Happy Halloween!
With our children having long left the nest, my wife and I no longer felt the urge to crave pumpkins. Besides, with so many neighbors exhibiting so much effort and creativity in the layout of their scary Halloween decorations, we had no hope of winning in a competition. Thus, we resigned ourselves to not being the overly festive types.

But, amazingly enough, Amazon came to the rescue, so we now have three skeletons on our porch, cat and dog and human skeletons, albeit small ones and a huge garage-door Halloween banner—all for less than $60.

This cute painting went over so big that we now have a copycat neighbor two houses down, using the exact same banner. (Their ten-year-old son loves cats.) It took me two hours to install it; my neighbor, over three hours. It wasn't a fair race, however, as I held the advantage of not having any "helpers," whereas he had three.

As there are at least another dozen cat-themed Halloween banners at Amazon, I plan on using a different one each year until we have the complete collection.

One day, I will carve the triode-symbolled pumpkin, possibly using a transmitting tube, such as the 211 or 813 or 845, to provide the internal light, if the pumpkin is large enough. (Only the heater/cathode need be powered.)

Resuming Where I Left Off
My last post, offered several ideas on the possible further exploiting of class-D amplifiers, which included a direct-drive ultra-high-voltage class-D amplifier to drive electrostatic loudspeakers. Returning to Earth, I end the post with the idea of combining two electronic technologies: vacuum tubes for their musicality and a solid-state power buffer for their low cost and robust current delivery, i.e. puissance. In other words, a hybrid power amplifier. What saved the idea, however, from the damning charge of DGNNAHA was that my balanced solid-state buffers and amplifiers were not conventional in any sense. Basically, I wanted to exploit the solid-state's advantages of low voltage and low cost but undo its harmonic shortcomings via the tube-based input stage. What is DGNNAHA ?

Almost 20 years ago in my Post 111, I came up with the DGNNAHA label (which stands in place of Dear God No—Not Another Hybrid Amplifier!), as I so disliked seeing so many lame and similar hybrid power amplifier designs in audio-construction magazines. While searching for my first mentioning of DGNNAHA, I encountered Post 291, which held the following observations:

"…let's recall a familiar experience. Your favorite DIY audio magazine (or web-site) loudly proclaims: A New Hybrid Amplifier. Whereupon, your soul plunges. You brace yourself for yet another tedious SRPP stage followed by power MOSFETs, a feeling matched by picking up a five-year old Time magazine in your dentist's waiting room, discouraging and disheartening in the extreme—even a recent issue of Time magazine is a dreary affair, but a five year old one is miserable beyond endurance.

"Because the SRPP-input-stage-MOSFET-output-stage hybrid is so simple, so obvious, for the last four decades it has been constantly reinvented by earnest audiophiles who imagine that they have discovered the Holy Grail of audio: all the sweetness and fluidity of vacuum tubes and all the heft and command of solid-state devices. The entirely expensive sounding results of wedding a little bit of expensive glass with a lot of cheap silicon. If only. Possibly the saddest short phrase in English. (The disillusioned defense attorney laments, "If only my clients weren't so guilty." Yes, indeed, if only.)

"Thus, DGNNAHA stands in place of Dear God No—Not Another Hybrid Amplifier!

"Imagine that a friend tells you that he is writing a great novel that—brace yourself—tells the story of a prostitute who, surprisingly enough, has a heart of gold; or a fabulously new story about two policemen, partners, one of whom is super nasty and the other, amazingly enough, super nice; or a radically new story about how a teenager's parents don't understand him... A cliché becomes a cliché because many like it.

"Some start life as a raw, powerful utterance, such as "she wore her fingers down to the bone," a gripping, horror-filled image, which when first voiced must have shocked and sickened: the image of white bone protruding through red, torn flesh. No more. Like the five-year old Time magazine in your dentist's waiting room and like an old worn-out 6SN7 that refuses to move the tube-tester's meter, it is now bereft of all vitality and vigor. Other clichés, in contrast, begin as garbage and continue as garbage, such as "There's a fine line between genius and insanity." Was Aristotle, Leonardo, Shakespeare, or Johnny Von Neumann insane? Alas, for the cliché, no fine line existed. Indeed, I am more likely to believe that there's a fine line between stupidity and insanity, having encountered that strange muscular, all-defiant stupidity that is not born of ignorance and timidity, but of supreme certainty and self-confidence. Prisons and insane asylums brim with those chock full of high-self-esteem.

