to induce oscillation. It is simply the behavior of passive parts, in this case, the input and interstage transformers.

As you can see, Cinemag has done a very nice job here. (Same photo as above, just tidied up a little.)

As a minor diversion, I should describe the "Golden Age" amplifiers I keep referring to. This aren’t just the amplifiers made in the Fifties and Sixties; it describes the majority of PP tube amps made since then, including today.

There were only a few basic Golden Age circuits, or topologies, as we like to call them. (Topologies omit circuit values, but are easily worked out once you know the tubes.) The first was the Williamson of 1948, but it had the drawback of marginal stability. Still, it dominated the US market until 1955 or so, when the much simpler Dynaco variant came in. (The Dynaco topology simply omits the driver stage of the Williamson and uses the phase splitter to drive the output tubes. More distortion but more stable.)

The Mullard became the prototype of many tube amps as the better-performing alternative to the Dynaco circuit, and is still widely used today. Let’s walk through it.

There’s a high-gain input tube, typically either a 12AX7 or a pentode like an EF86. This is direct-coupled to one half of a differential stage, with the other grid AC-coupled through a cap to ground. Because the grid of the diff stage is at 150 volts or so, the cathode is a little bit higher, maybe 155 volts. This requires a large value resistor that goes all the way to ground, so the diff stage is frequently called a "long-tailed pair". A current source could replace the resistor, but in practice, the performance is very similar to a current source, so it’s rarely done even in modern amps.

The diff pair are a pretty good phase splitter, and unlike the split-load inverter of the Dynaco circuit, audio-frequency balance is not too sensitive to load. It also has more drive capability than the split-load inverter, and unlike the split-load inverter, it has some gain, too. So a win all around.

And we’re not talking about a lot of parts here: 3 triode sections, and the output pair. A Dynaco is even simpler, with 2 triode sections, and the output pair. The only coupling caps with either circuit are between the grids of the output pair and the preceding circuit, so not really complex, and simple enough that a stereo chassis, running off a single B+ supply, is quite practical.

The point of the high gain (in the input section) is to give feedback something to work with. Feedback requires "excess gain" to work its magic; you need 20 dB of excess gain to get 20 dB of feedback, which will reduce overall distortion tenfold. In a pentode or ultralinear connected amplifier, the output impedance is way too high to use with most speakers. The feedback really comes in handy here: 20 dB of feedback reduces output impedance tenfold.

What limits applicability of feedback is loss of stability if too much is used (I’m not going to get into Nyquist Stability Criteria here, nor phase margin, settling time, etc.) In other words, if we slap in another gain stage and try for 40 dB of feedback, it will just oscillate. At full power. And take out a tweeter before damaging itself and letting the smoke out.

A more clever approach is wrapping local feedback around the most distorted stages, like the output section, and then add overall global feedback on top of that. This was done in the McIntosh, Citation II, and a few other amplifiers. This really gets the distortion numbers down, but clipping can get ugly, and settling time from transients can be an issue. Multiple feedback amplifiers can be quite sensitive to operating conditions. It’s more often seen in modern transistor amps as "two-pole compensation", and is not trivial to design.

Note: To puzzle out a schematic, by convention, signal flow is left to right, just like you’re reading this. To see what a tube is doing, look what the grid (the dotted line) is connected to. Often, there will be a coupling cap, typically 0.1uF. If it is much smaller than that, like 30 mmF or 30 pF, it is bypassing RF or has something to do with stability. Larger caps are cathode bypasses or power supply. The plates (the flat-topped dingus) is the output of the tube and typically heads to the right side of the schematic.

You usually have to stare at a phase splitter quite a while before the function becomes obvious. One side is quite simple, coming directly from the input tube, but the other side can be pretty weird. A diff stage can be puzzling, because the DC connection is a high-value resistor going to the other grid, and the AC connection just goes to ground through a 0.1uF cap. The "other half" is actually driven from its cathode, not the grid.

What gives away a split-load inverter, or "concertina" stage, are the equal cathode and plate resistors. This is a dead giveaway you are looking at an inverter, since no other tube stage uses equal resistors ... for one thing, it’s kind of useless for anything else, since gain is a bit less than unity.

I leave the "floating paraphase" as an exercise for the reader. I kind of like them, actually, because current drive for the power tubes is pretty good, although balance is only so-so.

