This page collects five variants of small single-ended amplifiers based on the EL84, all born from the same basic idea, to get a few real Watts, but with quality, using output transformers sized properly, sensible wiring, and safety criteria respected. On paper these are “compact” projects, because the final result depends far more on how they are built, grounding, physical layout, shielding, wiring, and power supply, than on the electrical schematic alone. Some photos in the article may look aesthetically rough or not very refined, but that is normal, in several cases these are hobby builds by customers who assembled my kits or followed my schematics, and as such they look like what they are, DIY builds, useful to see precisely because they show the practical, real side of self-building.

The five circuits are similar, but not equivalent, they differ in the driver tube, input sensitivity, dynamic behavior, the possible presence of global negative feedback, and in one case even the choice of rectification. What they share is the SB-LAB approach, transformers correctly matched to the real working impedance, a clean signal path, stable power supplies, and a mechanical and electrical build that leaves no room for hum, oscillations, or audiophile “magic”.

If you are considering an EL84 project, this page shows you two things, how a properly built single-ended amplifier with few tubes can sound, and, above all, how a tube amplifier should NOT be assembled. In particular, the “wooden base and flying wiring” theme comes up often, because it is one of the main causes of hum, instability, and real hazards (temperature, insulation, leakage, and the lack of a serious ground reference). Below you will find the five sections, each with its dedicated premium schematic, practical build notes, and where available, instrument measurements.

1) Pico 8284 – EL84 driven by ECC82

This Pico was born in a very concrete way, some time ago I sold a set of transformers for a Pico 8282 to a customer who then, during the build, ended up with unbearable hum and with a mechanical and electrical layout that made it impossible to “fix it with two wires”, because the real issue was structural. When a tube amplifier hums “in a terrifying way”, in most cases it is not the fault of the tube, the miracle capacitor, or the esoteric resistor, but of how grounds, wiring, current loops, and shielding were handled.

The two photos below show the customer’s build, useful as a didactic example, I will not comment on aesthetics, because aesthetics are not an electrical parameter, but here the point is safety and functionality. Building a tube circuit on a wooden base, without a metal plane to provide a ground reference and at least minimal shielding, is a recipe for guaranteed problems. Tubes run hot, and noval power tubes like the EL84 run quite hot, without proper thermal decoupling, without a metal plate, and without correct distances and mounting, you risk deformation, charring, insulation breakdown and, in the worst case, a fire hazard. Also, without a conductive chassis, “ground” becomes a concept, meaning a wire that runs wherever it happens to go, with return currents sneaking into the input stage and giving you hum and buzz regardless of how nice the schematic looks.

The Pico 8284 is the version with an EL84 driven by an ECC82, designed to be simple, stable, and practical. The premium schematic has no global negative feedback, this choice makes the circuit very direct and “clean” in philosophy, but it demands even more attention to layout and grounding, because when you do not rely on global feedback to tame gain and impedances, the real result is made by the iron and the wiring.

To turn a “humming kit” into an amplifier that works, the first step was to rebuild a proper mechanical base. I made a 3 mm aluminum panel as the structural support for the circuit, to get rigidity, a serious ground reference, and at least minimal shielding to the outside. This is not a whim, the metal plane lets you control return-current paths, reduce loop areas, anchor components solidly and, above all, make the result repeatable.

Once the metal “backbone” was defined, I had a carpenter build a raw wooden box, then I finished it myself, drilling, filling, sanding, staining, oiling, and finishing with shellac. The wood here has an aesthetic and mechanical role as a frame, it does not replace the chassis function. In the photo you see two cabinets because the other one is meant for a “Scherzo”.

In my builds I never give up an insulating plate (bakelite or equivalent material) under the metal plate. It is a workshop trick that makes a real difference, it lets you screw down components, terminal strips, and anchor points without inventing “flying” mounts, and it lets you manage wiring and grounds neatly. Above all, it reduces the chance of failures due to vibration, stress on component leads, and accidental contacts.

Another “non-negotiable” point is anchoring. A tube amplifier is not a PCB you can let dangle around, it is an object that runs hot, vibrates, gets moved, and must stay stable for years. Having tie-down and mounting points for components and wiring is literally vital, both for reliability and for safety. If you are building without anchors, you are not “minimal”, you are just building a future problem.

Finished Build

With a metal chassis, insulation, anchoring, and sensible layout, the Pico finally becomes what it should be, a small single-ended amplifier that is quiet, stable, and enjoyable, with adequate transformers and no “tricks” to hide basic flaws. Below are some photos of the completed build.

Measurements

The numbers below are useful only if read with the right context, we are talking about a few-Watt single-ended amplifier, where the goal is not “to make power”, but to keep coherence, low noise, and a serious frequency response with an output stage that does not collapse as soon as the load becomes real. Bandwidth and distortion at 1 Watt say far more than any claimed maximum power.