"Well, in my view, the SRPP cascading into the MOSFET output-stage is just that kind of electronic cliché, as obvious as it is wrong. We deserve better. The Hollywood legend, Samuel Goldwyn, put it best: "Let's have some new clichés." I wholeheartedly agree, but with some conditions of satisfaction.

"Conditions of satisfaction," what the hell is that?

"Simply put, conditions of satisfaction (COS in its legal acronym) are the criteria—the standards, benchmarks, guidelines, requirements, principles—that a new hybrid amplifier must pass in order to bypass the getting of the shameful DGNNAHA label. Here is my list:

1) No gratuitous tubes

2) No safety issues

3) No huge compromises in performance

4) No weirdness for the sake of weirdness

"Wow, John, wouldn't that list kill all the fun?

"In one of the few lovely passages in Immanuel Kant's writings, we find

“The light dove, in free flight cutting through the air, the resistance of which it feels, could get the idea that it could do even better in airless space."

"The poor bird does not realize that without air resistance, it would plummet to the ground. Only the mentally flabby fear the resistance imposed by rules. And surely, this list is not too onerous.

"Tubes should never be expensive replacements for LEDs. You know the unease you feel, if only subconsciously, when you are watching a movie which holds a gratuitous nude scene, a scene unnecessary, unwarranted, and unjustified by the film's story, serving only as a sneaky showcase for the leading lady's pert breasts and rollickingly round rump. Well, that is the sort of unease I feel when I see tubes being used gratuitously. Tubes that replace resistors needlessly; tubes that could be (and probably should be) replaced by a FET or transistor. Thus, if a tube appears in a hybrid circuit it must do something meaningful, not just lend a glowing, glistening luster to an otherwise dim, dull effort.

"On the other hand, there is the topic of tube porn, wherein tubes, used often superabundantly or whimsically, are the point, the whole point, the only point. In porn, nudity and sex are never gratuitous; they are essential, and plot and character development would seem gratuitously out of place. John Atwood and Christian Rintelen and I have had interesting discussions on this topic of tube porn, the outlandish display of vacuum tubes, for the sake of glowing glass, not necessarily sonic glory. I am much more disposed to greet tube porn with a smile because it is at least honest, something the serious film with gratuitous sex or nudity is not. In addition, in tube porn, cheap solid-state devices are never added to lower the price; indeed, the greater the expense, the closer to an all-tube lineup, the more decadent and opulent the results."

Yes, that was a dang long quote. Since pixels are effectively free, however, only a fool would be parsimonious in expending them—besides, it's fun (embarrassingly too fun) to quote oneself. Actually, I was about to leave only a link to this post, but I know that fewer than one in a hundred ever follows a link. Moreover, I am quite proud of the post. In fact, I wonder what type of coffee I was drinking when I wrote this post (Sumatran, perhaps?), as it strikes me as particularly fine. No doubt, some readers wonder which whiskey I poured into my coffee that day and further wondering what the hell do Aristotle, gratuitous nude scenes, and porn have to do with audio.

My view is that you can get away with having a distinctive, if not a snappy, prose style, as long as few recognize that is what it is. This echoes Raymond Chandler's famous observation in a 1949 letter to his British publisher, Hamish Hamilton:

"My theory has always been that the public will accept style provided you do not call it style either in words or by, as it were, standing off and admiring it."

Wasn't I just standing off and admiring it? Indeed. The shame. Returning to audio, my latest thoughts were along the lines of using a single-ended-to-balanced, tube-based input stage and a balanced-bridge, solid-state output stage.