I agree. This is a stable topology, taking full advantage of specialty transformers designed by two of the world’s top designers, and using vacuum tubes that are in current production as well as ample NOS stocks.

As mentioned earlier, it’s a very simple signal path, with only transformers and vacuum tubes, and fully balanced from input to output. Zero feedback, with the audio signal only propagating in the forward direction.

And I am very much looking forward to fellow enthusiasts hearing the Raven preamp and Blackbird power amp. I’ve been a voice in the wilderness for about twenty-five years ... neither a member of the SET fraternity (well, maybe on the edge of it) nor mainstream Audio Research/Jadis/Conrad-Johnson push-pull pentode big-watt amplifiers dominating the hifi shows. A handful of people built the Karna amps, but many abandoned the difficult project halfway through.

Don was one of the very few who persevered through two years of building prototypes that were far off the beaten path of mainstream tube gear. He’s had plenty of hands-on experience with the fiendishly difficult Citation II, the most complex amplifier of the Golden Age, and his own designs, the Valhalla (6L6) and Kootenai (KT88).

Of all the people I know in the industry, Don is the most qualified to honestly tell me what is unrealistic and pie-in-the-sky, and what is practical and a good solution. He’s been there and done that. Oh, and he has good taste, too, which isn’t that common in the industry.

You might think I’m being snarky about the "good taste" but I am perfectly serious. The industry has plenty of competent engineers, and whole hifi shows filled with high-powered marketers, but good taste? It’s not all that common, and I’ve been in the industry since 1973.

Another walk down Memory Lane. This time, we’ll go into the late Forties, when the Williamson burst on the scene. This English design wiped out all other designs in the USA until about 1955 or so, with the exception of the McIntosh and a few others.

How does it work? There’s an input tube, typically a triode like the 6SN7, direct-coupled to a split-load inverter, also called a "concertina" stage. This always has identical plate and cathode resistors, and gain a bit lower than unity. The plate output drives the upper half of the push-pull amplifier, while the cathode drives the lower half. Despite appearances, the voltages on top and bottom are equal and opposite ... provided the total loads match, as well.

The inverter is then cap-coupled to a separate push-pull driver stage, which is sometimes also set up as a differential stage, depending on the resistance presented to the common cathodes. High impedances move it towards a differential stage, with the limit being modern constant-current sources. 6SN7’s were typically used here, with later designs replacing them with 12AU7’s (which typically have more distortion).

The drivers are then RC cap-coupled to the output tubes in the usual way. The drawback of a classical Williamson are the two stages of cap coupling, which can introduce low-frequency instability unless the output transformer has extremely wide bandwidth. The Partridge transformer specified for the original design had one of the widest bandwidths of any output transformer ever made ... but lesser transformers introduced stability problems, sometimes "motorboating" at low frequencies, but more commonly long recovery times from overload.

The Dynaco, introduced in the mid-Fifties, took the drastic step of deleting the driver stage and its associated RC coupling, and driving the output tubes from the RC-coupled phase inverter. Although the open-loop performance was quite poor, rolling off around 100 Hz and 7 kHz, the 20 dB of feedback nicely corrected it, since the input section used a high-gain pentode and there was plenty of "excess gain" to drive the feedback network.

The Dynaco had the advantage of being the cheapest of all to build; a combined pentode/triode, the 7199, took care of the entire front end, and all that was left were a pair of EL34 output tubes and an output transformer. In addition to Dynaco, many receivers used this approach as well. It was simple, saved money, and saved space, which was at a real premium in a low-profile AM/FM stereo receiver.

Receivers in the early Sixties (Fisher, Scott, Sherwood, Harman-Kardon, etc.) all had Bass and Treble tone controls, an AM and FM tuner with two different IF strips, an FM multiplex stereo decoder, a stereo power amp with at least 20 to 35 watts/channel, and last but not least, a stereo phono preamp. All with vacuum tubes, in a very crowded chassis, with marginal ventilation and caps of much lower quality than we have today.

We don’t see many Williamson amplifiers today. The dominant PP-pentode designs are Mullards and Dynacos, depending how price-sensitive the amplifier is. The monster tube amps with 4, 6, or 8 output tubes per channel typically throw in a dedicated cathode-follower section to drive all those grids ... sometimes one cathode follower to drive them all at once, or preferably, each output tube gets its own cathode follower. The RC coupling is then moved to the input side of the cathode follower, and the CF directly drives the grids of the output tube(s). This easily provides independent biasing of each of the output tubes, which is important when that many tubes are used.