Power: 3 Watt RMS per channel
Bandwidth @ 1 Watt: 20 Hz / 36 kHz -3 dB
THD @ 1 Watt: 0.45%
Damping factor (DF): 1.3
Rout: 6 ohm

2) Pico 8484 – EL84 driven by ECC84

The ECC84 is often overlooked not because it is “bad”, but because many DIY builders do not know it, and what you do not know can feel intimidating. In reality, it is a very sensible choice as an EL84 driver when you want a brighter and more controlled behavior compared to an ECC82. The ECC84 is conceptually the predecessor of the family that also includes the better-known ECC88, with a Mu around 24 and a lower internal resistance than the ECC82, which helps drive the next stage more authoritatively, especially when layout and stray capacitances are not “laboratory grade”. With the same topology, this tube can give a sense of greater definition and better high-frequency control, precisely because it works better when it has to “push” current into input capacitances and real-world wiring. A huge practical advantage is NOS availability, you can often find excellent samples at reasonable prices and in quantity. There is also the PCC84, with a slightly different nominal heater, which in many contexts can work without drama if the heater supply is well designed and has margin, but as always the project must be considered as a whole, not “by feel”.

From a circuit standpoint, this version keeps the Pico approach, simplicity, few well-chosen components, and correct output transformers. The premium schematic is also free of global negative feedback here, so the same golden rules apply, good grounding, tidy wiring, a quiet power supply and, above all, a mechanical build with a metal plane and proper anchor points. The schematic is intentionally close to the previous variant, so anyone who wants to experiment can understand what really changes when you swap the driver, without introducing ten variables at once.

3) Pico 8084 – EL84 driven by EABC80 (or ECC83)

The EABC80 is an iconic tube from 1950s and 1960s radios with FM, it contains three signal diodes (unused here) and a triode with characteristics similar to an ECC83, but with a lower amplification factor (about 70 instead of 100). In other words, it is a smart way to get a high-gain driver using a common, inexpensive tube that is often surprisingly reliable. In many years of radio repair work, it is one of those tubes that is rarely found faulty, and that is a very practical reason why it is still easy to find NOS at modest prices. Its “mass-market device” vocation makes it perfect for anyone who wants to have fun without spending absurd money, but that does not mean giving up performance at all, it just means thinking technically instead of fetishistically. In this version, precisely because the available gain is high, the schematic uses global negative feedback, not as a fashion statement, but to make the amplifier manageable, with correct input sensitivity, a usable volume range, and a higher damping factor compared to the versions without global feedback. With a high Mu, without a stabilization strategy, you risk an ultra-sensitive and nervous circuit, which then gets “tamed” in random ways with odd pots or improvised attenuators. Here, instead, the right thing is done at the root.

Here is the premium schematic. It is possible to use a single ECC83 instead of the two EABC80s, wiring the two sections appropriately, in practice the EABC80 becomes an economical and robust route to the same goal, and the ECC83 remains the “classic” alternative for those who already have it or prefer to stay with a more familiar type.

Below I include a real-world testimonial from someone who installed the transformers and the recommended modifications, because in these projects the difference between “on paper” and “on the bench” is made by details such as the feedback phase, correct primary connections, and wiring that does not trigger oscillations. It is completely normal that, at first power-up, if the phase is reversed, negative feedback becomes positive and the amplifier starts oscillating, it is not a tragedy, it is electronics. The important thing is being able to recognize it and fix it in the right way.

Hi Stefano, I’m “M.T.T.” from Ravenna, the one you sold a pair of SE4k5-EL84 output transformers to, I’m writing from my second email address because the other one is having problems.

I installed the transformers and made the modifications you recommended… At first power-up a problem came up with the negative feedback (actually in this case it was “positive”) that made the amplifier oscillate. I solved it by swapping the primary connections of the output transformers, so the red (+H) was connected to the plate and the black (plate) was connected to +H.

Once the negative feedback problem was solved and the voltages were checked, I started testing with a signal generator and oscilloscope, and the amplifier’s improvement was immediately obvious.

The frequency response is linear from 20 Hz to 35 kHz, the square waves are almost perfect under all conditions (tests performed with an 8 ohm resistive load), and there is no ripple, or in any case it is not measurable.

The input sensitivity I measured is 0.8 V RMS on 250K (I did not have the 47K pot) for an output power of as much as 3.6 W RMS with both channels operating temporarily (measured up to the first hint of deformation of the sine wave on the load).

Moving to listening tests, you immediately hear the depth of the low frequencies, the “cleanliness” of the highs, and the lack of intermodulation distortion in loud passages, due to the low-value G1 resistor that prevents the grid from shifting away from its operating point.