Although the bridge output-stage configuration is what most modern class-D amplifiers use, originally it was used in class-AB output stages. The logic is: since a loudspeaker is an intrinsically differential device, why not drive it in a balanced, i.e. differential, fashion? The bridge configuration allows us to get four times the amount of power output from power-supply rail voltages than what we would get with the same rail voltages with the conventional single power amplifier. For example, with +/-24Vdc power-supply rails, a single solid-state power amplifier might swing voltage peaks of +/-20V, which into an 8-ohm load equals 25W. In contrast, two of the same power amplifiers in a bridge configuration would yield +/-40V peak voltage swings, which equal 100W into 8-ohm loudspeakers.

And since solid-state output devices are not current shy, the bridge configuration allows us to halve the B+ voltage, but we must double the net idle current. With the bridge configuration, current delivery is of key importance, as each amplifier effectively sees half the load impedance, 8 ohms appearing as 4 ohms, something no tube-based OTL amplifier could tolerate. What I do not like about the bridge configuration is that, like the Circlotron, it delivers a very push-pull signature, with far greater attenuation of even-order harmonics, leaving only the annoying odd-order harmonics.

For example, this simple bridge output stage enjoys two single-ended, class-A structures—constant-current source loaded no less, but as it runs the two MOSFETs in anti-current phase, it gives rise to the severe push-pull harmonic structure.

My intended workaround is to pre-flavor the drive signals with tube-based circuitry that delivers ripe single-ended harmonic signature.

Here was the design's overview:

In other words, I want the tube-based frontend to purposely undo some of the balanced output stage's push-pull harmonics by introducing heavily single-ended harmonics first. Not all tube circuits can deliver the desired results, sadly. For example, a tube-based long-tail phase splitter also converts a single-ended input signal into a balanced output pair of signals, but it, too, delivers more of a push-pull harmonic structure.

Ditto, the paraphrase phase splitter.

In both these circuits, the two triodes run in equal anti-current phase. In contrast, the split-load phase splitter holds only one current phase.

A super-intriguing variation on the split-load phase splitter is the totem-Pole phase splitter:

This circuit delivers a lovely single-ended cascade of harmonics. See my Post 444 for more details on its operation and the clever positive-feedback variation. Extra gain, however, is something we do not need in this application. Here is the totem-Pole phase splitter with a 6DJ8 tube and a lower B+ voltage.

This circuit delivers a balanced output gain of 1:50 peak-to-peak (+34dB). The PSRR is poor. With the addition of some Aikido Mojo we improve the PSRR substantially.

The 30k resistor and 47µF capacitor inject enough power-supply noise into the input triode's cathode to force an equal amount of noise at each output, which the following balanced power buffer stage will ignore. When measuring the PSRR of a balanced output in SPICE simulations, we use the following method: we take the Log of the net difference between the two outputs leaked power-supply noise and multiple the resulting number by 20 to get the answer in dBs.

    20 x Log(Inverting Output – Noninverting Output)

If both output leak the exact same amount of power-supply noise and in perfect phase alignment, the result will be infinite PSRR, although each output might exhibit a leakage of power-supply noise of 100%. The assumption here is that the following balanced amplifier will use its high common-mode rejection (CMRR) to ignore the equal amount of identical power-supply noise riding on the output signals. (Preferably, both outputs would only contain a fraction of the power-supply noise.) We adjust the Aikido Mojo resistor's value to attain the state of equality. Here is a simpler example of adding an Aikido resistor.

Without the Aikido Mojo, the PSRR is a poor -9.6dB at 100Hz. (By the way, this circuit delivers a lovely single-ended harmonic structure.) This Aikido Mojo technique differs from the approach I used in the circuit from my last post:

The Aikido Mojo is found in the 432-ohm cathode resistor value. In SPICE simulations, I discovered that this value brought the phase splitter's two outputs into alignment. By the way, since this circuit uses a lower B+ voltage and delivers lower gain, it will work better within a lower powered hybrid power amplifier or power buffer. For example, here is a suitable 60W class-A bridged output stage.

