A Very Modern Tube Headphone Amplifier

Once a discarded relic, over the years the humble vacuum tube has been rehabilitated in the arena of specialist audio. There are plenty of tube amplifiers now being manufactured, with a popular choice being headphone amplifiers that use a tube as a gain stage followed by an op-amp as a buffer with a low impedance output. This forms the basis of [Ratti3]’s amplifier, but with the added interest of a battery supply and a Bluetooth connection.

The tube circuit is a very conventional anode follower using an EF95 pentode. This provides plenty of gain and of course that “valve sound” beloved of audio enthusiasts, but suffers from an output impedance too high to drive a set of headphones. An NE5532 steps in for the op-amp buffer role, making for a very simple circuit. Power comes from a set of four 18650 Lithium-Ion cells with associated charger and balance boards, while a little switching boost converter provides the 100 volt HT for the tubes.

We’ve visited this type of amplifier before with a similar but much more rough-around-the-edges Chinese version. That had some astonishingly cheap Chinese tubes, but if we’re seeking better components it’s interesting to know just who makes tubes these days.

34 thoughts on “A Very Modern Tube Headphone Amplifier

  1. The guaranteed output short-circuit current of a 5532 opamp is 10 milliamps. If you have some magical 1000 ohm headphones it could be OK. The 5532 is not intended as a headphone driver and is not suitable for common 32-ohm headphones. This is why tube circuits use output transformers.

    1. Yes, the native output impedance* in most op-amps is rather high. While feedback does artificially drop this impedance, we eventually run back into the problem with the op-amps native output impedance all over again. That odor your smelling about now is the poor op-amp repeatedly moaning, “I’m sorry, I can’t.”

      The solution is actually pretty simple. A discrete buffer after the op-amp, and invite him to the feedback loop party.

      With IC based headphone amps, the imbalance between the native Zo and the load can get really wacky. The native Zo of the NJM4560 used in the Grado wood amps is around 600Ω. Not bad for Sennheisers, not good for Sonys. While feedback can help the latter, you wind up getting early distortion because the op-amp just can’t keep up.

      *The native output impedance of an amplifier is its Zo without feedback. Ideally you want this to be as low as possible to begin with, but many times you wind up with a higher Zo than your load, and you need feedback to tighten things up.

  2. Taking a closer look at the schematic, I’d be interested to know what the actual frequency response of the thing is, and which high-pass components dominate — at a casual glance it looks like the interstage coupling will start cutting below 100 Hz. There’s also no explicit high frequency rolloff, so it’s hard to tell where it will poop out or whether it will support high frequency oscillation.

    One neat thing about pentodes in a real transformer-based output is that they can be very nicely linearized by feeding one grid with a tap off the transformer (instead of the anode as used here). That configuration assumes you’re halfway there by running push-pull instead of running class-A like this though — pointless to do it with a single tube.

    1. I’ve not measured the frequency cutoff, I do have a scope but not sure how I can accurately measure it, I can’t say I’ve noticed bass missing below 100Hz.
      This whole thing was experimental and I was pleasantly surprised with the end result, sounds very good to my ears, I have other amplifiers that I bought from shops so I had something to compare against.

  3. I find it strange that in a world of digital recordings that anyone interested in audio fidelity would opt for something analog over DAC at the speaker. Using tubes makes even less sense because it introduces distortion.

      1. Because distortion is exactly what you want when listening in headphones!!1
        At the end of the day its just audiophool wankery, pretty glowing lights make the sounds so much better, as do $10K wooden knobs and your own separate power substation.

    1. I’d like to address your objection to the tubes in this project, but start by making some observations about “audiophile” equipment in general. These are my observations; there is no intent here to offend anyone, your mileage may vary.

      Let’s suppose someone marketed a super-high-performance amplifier, the “Unobtanium 2000,” that had a perfectly flat frequency response from DC to daylight, zero phase, frequency, and transient distortion and zero noise. What is the point of the obsession with audio component specifications? Well, the audiophile says, “I want to hear the performance as it actually sounded.” Is that a realistic objective? Let’s see…

      Let’s start with the fact that most people’s hearing is notoriously bad, and few have the capability to discern between a $100 dollar and a $1000 dollar sound system, let alone something that’s truly high-end.

      Even if they can, most likely, their equipment will end up in their living room, a room with square, flat, hard walls, floors, and ceilings… and standing waves. The damping in that room will have been determined randomly by carpets, curtains, and furniture, placed not for their acoustic affect but for practical and aesthetic reasons.