The listening tests were done with a PHILIPS CD624 CD player with BITSTREAM conversion, an inexpensive early-1990s unit that in various comparisons has outperformed much more expensive devices, and with Philips 22RH496 speakers, 3-way with acoustic suspension woofer, a bit hard to drive, but the amplifier showed no signs of strain.

I’m sending you some photos of the finished unit.

See you soon.

4) Pico 8084VTR – EL84 driven by EABC80 (or ECC83) with EZ81 rectifier

This variant was born from a specific request, a customer wanted to use an EZ81 rectifier. It is an understandable desire, because rectifier tubes are part of the “vintage” imagination, but the important thing here is doing it properly, meaning sizing the power transformer and the entire filter section correctly, respecting the rectifier’s limits and keeping the voltage stable. The circuit stays very close to the Pico 8084, but the power supply changes, instead of a diode bridge, rectification is handled by the EZ81, and that implies precise choices in current, capacitance, and series resistance, not “just put the tube in and go”.

Here is the premium schematic. Again, it is possible to use a single ECC83 instead of the two EABC80s, wiring the two sections appropriately. The presence of a rectifier is not an automatic sonic pass, what matters is the overall power-supply design, its dynamic resistance, residual ripple, and its ability to hold the operating point when the EL84 draws current in pulses.

5) Alimede – EL84 driven by 5842

Alimede is the “more grown-up” project in this collection, not because it has a hundred tubes or chases trends, but because it was designed with the goal of solving, technically, some typical limitations of small single-ended amplifiers, while keeping simplicity and coherence. It all began when I was given a pair of unknown monoblocks as “scrap iron”, not as a figure of speech, but literally devices with no technical sense, from which, however, some mechanical parts and reusable components could be salvaged. Instead of throwing everything away and calling it a day, I chose to show two things, the technical review of the starting point, and how you can turn a “hack job” into a real amplifier, with electrical criteria, operating limits respected, and measurable results.

Before the rebuild, it is worth pausing for a moment on the diagnosis of the “donor”, because it is the perfect example of how certain pseudo-HiFi is born from slogans rather than design. Here I show it live, from the technical side, an object that fully falls into the category of “impresentabili”. They were two single-ended monoblocks with an EL84 strapped as a triode, driven by a 5842, powered by a 6X4, and of course zero feedback. A compendium of audiophile clichés repeated like a mantra:

• Dual mono because separating the channels “sounds better”.
• Triode output because “triodes sound better than pentodes”.
• A famous driver because “famous tubes sound better”.
• A rectifier tube because “a rectifier sounds better than diodes”.
• Zero feedback because “no feedback sounds better than feedback”.

Facts, however, are not opinions. Let’s start with the power supply, the 6X4 is a small, delicate rectifier with limited current capability. Here it was followed by a CLCRC filter with huge capacitors and resistors chosen with no inrush-stress logic, 47 uF / 3 H 90 ohm / 330 uF / 1k5 / 330 uF. In this configuration, at every start-up the tube was being “pulled by the throat”, with current surges incompatible with a long and happy life. A quick simulation with PSU Designer immediately shows an operating-limit warning, with a 1.6 A inrush peak and a continuous current around 72 mA:

That means every power-up could have been the last, and even at steady state the tube was operating beyond common sense. Then there was the power stage, an EL84 as a triode into a transformer with a 10k primary. Real result, ridiculous power, on the order of a few tenths of a Watt clean, and a little over 1 Watt at full saturation. Yet it was “sold” as 3.5 Watts. This is not a matter of taste, it is an objective fact. The EL84 cannot deliver 3.5 full watts even in pentode mode.

The 5842 driver, with its cathode bypassed by a random capacitor, provided excessive gain, a few hundred millivolts at the input were enough to drive the output into clipping, making volume control a lottery. Moreover, under those conditions and with that layout, the tube picked up external interference, fluctuations, and even radio frequency energy, visible on the oscilloscope as trace “dirt”. To complete the picture, the heater supply was referenced directly to the full B+ voltage, resulting in heater-to-cathode differences well beyond the limits of several tubes involved. Avoiding those issues is not “tuning”, it is basic design hygiene. Even the original transformers, once measured, showed mediocre bandwidth, and in any case referenced to such a low power level that it was not very meaningful in a real system.

The point is not “I do not like how it sounds”, the point is that a circuit like that is wrong from top to bottom. Tubes are not talismans, they must operate within limits, with correct bias, healthy power supplies, and coherent transformers. Saying “but it sounds good” is not enough, it is like saying a Ferrari “moves” even if it only goes in first gear and cannot exceed 30 km/h with the engine always on the edge. If an object has structural problems, instability, stress, and risk, to me it is not “a matter of taste”, it is an unacceptable device.