Each MOSFET idles at 2A, which implies a peak output current swing of 4A, which further implies 64W of output into an 8-ohm load. Yet, this balanced power buffer runs off only ±20Vdc power-supply rails. Mind you, at idle this output stage dissipates 160W, so its efficiency is only 40%. Class-A isn't cheap. Another possible circuit is the Circlotron.

Circlotron Circuits
I cannot remember when I last covered a Circlotron circuit. Too long ago, no doubt. An auto-biased, class-A Circlotron would make a compelling output stage for a ripe tube-based input stage. Auto-biasing is easy, if we restrict ourselves to class-A operation, as the average current flow is identical with the idle current flow. This is very much like a constant-current source circuit, so it needs the same three things: a voltage reference, a current-sense resistor, and a DC gain stage to apply the negative feedback.

The 0.47-ohm resistor senses the current flowing through the MOSFET. The 2N4401 NPN transistor provides both the voltage reference in the form of its emitter-to-base voltage drop and the needed DC gain stage. If the MOSFET draws too much current, the transistor's current conduction increases, causing its collector voltage to fall, thereby lowering the MOSFET's gate voltage and current conduction. Negative feedback in a nutshell. Of course, we could get more elaborate by using an IC tight-tolerance voltage reference and an OpAmp to set the MOSFET's gate voltage, but the single transistor is compelling due to its simplicity. Unfortunately, its emitter-to-base voltage drop is a bit too high.

My workaround is to add a resistor to the B+ voltage to create a lower effectively voltage reference. In the following example, the current-sense resistor voltage drop is only 0.22 volts.

The leftmost MJE15032 NPN transistor's base still sees 0.443Vdc. Note that if the B+ voltage is increased, the idle current through the MJL3281A decreases, which is quite desirable in an output stage. Here is the graph that shows the current flow versus the B+ voltage.

Ideally, we would get a straight line whose slope would ensure constant-power dissipation by the MJL3281A, but the decline current flow beyond the nominal 20Vdc B+ voltage is a feature. Here is a power buffer Circlotron based on this auto-bias configuration.

Click on Schematic to see enlargement

The 20Vdc floating power supplies and the idle current flow of 1.26A per side, along with the necessary voltage overhead required by the transistors, limits the power output to only around 16W for 8-ohm loads. This puts the efficiency at about 32%, 18% short of the theoretical maximum of 50% for strict class-A operation. Of course, if we up the number of output transistors and increase the floating power supply voltage, we can get far greater power output.

The big question here is, how well did the auto-bias feature work? Very well indeed:

In this SPICE simulation, I varied the output transistors (MJL3281A) temperature from 25C to 125C, yet the total idle current flow increased by less than 10mA. Excellent. The next SPICE test I ran was the frequency response test.

With a -3dB high-frequency cutoff beyond 1MHz, we needn't worry about bandwidth. By the way, the Circlotron actually exhibits a gain of 2 (6dB). Well, what about distortion? At 1W of output, the Fourier graph doesn't look too bad.

In fact, its crazy good for a feedback-free circuit, save for the relatively higher 3rd harmonic. At full output, 16W, the push-pull harmonic structure reveals itself.

The THD is around 0.1%, but all the tasty even-order harmonics are greatly suppressed, leaving the irritating odd-order harmonics overstressed, creating the typical push-pull sound. Will the tube-based frontend be enough even-out the harmonics? Perhaps. My next idea was to include some negative feedback in the design by using a compound circuit.

The compound configuration places the PNP transistors under the control of an NPN transistor, thereby greatly increasing the linearity, but at the cost potential high-frequency instability and rougher clipping. In addition, it works best in class-A operation, as the transition between output devices in class-AB mode are sharper. On the other hand, the compound configuration is itself a form of auto-bias, as long as the controlling transistor is thermally isolated from the output transistors. The following SPICE-generated DC-transfer graph reveals the tighter auto-biasing.

In spite of a span of 100C, the idle current increases by less than 2mA (the graph shows only one side, not the total). The Fourier graph also displays an improvement.

Yes, the THD is improved, and the 3rd harmonic is further suppressed, but look at the 4th harmonic up to the 9th. They are worse. The negative feedback made these worse, but did decrease the output impedance and improved the auto-biasing. at full output, thing actually look better.

