RH813 – The Flagship Project

After designing several amplifiers with pentode and beam tetrode output tubes, and trying multiple options in the Universal and TTA amplifiers, the time has come for a “exclusive tube type” project. This amplifier was designed for the 813 beam tetrode tube, and due to size, voltage, heater voltage, and socket constraints, it is not possible to make it more universal than it already is. GK71 output tubes could be used instead of 813, but it would imply both different socket wiring and cap, and different heater voltage.

This project is by all means a “flagship”, in terms of output power, complexity, and costs involved. While I have tried to make it as easily reproducible and simple as possible, the voltages are nevertheless higher than in most common DIY designs, and the output power requires very serious output transformers. In order to make the amplifier as smooth as possible in operation, several additional features are introduced – thus in the end anyone can operate it without particular technical awareness: just turn it on, and enjoy.

The complexity and costs involved in this project imply that without the help of several DIY enthusiasts who post at the Audio-Talk forum, in particular Paul Barker, and the help of Per Lundahl of Lundahl Transformers, I would not be able to present it. Thus I would like to express my gratitude to those generous people before exposing the basics of the project.

 

Design Goals

The first and foremost design goal with this amplifier is achieving high output power – 30W and more, while using only one output tube per channel (i.e. pure SE operation). The 813, being a beam tetrode tube, lends itself very conveniently for that task, particularly having in mind that beam tetrodes and pentodes are more efficient than triodes.

Most high power SE amplifiers are built with transmitting triodes, like the 211, 845, or GM70. Leaving aside the question whether all these amps really achieve above 30W power ratings, truth is that all are operated at very high voltages, the B+ being 1kV and above. All voltages involved in tube amplifiers are dangerous, but it seems that everything above 500V requires extra care both in components choice, and how the amplifier is approached and measured. At high voltages arcs jump easily, and people can get electrocuted. That said, my idea was to keep the amplifier operating at voltages as low as possible, while achieving the 30W and more output power set as a goal. Most common digital multi-meters can measure DC up to 1kV, thus I assume that keeping the power supply around 800V is reasonable enough choice.

High power SE amplifiers with transmitting tubes operated at high voltages imply another difficulty – the fact that the driver tubes are usually operated at much lower voltages. While this can be easily solved with an additional power supply, my design goal is to keep it reasonable and simple – one power supply should be enough for all stages of the amplifier.

 

Design Choices

Keeping the B+ around 800V greatly simplifies the requirements for the capacitors – I have chosen to use motor run “cans” rated 600V AC, which should be safe for at least 1kV (and probably more). Besides the fact that those caps impart a particular sonic signature to the amplifier, used in this application they can be expected to last very long – the DIY-er will probably never have to change them. Another option is serial connected caps, where caps rated 500V DC WKG in series would be more than adequate – and due to the nature of the tubes used, and design choices made, even 450V DC WKG caps in series would probably be safe enough. Of course, the standard 400V caps could be used in a series of 3, for a very safe total rating of 1200V. The capacities required are relatively low, thus keeping the power supply as cheap and feasible as possible.

Putting capacitors in series decreases the total capacity, but by adopting a power supply with 2 chokes ripple is easily kept at bay. The example power supply illustrates the ripple at the load being just about 5mV (much less than most SE amplifiers), while ripple at the first cap is less than 150V – easy task for any motor-run cap (which is why they last so long in tube amps, and should be applied to the input cap position). With a hybrid bridge, output voltage will be higher for the difference in voltage drop between a 5U4 and the solid state diode used for the lower part of the hybrid bridge. The two chokes used, LL1638 – have a DC resistance of just 36 ohms, which helps improving the performance of the power supply: if different chokes with higher DCR were to be used, the additional voltage drop across the chokes should be accounted for.

The 813 beam tetrode has rather high CCS (Continuous Commercial Service) ratings, with 2250V maximum anode voltage, 1100V maximum screen grid voltage, 180mA maximum anode current, 100W maximum anode dissipation and 22W maximum screen grid dissipation – all intended for class AB1 operation, and even higher for ICAS ratings… at about 750V across the tube and slightly above 100mA cathode (total) current, this tube is literally “ticking”. Ug2 is set just below Ua, thus no single parameter is at more than 70% of the CCS rating, which makes this a very safe operating point, and probably should allow the output tube the longest possible lifetime.

