In this article I present a practical single-ended amplifier project based on the 45 tube, one of the most appreciated directly heated triodes when you want just a few real watts but extremely high sound quality. The 45 was born at the end of the 1920s, under the UX245 designation, as an output tube for domestic radio sets, and over time it became a reference for those seeking natural timbre, microdetail, and an exceptionally correct midrange, provided the circuit is simple, properly dimensioned, and free of unnecessary compromises.

The core idea is the one that I enjoy the most: hands-on experimentation, components chosen with good sense, transformers designed and wound to spec, and a truly obsessive focus on power supplies and wiring, because with these triodes every detail is clearly audible. All of it without chasing trends or fetish items, but with a clear goal: to obtain an amplifier that is quiet, stable, enjoyable even at low volume, and capable of a top-level mid and treble, without giving up a credible low end that is smooth, well blended, and believable.

The first prototype

On top of that, there is a deliberately “handmade” aspect and, if you like, a bit of romance too: this prototype was born as a restoration project, not to cut costs but purely for personal enjoyment. Recycling an old electronics chassis, bringing it back to life, and letting a triode power amp grow inside it is something that has always amused me, especially when the goal is to experiment without any fixed aesthetic constraints. This was in fact one of my first builds of this kind, and it reflects a very direct, functional approach. If I were to rebuild the same project today, the look would certainly be different, but the experimental spirit and the technical substance would remain exactly the same.

First of all, I did what should always be done: I drew up the schematic, defined clearances, weights, and dissipation, built the transformers, and only after that did I start studying the mechanical layout. This avoids weird compromises, forced wiring runs, and improvised solutions that you end up paying for later in noise, hum, or poor service access.

With a triode single-ended amp, layout matters. A lot. Filament currents are significant, ground returns must be managed logically, signal paths have to stay clean, and transformers and chokes must be oriented and spaced with care. Even a simple project can become complicated if the mechanics do not help.

The rust was removed with a small Dremel grinding wheel, working slowly to avoid deforming the sheet metal and without taking away more material than necessary. I then reassembled everything with new rivets and screws, because on an old enclosure the hardware makes the difference: if you leave mechanical play, over time you get vibrations, creaks, and contacts that oxidize. At that point the structure was ready for powder coating, an ideal choice when you want a durable, uniform finish, especially on a restoration that needs to come back as a “serious lab” piece rather than remain a wreck. In the photos below, the before and after…

To “erase” the past of that chassis, meaning old unusable holes and signs of previous builds, I also added solid walnut side panels finished with shellac. This is not only aesthetics: wood helps cover and stiffen, it makes the result cleaner, and it adds that handcrafted touch that looks perfect on a triode device, without turning it into a “display case” object. It must remain an amplifier made to play and to be tested.

Once the mechanics were completed, I wired a simple circuit with few tubes and a very clear philosophy: just enough gain, solid drive for the output tube, very carefully designed power supplies, and components consistent with voltages and operating conditions. A project like this should not “impress” with unnecessary complexity. It must do the fundamentals well, because 45 tubes (and even more so vintage UX245) reward exactly this approach.

The tubes used are two ECC83 / 12AX7 for the input stage, optionally replaceable with 5157 for voltage gain, followed by a NOS 6H6Pi (USSR) used as a buffer and driver stage for the output tube. This choice is not accidental: ECC83s operate at very low currents and have a high output impedance, which makes them poorly suited to directly drive a tube like the 45. Adding a dedicated buffer instead provides the output grid with a low-impedance source, improving control, dynamics, and overall clarity.

The 6H6Pi can be replaced without particular issues by many other double triodes, including E182CC and E180CC, and more common tubes such as 12BH7 also work correctly. In theory ECC88 and ECC82 can be used as well, however in listening tests I obtained better sonic results with more “muscular” triodes like the ones mentioned above, especially in terms of drive solidity and timbral naturalness. The output tubes are two UX245 RCA NOS from 1928, while the rectifier is a Philips GZ34 NOS.

The choice of the driver chain is not accidental: with triodes like these you need a driver capable of current and control, not just “a bit of signal”. The result, when the drive is correct, is a sound that is freer, more believable on microdynamics, and with a midrange that never becomes aggressive or “shouty”.

The output transformers visible in the photos of this article are my SE5K6-UNI, an older generic model I used at the time and that should now be considered obsolete and no longer orderable. As experience and measurements progressed, today I offer a dedicated transformer instead, the SE4K6-45, optimized specifically for the 45 tube. In a single-ended amplifier, the output transformer is not an accessory, but a substantial part of the amplification itself. A model designed specifically for the operating point and characteristics of the 45 allows the tube to be used more effectively, improving control, linearity, and overall naturalness of the sonic result.

The chokes that appear in glossy black are 1950s NOS chokes (properly repainted because they were originally yellow) from which I derived my current 16S64. The idea is to have serious B+ filtering, with a choke that actually does the job, reducing ripple and grime without having to throw huge capacitances everywhere. The power transformer is of course custom, sized for the real requirements of the circuit, with margin and sensible management of filament and high-voltage supplies.