All the odd-order harmonics are further suppressed by 10dB compared to the first Circlotron. But as I had to place a capacitor at the controlling NPN transistor's collector to ensure high-frequency stability, the high-frequency bandwidth is truncated a bit.

Thus, I concluded that the compound arrangement was not the answer, in spite of some of its features. My next idea was to try output MOSFETs in place of the output transistors.

If nothing else, MOSFETs are far more robust than bipolar transistors in actual use, as they survive when their bipolar brothers go up in smoke. They are also cheap and modern production has improved to the point that they pretty much come pre-matched right out the shipping tube. I did worry, however, that their relatively high input capacitance would too greatly limit their high-frequency bandwidth.

With bandwidth beyond 1MHz, no worries were justified. Next, we test the auto-bias feature.

Certainly, good enough. Now the Fourier tests.

Surprisingly enough, the harmonic structure at 1W of output pretty was much is identical as the first Circlotron with bipolar transistors. At full output, the push-pull harmonics blossom and the results are actually worse than the first version.

Bummer. Or are they actually all that bad? Note that the 10kHz test tone frequency, not the 1kHz test tone used with the other variations. Okay, what to do? How about adding an OpAmp to the mix?

To make the OpAmp happy, I have lifted the input signal DC reference to half of the OpAmp's power-supply voltage (10.18Vdc). Why does this make the OpAmp happy? Ideally, the input voltage should cut the entire power-supply voltage in half, so 0V with a pips and 1/2 with a monopolar power supply, as the input stage needs all the voltage headroom it can get. Not so with a perfect OpAmp, but perfect OpAmps do not exist.

The MUR410G rectifier is used as the voltage reference. With such trivial current flow through the rectifier, the voltage drop is a mere 0.36V. By the way, I used the four 249k resistors to prevent the use of 5% or 10% 240k resistors by readers, as we do not want a DC offset at the output terminals for the loudspeaker. (The two 0.3 resistors should be tightly matched, for the same reason.)The way we sidestep the potential DC offset is to have both halves of the Circlotron draw the exact same amount of current draw. I used the OPA2107 and IRFP240 SPICE models in my simulations. The auto-bias feature works, as a flat line developed in the temperature sweep. Next comes the frequency response plot.

Why is there slightly less than unity-gain output? Note where the negative-feedback loop terminates, not at the outputs, but at the 0.3-ohm source resistors. The loudspeaker sees the two resistors effectively in series, which cause a slight insertion loss. Why not terminate the 0.1µF capacitors at the outputs? I tried that first, but it resulted in higher distortion. Speaking of distortion, here are the results for 1W of output.

Okay, that's stunning. No doubt, too stunning. But if we assume a one hundredfold amount of too good, it is still stunning. At 16W of output, it worsens, but only by a tad; moreover, we do not see the push-pull harmonic structure.

Mercy! That's fine. Since it is impossible for me not to seek further improvements from any circuit, one idea I had was that we could give the OpAmp a regulated B+ voltage.

The 12V voltage regulator cleans up the B+ voltage and works as an intrinsic current limiter on the output MOSFETs, as it limits the potential positive output voltage swing from the OpAmp. Of course, an RC filter and zener could be used configured as simple shunt regulator. In addition, the 12V might need to be replaced by a 15V regulator or an LDO adjustable regulator, such as the LM1086. By the way, the OpAmp must be unity-gain stable.

If we want more output power, we must increase the floating power supply voltage. With 48V floating power supplies, we can easily get 100W into an 8-ohm loudspeaker. The total idle current, however, must exceed 5A. Switch-mode power supplies with 48Vdc output voltage are readily available. Each should be able to deliver at least 3A of current. This would leave the heatsinks as the most expensive component of the amplifier, save for possible NOS tube purchases.

In contrast, building conventional following power supplies is made easier by most toroid power transformer holding two secondaries. For the 16W version, a 100VA transformer with two 15Vac secondaries should work well. The tube-based front end will need its own power supply, but as it draws so little current, a 30VA toroid would be massive overkill. By the way, I assume monobloc construction here.

