History: The first version
I wanted to build an amplifier just for myself, made the way I liked, with no design constraints and a bit of creativity. I also had various leftover materials and salvaged components to reuse at zero cost, without qualitative compromises though.
The nickname of this amplifier is “the blasphemy” because I basically took a handful of audiophile dogmas and did the exact opposite of what the “commandment” says, yet I still obtained an amplifier that sounds veeeery, veeeery good, much better than a lot of stuff out there that follows the commandments. Let us see in detail the 7 commandments broken here:
- Tubes with a top cap do not sound good: An audiophile dogma would have it that tubes with a top cap do not sound good. In reality it is likely that those unable to wire properly introduce even greater interference into the circuits they build and conclude that the top cap is to blame, then they spread the word and everyone else follows blindly.
- Radio and TV tubes do not sound good: yet another audiophile dogma says that only “audio grade” tubes can sound good, only to then realize that “audio grade” tubes in many cases were born for radio and TV applications. In this amplifier the output tubes are for radio transmission, the preamp tubes are for radio reception and the little buffer is a TV tube.
- Pentodes do not sound good: the dogma says only triodes sound good. In this circuit both the output tubes and the preamp tubes are pentodes used as pentodes.
- Variable mu tubes do not sound good: the ARP34 are variable mu pentodes, but it is enough to keep the operating point away from the mu change area so that they are perfectly linear.
- Cathode followers do not sound good: another audiophile dogma claims that a follower buffer makes a circuit sound worse. In reality you put it in (here the ECC84) to avoid overloading the driver which, coming out at very high impedance, ends up distorting more than it should.
- Negative feedback does not sound good: another controversial audiophile dogma that I have already discussed extensively here.
- LEDs, diodes and other SS components make it sound like solid state: yet another belief of some (not all) audiophiles who, a bit like with negative feedback, think that solid state components are “infectious” and that their contact with the circuit can “infect” the sound, inoculating the disease of solid state sound. Here the ARP34 has the cathode biased with a white LED, such biasing executed correctly is like biasing it with fixed bias, therefore no capacitors under the cathodes.
There are many other dogmas, but I only broke these 7 😛 In the end there are so many things that are supposedly not good for sound, listening to the usual gurus, that you could not do anything except adopt very limited, timeworn circuit solutions that very often end up with having to shell out a lot of money. Many of these people willingly tie their hands and feet, brainwashed by the relentless proselytizing of forums and trade magazines, unknowingly ending up in a whirlwind of compulsive spending to buy components made expensive by fashion, often without achieving results that are worth the money spent. Many friends and clients who had the chance to listen to “the blasphemy” and its second version were astonished by what it manages to do by breaking rules (senseless ones) and with modest cost components. I say cost and not value, because the value of a thing depends on what that thing manages to do, while cost is something dictated by the market and value and cost do not necessarily match. Let me give the example that, going back about 20 years, audiophiles literally spat on the Fivre 2A3 bi-plate, they did not want them even if you gave them away, whereas today they are willing to sell a kidney to have a pair. Or they are willing to spend a lot for things that are worth nothing.
Let us take a quick look at the tubes used, 5C15 fivre…
I own about 80 of these tubes from military surplus, all absolutely NOS. I bought them from a radio amateur at a bargain price. It took me a while to trace their history, by reasoning it out (because I have no official sources). What I know for sure is that these tubes were produced by Fivre during World War II, when due to the embargo our factories had lost contact with American factories and therefore produced things as they saw fit or by copying tubes found in equipment that had crashed on Italian soil.
The 5C15 fivre is a transmitting pentode with directly heated filament. In the few official data available we know it has 15 watts of plate dissipation and a 4 volt filament. The glass envelope and socket format is that of the 807. Searching for the 5C15 designation you find on Radiomuseum at this address http://www.radiomuseum.org/tubes/tube_5c15.html a tube of English production that apparently has the plate built with the same shape, but this is the only feature in common, in fact the English 5C15, besides having a tubular glass, is a tetrode and not a pentode, the third grid present on the Italian version is missing.