The signal capacitors are all Sprague Vitamin Q. In the photo they are enclosed in heat-shrink tubing, a choice made to insulate the body when it may sit at elevated potential, prevent accidental contact, and improve mechanical robustness. The output tubes are self-biased, with high-quality NOS electrolytics bypassed by an audio-grade polypropylene. Here as well the logic is simple: stability, low noise, and a bass response without artificial bloating.

The 45 filaments are powered in DC. In the original prototype, from many years ago, I used a passive cell with about 44000uF of smoothing, a solution that was already effective back then to obtain absolute silence on the speakers. With directly heated triodes, hum is not a minor detail but a real design specification: if the amplifier must be truly listenable with high-efficiency speakers, the idle noise simply has to disappear.

Today, taking advantage of technology progress, I would not hesitate to adopt an even more radical solution, using supercapacitors on the order of 5 farads for the filament supply. This way the tubes operate in conditions very close to a battery supply, which remains one of the best solutions in absolute terms for DHTs. In my view, this approach is clearly preferable to the small “active” supplies full of transistors that are fashionable, because it eliminates noise, spurs, and unpredictable dynamic behavior at the root, while keeping the system simple, stable, and consistent with the project philosophy.

A nice little detail is the rotary power switch, a NOS component probably dating back to the 1950s or 1960s. It is one of those details that seem secondary, but in reality they tell a different way of building: solid mechanics, generous contacts, a true “industrial” feel. And, not negligible, if chosen well it is also reliable over time.

Below I report the description and listening impressions from the current owner of this Tulipa.

How does it sound? The result went beyond my most optimistic expectations. As soon as I turned it on I did expect a first-rate mid and treble, but I was not so sure I would also get a smooth, engaging bass that was perfectly blended with the rest of the audio band. For years I had convinced myself that it was “impossible” to have a truly credible low end with triodes of this type, but then, after hearing a properly modified 300B built the right way, I admitted that hope was reasonable. Here the difference comes from three things: a serious output transformer, clean power supplies, and correct drive of the output tube.

Of course you need the right speaker. You need speakers of at least 90/100db efficiency to fully enjoy the peculiarities of this tube. Let’s say from an Altec 19 upward, Tannoy, and in general systems with real sensitivity and impedance that is not “capricious”. They also play with the 96db of Klipsch Heresy, but it is from the Cornwall upward, staying in PWK’s house, that you can listen at a totally involving volume, even with energetic genres. Power is what it is, but when the watt quality is high, the sense of presence and control can be surprising.

With the construction precautions applied, hum is nonexistent even with Emission Labs 45 (Mesh Type), which until now had been almost unlistenable in other power amps. For me this is a fundamental point: if a DHT single-ended amp hums, it is not “character”, it is a problem. Here, instead, the silence is there, and when there is silence you hear everything else, meaning microdetails, harmonic tails, room ambience, and the breath of instruments.

The sound has reached peaks that are hard to surpass because we are close to the reality of the sonic event. I am not saying it poetically, I am saying it technically: the sense of naturalness, the absence of grain, and that ability to make a voice or a bow believable without “special effects” are exactly what 45 triodes can do, if they are not mistreated. And at that point it also becomes hard to justify certain prices for famous-brand gear that, when you get down to it, often leaves a bitter taste.

I wanted to add that the amplifier was built around the RCA Radiotron UX-245 and it is precisely with this tube that, in my personal opinion, you achieve the most flattering results. The EMLs come right after, even if in the first moments they can seem preferable. Standard 45s (with the non-Globe or non-Balloon shape) of course also have the magical sound of this tube and, for those who have never heard 245s, they may already think they have arrived. But this amplifier brings out every tiny nuance, and the sense of involvement and satisfaction you get with the RCAs, unfortunately, cannot be fully described. It is one of those things you understand in ten seconds when the first track starts.

Salvatore’s build (2021)

In 2021 this project was revisited and optimized in a version built by Salvatore. The original version used a pair of NOS chokes for B+ filtering of the two channels and simple CRC cells for filtering the 45 filament supply, while the rectifier tube was a GZ34.

In this version I optimized the power supplies using a 5V4G as the rectifier, smaller than the GZ34 but perfectly capable of feeding the modest current demands of this circuit. I used a single CLC cell (with an SB-LAB 16S64 choke) for both channels, because there were no crosstalk issues in this configuration and the layout allowed for a well managed ground return.

I improved the filtering of the 45 filament supply with an “RCLCRC” cell using a pair of 16S63, where the last capacitor is 33,000uF. This cell not only provides a supply as clean as a battery (on the oscilloscope you do not see any difference between off and on), it also turns on the filament with a very soft start over several seconds before reaching 2.5 volts. It is definitely a great thing to preserve the delicate filaments of these tubes, which are one of the most critical points of the whole system, both for reliability and for residual noise.