The three-tube 6SN7 circuit board for the Dyna ST70 converts it to a Mullard circuit, with lower distortion and stronger drivers. Since nearly all the ST70 circuitry is on the single circuit board (for both channels), swapping that board basically gives you a new amplifier ... while retaining the power supply, chassis, and transformers. Lots of ST70 variants, since so many were made and are still kicking around. And the output transformers are pretty good.

Of course, if you are replacing the power transformer and upgrading the power supply, you might as well build on a new chassis, and have an all-new amplifier. Nothing wrong with a 6SN7 Mullard circuit and modern power supplies ... that will take you into the $3000 to $10,000 quality bracket right there.

And of course a part-Mullard circuit is perfectly acceptable for a PP 300B amplifier. Unlike a PP pentode amplifier, though, you need about two to three times as much voltage swing in the driver, so a Dynaco circuit is definitely not the right choice.

A Mullard PP 300B works as follows: input tube direct-coupled to a long-tail pair (or CCS) of triode-connected 6V6 drivers. These in turn are connected to a PP interstage transformer with a modest step-up ratio, between 1:1.4 and 1:2. The interstage then drives the PP 300B grids. This would be a non-feedback amplifier, so good power supplies are required. I would imagine a number of the PP 300B amplifiers already on the market use this topology.

Back in 1993 when I was trying many different amplifiers on the newly completed Ariel speakers, I came to the conclusion that a stock ST70 was the minimum acceptable standard for hifi. Most transistor amps fell below this mark, and most tube amps exceeded it. Restored Mullard designs did very well, as anyone might expect.

What I did not expect was the performance of the Ongaku and the Herb Reichert Silver 300B. I expected all SETs to be terrible, but those two were the best sound ever on the Ariels (which are 92 dB/meter efficient). After hearing several other SETs, I was struck how variable they were. A few were superlative, almost otherworldly, but many others were pretty bad. One was so terrible that Karna and I just burst out laughing ... it sounded like a 1960 transistor radio left out in the sun too long. That’s how variable SETs are ... all over the place.

By contrast, a competently engineered Mullard with a decent power supply is almost guaranteed to sound pretty good. And the best ones are superb.

There's no feedback, and there's gobs of 2nd-harmonic distortion. Parts coloration works for or against that 2nd-harmonic distortion. Also, because power-supply rejection is zero, power supply coloration is right in your face.

Driving DHTs takes two to three times the swing of pentodes or beam tetrodes. This exposes driver nonlinearity as well. Even worse, in some designs, driver and output distortion partially cancels ... but only at certain power levels. So the distortion signature is strongly level-dependent, which is very undesirable.

The problem with two successive stages that are balanced and DC-coupled to each other is that DC drift is a big deal. A 1 volt shift on a 150 volt plate is normally inconsequential, but becomes a serious concern when the grids of the following stage have a 1 volt offset between them ... which is what DC coupling does.

A Mullard sidesteps this by direct-connecting the plate of the SE input stage to ONE driver grid. The other grid (of the driver) is AC-connected to ground through a 0.1uF cap and DC-connected to the other grid via a 100K ~ 220K resistor. As a result, the two driver grids always DC-track each other.

By contrast, if the Mullard input section is replaced with a DC-connected balanced or diff stage, then DC balance and drifting of the first stage becomes critical, requiring a servo circuit to always keep the plates of the input tube exactly matched. No thanks.

The Blackbird is fully balanced, input, driver, and output, with DC balance issues resolved by using transformer coupling. Transformers are incapable of passing DC from primary to secondary, since the coupling is magnetic. Charge/discharge issues associated with capacitors, as well as potential coloration, are also avoided since cap coupling is not used anywhere in the forward path.

The hard part is getting transformers of high enough quality ... this is where working directly with the transformer designer, making them a part of the design team, is essential. These are not off-the-shelf parts.

A minor side benefit is avoiding turn-on pops and clicks, since the circuit remains balanced in all modes of operation, without relying on servos to maintain balance.

As mentioned above, a part-Mullard is great way to build a PP DHT amplifier. Not too complex, a well-known circuit that behaves predictably, and capable of scaling up the driver so it has enough power to motivate DHT grids.