Later I was able to acquire the curves of these tubes on a Sophia curve tracer and the tube proves to be extremely linear and suitable for audio use both connected as a triode and used as a pentode. By clicking here you can download the PDF of the datasheet I created with all the data I found about this tube, including the curves acquired with the Sophia: ds-5c15fivre
From these data and with a bit of research I then discovered that electrically the 5C15 fivre is a copy of the famous 307A – VT225. A tube much appreciated in audio use, especially in the USA, where they are plentiful. You can search for “307A tube” on Google to find a ton of high-fidelity tube projects and builds using this tube. Thus the plate curves of the 5C15 fivre and those of the 307A, as well as the pinout, are identical. The only difference that remains is the filament voltage which is 4 Volts for the 5C15 Fivre and 5 Volts for the 307A. This difference suggests that the filament of the Italian tube is more robust, in fact lower voltage, higher current equals thicker filament.
ARP34
ARP34 is a small military pentode of English origin, also produced during World War II, used in receivers typically as an intermediate frequency amplifier. I found about fifteen of these tubes in a lot purchased some time ago. Apparently useless in radio restoration, in reality it is an equivalent of the EF39 and several people have used it successfully in audio preamps and even as a driver for the 300B (although I believe it is too weak to be able to properly drive a 300B). The tube is variable mu, however the zone where mu suddenly drops is found low down in the curves, when the grid becomes quite negative. With an input signal of a few volts peak to peak it is very easy to bias it staying within the first zone and thus make it operate in a perfectly linear way.
PCC84
Not much to say, the PCC84 is a TV double triode, progenitor of the better known P/ECC88 / 6922 (those who turn up their nose, “haaaa the 84 is TV, the 88 is audio..” well, disillusion yourselves, the 88 is also a TV tube, born for the same purpose as the 84). I needed a little buffer, although the 5C15 as a pentode needs just 50 Vpp to be driven, the ARP34 was already sagging a little (not much) when driving it. In fact I do not understand how some can be so naïve as to drive a 300B with it. Long live distortion, meh. Anyway I had some lying around and I never give these away either because common audiophiles are too picky to accept tubes with a designation they have not heard at least a million times. Since deriving pentode models on SPICE is a pain, I experimented a sketch of the circuit on the good old wooden board, just to understand what I could get out of it, how to power it and whether there were any particular problems to watch out for in the final assembly.
The output transformer I adopted is the SE5K6-UNI. In the image below you can see the load line and the chosen operating point, the 5600ohm of the SE5K6-UNI fit like a glove:
In the final version of the circuit, after some empirical testing, I kept G2 at 350 volts and G1 at about -47 V, to reach a neat 4 Watts RMS. More or less the same power can also be obtained in triode, but a stronger drive is required and since I always see only triodes I wanted to do everything with pentodes, both preamplification and power section, for a change. I also had some transformer cases plus a power transformer and a working choke left over from a demolished Chinese amp.
So I decided to recycle them to make something with an aesthetic a bit different from my usual builds. I made a cabinet in elm wood polished with shellac and the aluminum plate totally home made, without relying on expensive workshops.
Here are the Chinese small cases repainted…
Output transformers potted in paraffin, the cardboard is used to keep them from contacting the metal plate.
Mounting plate complete, the octal sockets and the noval one are absolutely salvaged parts.
Also to do something different I made a solid state “servobias”. It is a small circuit based on a pair of op amps that measures the plate current of the output tubes and varies the grid negative voltage, automatically adjusting the bias without the need to intervene with screwdrivers on trimmers every time, to tweak the bias of tubes that need to settle and are no longer new or when they are changed. You just insert any tube in the socket and turn on the amplifier, the circuit in about ten seconds settles its bias at a very precise value and keeps it stable over time even if there are variations in the mains or in the transition from cold tube to hot tube after a few hours of operation. Like in self-bias, with the difference that the tube polarization is in effect a fixed bias, therefore there are no capacitors under the cathodes that can introduce colorations in the sound.
I then started assembling the actual circuit. The chokes you can see, believe it or not, come from faulty and scrapped PC power supplies. Some power supplies (the serious, heavy ones) have filter chokes on the low voltages, these specifically are 60 mH with an RDC of 1,8 ohm, absolutely with an air gap (thus suitable for filtering direct current). I used them to power the filaments of the 5C15 with direct current, creating a CLC cell preceded by a Schottky rectifier.
This is the finished unit, the knob is temporary, made with the 3D printer. As soon as I have time I will have one made in turned aluminum.
Instrumental data:
Maximum power 4 Watts RMS per channel
Bandwidth @ 4 Watts: 10 Hz – 40 kHz -1 dB
THD @ 1 watt 1.4%
Damping factor DF: 5.71
Rout: 1.4 ohm
Input sensitivity: Clipping with 4 Vpp at the input.