The rest of the circuit remains almost the same as the original, aside from a few small changes here and there to suit the specific wiring, improve component placement, and make the result cleaner in terms of serviceability. Below are photos of the customer’s build that was brought to me for verification measurements and a general check. Here is Salvatore’s build:

Good evening. I am attaching photos of the finished project.

I have only taken a few measurements so far, as soon as I am less busy with work I will take more. In the audio test with my DIY speakers, a Filippo Punzo project, 4ohm, 98db, I have to say the result is incredible… at 1/3 of the volume pot it fills my 5×4 room. The dynamics across the whole frequency range even at very low volume are incredible, it is balanced, it does not overdo it on any musical genre, detail and sweetness are in charge. I did not expect such an articulated, precise bass that goes down very well, even more than my SE 300B. For now I thank you for the patience in answering my questions and I will definitely do other projects. As soon as I take other measurements I will attach photos. Maybe one day, if possible, I will bring it to you for more detailed measurements.

Salvatore

This feedback is interesting because it confirms two things I often repeat: first, perceived dynamics do not depend only on watts, but on how the amplifier handles transients and micro-contrast; second, a well made single-ended amplifier can deliver surprisingly readable bass, not “big”, but articulated and believable, if the speaker interface is sensible.

Another completed build by another customer (2023)

In 2023 I saw another completed build by a different customer. Here too, the best part is seeing how a seemingly “niche” project can be replicated with excellent results when the guidance is clear and when transformers, chokes, and power supplies are sized as they should be. Below you will find some photos of the finished build.

The builder’s comment is at the bottom of the page in the comments…

My most recent build

Years later I revisited this project by building a more recent version, still based on the same circuit philosophy, but with aesthetics and mechanical care far superior to the first prototype born from a reclaimed metal chassis. In this version I had more freedom to design the mechanics, optimizing component placement, wiring, and grounding paths, as well as the overall finish: cleaner panels, coherent drilling, an orderly front panel, and a final look that is clearly more “definitive”. Below you will find photos of this updated build, showing how the same project, with accumulated experience and mechanics designed on purpose, can become not only more elegant, but also more service-friendly and more rational from a construction standpoint.

Let’s look at the measurements:
Maximum power: 2.25 watts RMS per channel.
Bandwidth @ 1 watt: 10Hz – 28kHz -1db
THD @ 1 watt: 1.3%
Damping factor DF: 3.6

Two quick notes to read these numbers the right way. The maximum power of 2.25 watts RMS per channel is perfectly in line with a 45 single-ended set conservatively and sensibly, meaning without pushing the tube beyond what is reasonable. The 1 watt bandwidth, from 10Hz to 28kHz at -1dB, instead reflects the quality of the output transformer and the good overall balance of the project: real low-end extension and a top end that does not collapse early, without any tricks.

The 1 watt THD at 1.3% is typical of this family of amplifications, and must be interpreted with the usual caution: it is not only “how much” it distorts, but “how” it distorts. In a single-ended triode, much of the distortion is even-order harmonics and it grows progressively, often perceived as naturalness and timbral density, up to the point where you reach listening levels outside the intended use for a device like this.

The damping factor DF of 3.6 is consistent with a single-ended amplifier with little negative feedback. It is a value that calls for a suitable speaker, but it is also part of why these amplifiers, with the right speaker, deliver a “lively” bass and a very particular sense of presence, without becoming loose or confused.

Spectrum analysis @ 1 watt

The 1 watt spectrum lets you understand the circuit “signature” at a glance: harmonics that decay regularly, no abnormal spurious components, and a well controlled noise floor. It is a very useful way to confirm that the build, grounding returns, and power supplies are behaving as expected.

Bandwidth on resistive load @ 1 watt

The response on a resistive load is the baseline that tells us how linear the amplifier, together with the output transformer, remains under ideal conditions. It is also a good indicator of headroom at the band extremes, and it often anticipates what will happen later with square waves and reactive loads.

Response on reactive load @ 1 watt

A reactive load is what most closely resembles a real loudspeaker. Here you see whether the project remains stable, whether resonances appear, whether there are odd bumps or a tendency to oscillate. With low-feedback single-ended amps this test is particularly important, because it tells you how comfortable you can be with difficult speakers.

Square waves at 100Hz – 1kHz – 10kHz @ 1 watt

Square waves are a very “visual” test to understand the balance between the extreme low end and the extreme high end, and to observe any overshoot, ringing, or slew limitations. At 100Hz you mainly look at low-frequency behavior and any tilt due to primary inductance limits; at 1kHz you check the overall response; at 10kHz you stress the high end and the transformer behavior with parasitic capacitances and leakage.

Triangle wave 1kHz @ 1 watt

The 1kHz triangle wave is useful to check dynamic linearity and the possible appearance of “edges” or curvature that indicate speed limits or nonlinearity in some stage. It is a simple test, but on certain designs it says a lot, especially when you compare it with the perceived cleanliness and naturalness in listening.