Scope pr0n. Same pix as earlier, just more zoomed in.

Don Sachs scope photo of Blackbird at 30% power (8 watts) at 1 kHz. Zero feedback, with no grid resistors to "trim" the response.

A quick note: Don and will not be making the part-Mullard or the suggestions made by others in this thread. The Raven and Blackbird are where our attention is, and that’s where it will stay. Our focus this year, and the next, is getting production moving smoothly, making sure the Raven and Blackbird are reliable as possible, and growing the customer base.

On my part, I’ll be completing the long-awaited "Beyond the Ariel" speaker project over on DIYaudio, with the assistance of Troy Crowe in Canada. I do my best work collaborating with others, and Gary Dahl, Bjorn Kolbrek, and Thom Mackris have made a real difference on that project. I’ve been meditating on an appropriate name for the speaker, and "Phoenix" feels right, considering how many times it has been re-born.

Try removing the 56K load resistor entirely and give it a listen. You might like it better. Remember, with conventional signal sources, there is almost no ultrasonic content, so the ringing in the transformer is never stimulated. And getting rid the load resistor has no effect on circuit stability or DC stability, since the grid sees the DCR of the transformer, which is a few K at most. All the grid resistor does is make square waves look pretty. It has no other function.

Now, in a cap-coupled circuit, it is absolutely necessary, since grid current, as small as it is, has to go somewhere. With transformers, it just goes through the secondary, and the current is so minuscule there is no effect on the core.

Alex, remember, you cannot harm your amplifier if you completely remove the grid-load resistor. With a transformer, runaway from DC instability is impossible. There is always a DC path through the secondary ... as long as the secondary is intact.

If you have an oscillation lurking in there, at some high frequency like 5 to 20 MHz, that's a different story, and unrelated to the value of the grid resistor. 5 to 20 MHz oscillations, even at a very low level, will absolutely make the sound bright and unpleasant. If they are -40 dB down, you will never see them on a scope ... that's no more than a trace width. You need an RF spectrum analyzer to sniff out the little monsters. They look like little spikes rising out of the noise floor (which should be very smooth).

If you suspect this, you need a grid-stopper in series with the grid pin, like 100 to 500 ohms of carbon-comp resistor soldered no further than 1/2" from the grid pin. (NEVER use a wirewound for a grid-stopper.) That will kill self-oscillation.

I would try a grid-stopper first before futzing around any more. The only way you can solidly rule out self-oscillation is use an RF spectrum analyzer that's good to at least 20 MHz, preferably 100 MHz. These things aren't cheap, and only have one use, chasing out RF nasties. Low-level RF oscillations are surprisingly prevalent in high-end audio equipment, with poorly designed regulators as the usual culprit.

Try the grid-stopper first before anything else. After that, play around with various value of grid resistor, including nothing at all. It should not be sounding bright, unless something is wrong.

Hi there, Pindac!

The decision to export out of North America is largely up to Spatial Audio. Don and I are the technical advisors and consultants, but exporting is a business decision. We can suggest and advise, but we do not make the final decision.

As Don has mentioned, exports to Europe would have to meet rigorous EU safety standards, and an EU servicing center would be wise. The EU is a big place with many technical and legal requirements that are quite different from the North American market, which has one dominant language, safety standards, and electrical power. If you can sell it in Los Angeles, you can sell it in Toronto, and everywhere in between.

Sure, there are small audiophile manufacturers who import on a "grey market" basis into the EU and the UK. That’s fine until you get caught. Don, myself, and the team at Spatial intend to be on the straight and narrow when it comes to regulations ... we are not Tesla, Apple, or Microsoft, with armies of lawyers to smooth the path into new markets.

Back when I was at Audionics in the Seventies, we eventually surmounted the many EU technical regulations and sold our products into the European market. But it took several years, and we didn’t attempt it until we had significant sales volume in the domestic market. Sure, it’s easy to order a Monolith power transformer with multiple input voltages. That doesn’t make the finished product legal to sell in the UK or the EU.

The unusual thing about the US market is that it is really easy to sell into ... tariff rates are some of the lowest in the world, technical safety requirements are not too severe, and the market is huge and easy to serve. Markets everywhere else are different ... more fragmented, higher tariff barriers, multiple languages, many different technical standards, and other obstacles.