Music Recommendation: The Haden Triplets, The Family Songbook
Charlie Haden, the well-known jazz double-bass player, has triplet daughters, fraternal triplets, Petra, Tanya, and Rachel. All three are excellent singers and form the group The Haden Triplets; with two albums out, the first (2014's The Haden Triplets) was produced by Ry Cooder.

If you expect something out there, avant-garde, experimental based their father being Charlie, you will be wrong, as the album is very traditional folk-country-Americana. But a hint of the jazz tradition appears in their cover of Kanye West's song, Say You Will. (Not Foreigner's song of the same title.)

Jazz, like America in the past, often takes the discarded, the forgotten, or trivial and transforms it into something magical. Nina Simone's fabulous cover of "I Put a Spell on You" is a great example. It was written and first recorded by Screamin' Jay Hawkins in 1956. Give the original a listen and hear hers, and you will instantly realize her genius.

The song, "Say You Will," is about calling an ex-girlfriend to come over to have sex, but she doesn't show up. The Haden Triplets cover transmutes the song into something achingly heartfelt and mournful. Magic.

Since three singers sing, we should be able to hear three voices, making this album a good test of our systems.

//JRB

Did you enjoy my post? Do you want to see me make it to post 1,000? If so, think about supporting me at Patreon.

Adobe AI Summary:

Here are the main points from the document:

Halloween Introduction

1. Author shares a lighthearted Halloween anecdote, highlighting creative neighborhood decorations and their own display using Amazon purchases.

Hybrid Amplifier Design Philosophy

1. The author discusses their ongoing exploration of hybrid power amplifiers that combine vacuum tube input stages for musicality with solid-state output stages for robust, low-cost power delivery.
2. Critique of generic hybrid amplifier designs labeled “DGNNAHA” (Dear God No—Not Another Hybrid Amplifier!), especially those using conventional SRPP tube stages with MOSFET outputs, which the author considers clichéd and lacking in originality.
3. Lists key design criteria for acceptable hybrid amplifiers:
• No gratuitous use of tubes
• No safety hazards
• No major performance compromises
• No design weirdness without purpose

Design Approaches and Technical Solutions

1. Explores balanced (bridge) and Circlotron output stage configurations, emphasizing:
• Benefits: more output power from low supply rails and robust current delivery
• Drawback: bridge configurations can overly suppress even-order harmonics, producing a push-pull, often less appealing harmonic signature
2. Suggests compensating for push-pull harmonic structures of solid-state stages by deliberately adding single-ended harmonic coloration in the tube frontend, favoring specific phase splitter topologies (like the totem-pole split-load).
3. Details PSRR (Power Supply Rejection Ratio) improvements using “Aikido Mojo” techniques and presents several practical tube and solid-state circuit examples, all evaluated with SPICE simulations for harmonic distortion, bandwidth, and thermal stability.
4. Evaluates multiple output stage variations:
• Class-A Circlotron using bipolar transistors (with and without compound/transistor feedback)
• MOSFET output stages with and without op-amp drivers
• Each approach is tested for linearity, bias stability across temperature, efficiency, harmonic structure, and output power
• Notes that MOSFETs are robust and modern types come well-matched
• Use of op-amp drivers can achieve stunningly low distortion
• Power and efficiency trade-offs are discussed for each variant

Practical Construction Notes

1. Considers power supply options (floating supplies, use of toroidal transformers) and monobloc amplifier construction. Recommends practical transformer ratings for different power versions.
2. Emphasizes the importance of correct resistor and component matching in preventing DC offsets at the output.

Music Recommendation

1. Recommends The Haden Triplets – “The Family Songbook” for both enjoyment and system testing, specifically highlighting their cover of Kanye West’s “Say You Will” for its emotional impact and usefulness in evaluating audio systems' ability to resolve voices.

This summary captures the technical focus on improving hybrid amplifiers by thoughtful circuit design and the balancing of tube 'musicality' with the efficiency and power of modern solid-state electronics.