The unit sounds very good, let us see the spectrum analyses:
Square wave at 100 Hz
Square wave at 1 kHz
Square wave at 10 kHz
The circuit generates second and third harmonics at about the same level. Although many say that the harmonics must be “stepped”, this is not an absolute rule and when the overall THD level is very low you do not hear sonic differences. On the contrary you can clearly hear the capacitors used. So there are no sounds of broken glass or robotic voices as many believe, who obviously rely only on hearsay and have zero experience with actual instrumental measurements. The reality is that although I now have these results, by changing tubes with others, even of the same type or brand, harmonic generation also changes and in a more or less random way one can also experience stepped harmonics simply by changing tubes. I explain these phenomena well in this article to demonstrate that instrumental measurements are anything but useless as some golden ears believe, instead they explain many things.
Small subsequent tweaks
Since the sound was not yet perfect I added 2 polypropylene bypasses on a couple of electrolytics in the power supply, clearly improving the clarity in the high range.
The second version “GREY”
After making small modifications and improvements to the first prototype for a few years, I had probably reached its maximum, already very high but which could not be improved further due to limitations of some components employed. In addition the Chinese power transformer started to buzz after a few hours of operation. So, given the excellent experience I had, I thought of rebuilding it seriously and without savings. I started by designing a chassis in 3D and began assembly, this time with all new parts and not salvaged ones.
New, more advanced servobias circuit
Suppressor grid hack
As I wrote at the beginning, the 5C15 fivre is a copy of the 307A. By consulting the 307A datasheet one learns that this pentode was built for RF transmission in amplitude modulation, where basically the RF carrier reached G1 while its modulation was obtained by sending the audio signal to G3. This means that G3 has a construction that makes it relevant to the electron flow and therefore it was worth investigating a possible polarization of it.
After some tests on the uTracer I found an optimal operating point for G3 at +40 volts. In the graphs below you can see the pentode curves with G2 at 350 volts, on the left with G3 at 0 volts and on the right with G3 at +40 volts. As you can see, the slightly positive bias of G3 straightens the hump of the plate curves making the tube much more linear and at the same time lowers the G2 current. In practice, G3 at 0 volts becomes a brake for the electrons traveling between G2 and the anode. By biasing it positively you remove this brake. By biasing G3 at an even higher voltage than 40 volts it starts drawing current and the plate curves drop again. For those wondering, it is not useful to connect G3 to the anode because you get a tetrode with all the stability problems of tetrodes. I could not acquire curves in this mode because the tube oscillated and froze the uTracer CPU every time. With this trick I managed to get 1 more watt at the speaker and lower distortion.
G3 = 0 volts G3 = +40 volts
Assembly in progress…
The top side
I replaced the PCC84 with an ECC82 (but only because I had run out of P/ECC84). In this specific circuit nothing would have changed at the sonic level. Note the ARP34 “naked” 😳 , if you look at the photo at the beginning of the article the ARP34 have a gray coating, it is a conductive paint connected to pin 1 that acts as an RF shield, very important in a radio receiver. Almost all the ARP34 I had available, however, had the paint damaged by the years, which crumbled even though they were new and never taken out of the box. Since for audio use this shield was not found to be important, I cleaned them with a scouring pad, aesthetically they are very beautiful with a mirror getter that covers almost all the glass, leaving only a partial view of the internal electrodes.
The aesthetics of the cabinet deliberately echo the look of vintage equipment, including the color.
Instrumental data:
Maximum power 5 Watts RMS per channel
Bandwidth @ 4 Watts: 10 Hz – 40 kHz -1 dB
Total harmonic distortion @ 1 watt: 0.34%
Damping factor DF: 5.71
Rout: 1.4 ohm
Input sensitivity: Clipping with 4 Vpp at the input.
Spectrum analysis
Bandwidth at 1 watt on resistive load
Bandwidth at 1 watt on reactive load
Square wave 100 Hz, 1 watt
Square wave 1 kHz, 1 watt
Square wave 10 kHz, 1 watt (improved compared to the first version)
Triangle wave 1 kHz, 1 watt
Currently my milkyway is playing paired with the granny 27 and the combination of the two is truly spectacular. The bass precision and the refinement of the mid high range are truly a luxury, you can hear every tiny sonic detail, clean, clear, pleasant and never annoying. Female vocals are truly celestial.