      Similarly, the listening position in that room will be dictated by the position of the most comfortable chair, not phase cancellation or propagation differences between the left and right ear.

      The speakers may or may not be impedance matched, may or may not be phased properly, and in any event, depending upon their design and whether they are raised, at floor level, or mechanically coupled to the floor, will introduce effects of their own.

      While some audiophiles have equalizers with noise generators and a calibrated microphone for tuning purposes, I have never personally met anyone–including self-described audiophiles–who made proper use of it. Rather, the equalizer and tone controls are usually set to what “sounds good,” meaning the bass and treble are jacked up.

      The better your equipment, the more the details of its deployment–not the equipment–dominates any perceived “distortion” of the intended music.

      Not only that, the desire to hear the performance “as it sounded” is probably unrealistic to begin with. Do you know, for example, what a snare drum REALLY sounds like? If you’re not a drummer or a musician who’s played with a drummer you don’t, regardless of how many CD albums you own.

      Why? Because typically, a snare drum is not recorded with an ambient-air microphone—you don’t record it the way a person would ordinarily hear it. Rather, it is close-miked, both top (to capture the crack) and bottom (to capture the rattle of the snare.) A recording engineer then decides on the ratio of those two mikes at mixing time.

      In addition, the drum is limited, compressed, and is usually tied to a gated reverb to “fatten” it. In the end, you get a great “punchy” snare sound, but in reality, it is a sound that has been synthesized and created through clever manipulation. It doesn’t, and has never, existed in real life.

      The same thing is true with instruments like guitars–especially electric guitars. There is no end to the “space noises” they can create. These days, it’s true even of vocals, which while always compressed and EQ’d, are now often “auto tune” corrected, as well. There is no faithful “RE-production” of a performance, because the “production” was artificial to begin with.

      My point is this: In the engineering/science world, if I’m making an electrical measurement–an objective activity– then absolute linearity, frequency response, and freedom from distortion in my signal chain is of great importance.

      Music, on the other hand, is not objective, but entirely SUBJECTIVE–from the point at which it is created/recorded to the point it is delivered to your ears. NONE of the objective equipment specifications really matter, so long as the end result is perceived to “sound good”– and what “sounds good” varies from listener to listener.

      In the end, the guy with the $10,000 stereo amplifier with 500 watts, DC to daylight response and 0.0001% distortion, won’t enjoy a Beatles, Rush, or Floyd album any more than I will on a semi-decent stereo of modest specifications.

      What he DOES get for his money is cool, a possible investment, a conversation piece, something to impress his friends with, and bragging rights of the same kind enjoyed by people who enjoy hot rods, monster trucks, racing boats, and custom hunting rifles.

      As to your comments about tubes introducing distortion, I refer back to the subjective nature of music and point out that some kinds of distortion are not only tolerable, but may even be desirable.

      One of the reasons tubes are appreciated in audio/music circles is precisely because of the distortion your are worried about. Tubes tend to produce distortion comprised of even harmonics. An even harmonic is related to the fundamental by a musical octave. The distortion artifacts are therefore *musically* related to the fundamental, so their appearance has the effect of “fattening” the sound. It’s like the difference between hitting middle “C” on a piano keyboard all by itself, or hitting middle “C” and every other “C” on the keyboard simultaneously.

      For years, when I used to perform, I played through power amps produced by a company called “MosValve.” MosValve amps used FETs in lieu of tubes (the benefits being reduced weight, no heat, less fragility, and greater reliability), but were purposefully configured to distort the way tubes will. They were awesome.

      By the way, analog magnetic tape exhibits this behavior too and back in the day, I knew recording engineers who would purposefully run their tapes just a bit “hot” to capitalize on it. BASF, if I remember correctly, actually produced a tape formulation that purposefully enhanced this effect.

      1. I mostly listen to classical music these days. Is all that post processing done on classical recordings? I know a professional string quartet musician who does his recording with an iPhone and one or two microphones (if even that), then self publishes the CDs.

        My stereo system is the best since I built the tube amp from a kit, and the speakers from a kit. Something about investing the labor yourself that makes it sound good.

        1. Really depends on the particular artist, studio, etc – but expect a fairly significant degree of post processing in all ‘studio’ music. For one thing things like drums are inherently very loud, which is why some venues hide the drummer behind a screen etc – so any recording will have a great deal of mixing to get the right level of those louder instruments to the softer ones (and of course even live the sound engineer will be mixing to get the ‘right’ sound).