SB Alimede is Born

Before Merlòtto, I built this small amplifier starting from what was truly worth saving, tubes, sockets, a few selectable components, a power transformer and a choke, plus hardware. The donor’s output transformers could not be reused, the 10k primary impedance and triode-strapped EL84 operation led to too low efficiency and an incoherent match. I therefore adopted a pair of my SE4K5-EL84 output transformers with 4500 ohm primaries, consistent with the goal of obtaining 3 real Watts, controlled and measurable.

I chose a topology often called Shadeode or, more descriptively, Partial Feedback, and combined it with cathode feedback to increase damping factor without resorting to heavy global negative feedback. The concept is simple, make the EL84 behave more “linearly” dynamically, while keeping the power capability of pentode operation, then refine control and output impedance with a well-reasoned local feedback strategy. A similar idea, conceptually aiding control, can also be found in the STC configuration described in the Luna. The premium schematic of the unit is shown below (click the thumbnail to enlarge).

Rectification is handled by a pair of 1N4007 diodes. I will say this clearly, it is not a “cheap” choice, it is the correct choice for the available voltages and the reliability goal. In that context there was no sensible margin for a rectifier tube, inserting one would have forced smaller capacitors and larger series resistances, increasing the supply’s internal resistance and creating more problems than benefits. If you want a rectifier because “it looks cool”, you are doing aesthetics, not design.

The first stage is a 5842 biased with a selected green LED at about 2.00 V, bypassed by a very generous low-ESR 1500 uF electrolytic. This type of biasing, when done correctly, behaves in practice like a very stable bias, close to the idea of fixed bias in terms of performance. The 5842’s 10 mA current keeps the LED conducting, the large capacitor preserves stability even when the signal drives the tube close to cutoff. For extra robustness, I added a few mA of additional current through the LED via a 100k resistor from B+, so the cathode voltage stays steady even under worst-case conditions, such as 20 Hz signals near clipping. This is the kind of “invisible detail” that separates a well-designed amplifier from a circuit built with poor workmanship.

The 5842 drives the EL84 in pure pentode mode and receives the partial-feedback signal through resistors on the order of 90k. In this configuration the EL84, while still a pentode, behaves more like a triode dynamically, but retains pentode power capability. I then completed the balance with cathode feedback on the output tube, bringing the output transformer secondary “under” the cathode, to increase damping factor and make the load response more controlled.

Under the EL84 cathode you will find a bias and bypass network designed to be serious, not “thrown in”, resistors combined to reach the desired value, a properly sized electrolytic as the main bypass, and in parallel selected non-polar capacitors to reduce dissipation factor at frequencies where the electrolytic becomes less ideal. You can also achieve a similar result with a good polypropylene capacitor of adequate value, but the key remains the same, components are not chosen “at random”, they are chosen with measurement and coherence to the goal.

The result is not a marketing claim, it is a set of measurements, 3 real RMS Watts, with a higher damping factor compared to the Pico versions without global negative feedback, and with bandwidth that demonstrates the quality of the iron and the stability of the circuit. Below is the frequency response at 1 Watt RMS and, next, spectrum and square-wave shapes. The amplifier’s power is a real 3 Watt RMS, full and clean. The damping factor DF is 3.33 and the bandwidth is 12 Hz / 42 kHz -3 dB @ 1 Watt RMS, below the graph…

Harmonic distortion and noise, measured at 1 Watt, complete the picture, a single-ended amplifier can be “musical” without being muddy, and it can be simple without being approximate. The graph below is a useful reference to understand what you are really hearing when you say “clean” or “dirty”. Harmonic distortion is 1% with a signal-to-noise ratio of -65 dB, still at 1 Watt, below the graph:

Square waves at 100 Hz, 1 kHz and 10 kHz are another immediate way to “see” stability and transformer behavior. They are not a forum fetish, they are a practical test to understand whether you are chasing an idea or whether the circuit is actually working well into a controlled load.

Let’s look at the build, a mahogany chassis finished with shellac, and powder-coated aluminum plates. Here too the philosophy is the same as the Pico, but taken to a more “definitive” level, rigidity, order, short paths, anchor points, and a logical separation between power supply and signal. The photos below show the finished assembly.

Sonically, Alimede is designed to be fast, transparent, and controlled, extended and readable highs, detail present without becoming “glassy”, and above all low-frequency grip that, for a 3 Watt single-ended amplifier, is anything but trivial. It should not “erase” details, it should let them emerge naturally, as a truly HiFi system does when it is well designed. People who listen to it are often surprised because these “only 3 Watts” are usable Watts, not brochure numbers. I also include a comment from a reader who purchased the premium schematic, because it describes the project’s goal exactly, a small amplifier for real listening, not a display piece.

Stefano, I finished the Alimede build a couple of days ago and I’m extremely satisfied! For its size, it has a surprising sound in terms of transparency, balance, speed, and control. It is only 3 W, but perfect for my late-night listening at low volume. You did a great job!!! A real miracle, considering the little wrecks you started from. Thank you very much and kind regards. See you soon, G.