          With Classical you also probably get some deliberate reverb and building acoustics artificially added – as recording each instrument or section so you can balance them properly means you don’t get that buildings acoustics showing up in the recording in the same way as just putting a mic in the audience…

          If you really want processing free you probably have to go for an audience recording of a live a show, and even then it won’t be a true representation, but it will be about as close as it can get to actually being there.

      2. I would hope the one thing a guy with stupidly over the top sound system gets is real silence when it should be – I’ve got a pretty good amp, but its always humming (its own transformer noises), the noise out in the speakers is almost nil, but its there. However when the whine is loud enough to hear from the amp directly a little noise in the speakers…

        It’s a problem I only really notice now I do so much music and video watching from an entirely solid state computer (Pi4) at night – as the PC fan noise largely masks it when its on (and my PC is rigged with ‘silent’ fans)… But still if somebody wants to drop the supercar money on a sound system, I’d hope its better than mine at silence, even if in general it sounds about as nice…

        I’d also hope they actually put in the effort to outfit the room its in so the comfy chairs are in the sweet spot… I mean if you have that much money you surely have the money to waste an entire room just for getting the very best music playback…

    2. Guess what, I have a DAC and I have other ampliers, this project was to kick start my passion for electronic engineering, I’ve not done anything like this since I left college 22 years ago and worked in IT.
      The end result was so good I now use this amp whiilst I’m working.

      It’s not always about hifi, tubes are meant to add distortion. I’ve tweaked everything to what I think sounds best, for me different tubes and op amps create a different sound.

      1. Check out Vinyl Savor, Bartola Valves, diyaudio.com (tube section), Tubecad.com and do a google search for Dmitri tube spice modeling. Also Mr. Karlson’s lab and El Paso tube amps on youtube. Tubes are not meant to add distortion. These gentlemen may prove you wrong in that. They can be very high fidelity devices with little complexity. Everything you want in HiFi. Check out directly heated triodes and RF transmitter tubes. Some good ones: 45 Triode, 2A3 triode, 6C33, 833 Triode, 809 triode, 6AK5 pentode, and lots of others. Tubes went away due to size and high voltage supplies not necessarily that they were ‘wrong’ IMO.

    1. The pre-amp output is probably a little less than 750 ohms, the parallel value of the 3k plate resistor and the 1k resistor to the opamp inverting input. That will be lowered by the effective plate resistance of the tube, which would be 300k if the plate weren’t connected to the second grid.

      The main outputs should be a fraction of an ohm.

    2. I only have Sony MDR-1R and Master & Dynamic MW60, the Sony being neutral and the MW having slightly more bass. Both are 32ohms.

      I would imagine this would be similar to any other op amp headphone amplifiers, it would proably struggle to play > 150ohms.

      With my headphones puting it halfway up is too loud (it would cause deafness and it did sound clean when I tried it).

  4. Why all the gain? Given that a volt or less is a line level signal the only thing needed is current gain not voltage. Oh, for turning up weak YouTube audio with it’s fewer bits to start with. I’ve heard minus 30dB. Tubes, battery power what the heck, but blurtooth really. Only the latest 5.0 specs may be of any use with such esoteric use, along with high def sources. Anything CD spec is a waste.

    Cathode output usually is the way to get low impedance out of a tube without a transformer.

    1. “Cathode output usually is the way to get low impedance out of a tube…”

      Interestingly, this tube can put out 20 mA with no difficulty. That’s more than the guaranteed minimum of that output op-amp.

      I know, I know, that means you will only get around 7 mA rms into the load, but that’s still darned loud into headphones.

      You’ll need around 5 Vp-p to drive it this way, so you’ll need a preamp.

  5. I’m an old tube guy who still works on ‘ancient’ vacuum tube stuff from the 1920’s through the 1970’s. And I have to say that you have a very impressive project!

    You should ignore those who want to pick your project apart. Sure, it doesn’t have the fidelity of most modern equipment, but who cares? Anyone who has heartburn with your work should pick up from your are at now and and evolve your design to something better for himself. And anyways, most don’t know how horrible some classic designs perform compared to some of today’s high end standards.

    It’s a vacuum tube project and I’m impressed that it’s battery powered without the need for an A and B battery! I have a few of those radios, BTW.

    1. Agreed! Building a quiet high voltage power supply that runs from a battery takes effort. I used to build electronics for a neuroscience research lab. We needed a constant current source for a finger tap stimulator (the sensation feels like being touched lightly with a pencil eraser). An engineer from Pitt designed and built the stimulator using photo flash batteries for safety and to get the high voltage necessary to get a decent constant current pulse from an FET current source. The problem is that photo flash batteries were terribly expensive and we were using a lot of them. I doubt those batteries are even made now.

      I tackled the problem by looking at what Grass Instruments did for their SD-9 physiological voltage stimulators which had an isolated 100 volt output. They used push-pull oscillator driving a resonant air core transformer. I built a similar circuit using a glass test tube as the coil form. I tried using a 50 ml syringe barrel, but the plastic got hot at the ~ 1 MHz frequency I chose. No local supplier had traditional paper forms. I needed about 200 volts for the current source supply, so I used a voltage multiplier rectifier with Schottky diodes. The filter caps could be small silver-mica at that frequency. The scientist doing the experiment said he could still see fuzz on the EEG traces on his oscilloscope. I cleaned up the common mode noise with the highest quality 600:600 ohm mixing console transformer I could find with the primary in series with one side and the secondary in series with the other power supply lead, That did the trick.

  6. PAIA in the nineties sold a preamp “for that tube sound”. An opamp feeding a 12AX7, I’m pretty sure the tube was running off very low voltage, a “starved circuit”. The premise was to get a tube into the signal line, and that was enough tor that tube sound.

    There was the usual article, in “Electronics Now”. Apparently June 1994.

  7. I have read apocryphal stories of the famous Bob Carver throwing down the gauntlet, that he could make a solid state amplifier sound like any other storied amplifier. Carver would take a sample of the exotic electronics to his hotel room and emerge hours later with his creation which confounded audiophiles who were unable to distinguish Carver’s sound-alike from the exotica.

    Both my parents were musicians. Both played in a swing band. Father played trombone and mother played piano or keyboard. My mother was classically trained pianist and taught piano as well. Their home hi-fi: Archer stuff from Radio Shack. Once you get to 35 or 40 years old, all that golden eats stuff is much less important and the story being told by the music matters more than whether the concert master is three feet or four feet from the next chair on the sound stage. That is going to be even more true of this generation who seem intent on blowing out their cochlear hair cells by listening at insane levels in their cars. One of my friends is an audiologist and he tells me he feels pretty sure about his job security.

    1. This is a nice looking project. I believe if I were building a project involving vacuum tubes that I would have opted for a cathode follower circuit with 12AX7 or 12AU7 (ECC83/ECC82) tubes and no solid state output. The op-amp which sort of negates the purpose of having vacuum tubes present. The complication being that in the cathode follower role, those tubes’ plate voltages were often run around 300 Volts DC so the designer had working space on the load line. It also depends on the headphones’ impedance. In the 1960s headphones could be found with impedances up to hundreds of Ohms. I don’t follow audiophile stuff closely these days, but I’ll bet most headphones are at most tens of Ohms impedance thanks to typically being driven by transistor or IC output stages.

  8. My personal hi-fi system contains an “A502 DC Stereo Control Center.” I built it from a kit in the mid 1980s from an importer called Monarchy Engineering, Inc. the A502 is a relabeled Fujitech Luxkit imported from Japan. Luxman used very interesting design where everything is DC coupled. The differential stages are all discrete FETs and output stages typically are matched bipolar transistor transistors. There are several handfuls of op-amps present, however the op-amps aren’t used directly in the audio path. The op-amps are used to maintain the Q-points of discrete components and the op-amps have cut-off frequencies in their feedback loops of a few Hz. It was a pretty challenging kit to build because the assembly manual is in Japanese and I don’t read Japanese. Building the wiring harness was the the toughest part because it called for specific color and length wires. I figured out the Japanese characters for colors from the resistor color code chart in the front of the assembly manual and fortunately numbers are Arabic numbers. The amplifier stages were to be adjusted open loop as close as possible to zero volts offset and only after everything was adjusted would links be soldered in to connect the what I call “servo” op-amps.

    In its day, the A502 cost about $250. That would be like $500 or more now. That’s still cheap compared to golden ears audiophile equipment with pretentious sounding multi syllable names. I took my preamp to a number of listening parties to go up against fancy equipment. The audiophiles agreed they could hear differences between the various equipment, however none would commit to a winner preamp. I still use my A502 preamp daily in my modest A/V set-up 35 years after I built it.

  9. At least the tube actually works. I’ve seen some of these headphone “tube” amplifiers where the tube is just there for wankery show. At best just the filament is hooked up, or some dumb shit like putting a blue LED under the tube, because everyone knows a tube’s filament glows blue lol.

  10. Tubes are actually the most linear devices. You don’t require much if any feedback to get them to work. Our lead RF lead engineer claimed they were the most linear devices ever made. Tube sound is a myth from people with little exposure in both design and use of tube circuits. If it has a tube ‘sound’ your circuit is not designed right.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.