Screen grid voltage was chosen to be just below the anode potential. This choice is dictated by the peculiar curves of the 813. The datasheet shows a slight kink in the lower part of the curves (well, they are actually more like straight lines), which is the least pronounced at low screen grid voltages, while it increases with the increase in screen grid voltage. Thus we might imagine that 400V Ug2 would be a good compromise – but unfortunately such choice would limit the power output to maybe 15W, which is definitely not worth pursuing in a project of this magnitude. While at 700V Ug2 the kink is more pronounced, the spacing of the curves allows for approximately 35W power output at several load resistance settings.

The curves show that 35-36W output can be achieved at similar operating points with both 10k and 5k load resistance. After trying (approximately) 5k, 7k, and 10k load resistance, I have chosen 5k as the load resistance for this amplifier, on the basis of “best sound”. The increase in load resistance does not yield appreciable improvement in performance – actually, both output power/distortion ratio and sound presentation are favorable with an approximately 5k load.

 

Voltage Regulation Tubes and Driver Section

The driver section is very similar to what already applied to 2^nd generation RH amps – RH300B and the RH2A3 family of amplifiers (RH-TTA). The ECC81 acts as driver tube, while the cathode follower tube makes life easier for the driver providing high input impedance, while at the same time it drives the output tube with low output impedance. Although the amplifier operates in A1 class (driving the output tube in voltage only), low output impedance is beneficial with most direct heated tubes due to the gradual increase in control grid current with the increase of driving signal values. Besides that, the 813 requires a relatively low valued grid bleeder resistor – another good reason to apply a cathode follower to driving this tube.

Just like in the other amplifiers mentioned that share the driver topology – the cathode follower tube could be almost any tube that can meet the current draw, voltage across tube, and anode dissipation requirements. The cathode follower tube is direct coupled to the driver tube, thus it is the driver tube that sets the operating point for the cathode follower tube – the current draw and voltage across the tube will be almost identical for any suitable tube used. In this amplifier I have chosen to use the 6SN7, but the same task could be performed by ECC82 or ECC81 family members – or other similar tubes like the 6J5, 6N7… For this driver topology it is important to correctly set the potential of the heaters circuit by connecting the heaters to a voltage divider like shown in the schematics. The driver tube will have the cathode at a very low potential, while the cathode of the cathode follower tube is going to see a relatively high potential, dictated by the voltage at the anode of the driver – depending on the particular tube used and a few other details, this will range between 100V and 160V DC – while the maximum allowed cathode-heater potential difference is 100V DC.

The purpose of the voltage regulator tubes in the circuit is not strictly shunt regulation of the voltage feeding the driver and cathode follower tubes. Actually, regulating this voltage is not particularly necessary in a class A circuit where there is little variation in current draw between minimum and maximum output power. Thus the purpose of these tubes is to facilitate smooth operation and to present a fail-safe feature.

If a directly heated rectifier tube is used, at power up the B+ will rise quite fast to its maximum level. Since the output tubes are directly heated and compelled to draw a set current (CCS circuit below the cathode), they will start drawing current immediately, and B+ will not rise above normal operating levels. This means that until the driver tube starts drawing current, the anode will be at a very high potential, and will set the grid of the cathode follower tube very positive as well. As the cathode follower tube gradually heats and starts drawing current, the cathode will be at a very high potential and arcing may occur between the elements inside the tube.

While there are several other mechanisms that could avoid this condition (pre-heating the driver tubes, placing a diode between the anode of the driver and the cathode of the cathode follower – allowing the current to flow across the anode and cathode resistors in series until the cathode follower tube heats up and starts drawing current, placing the cathode at a higher potential than the anode of the driver tube, which in turn makes the diode “disappear” from the circuitry), the solution with the voltage regulator string is in my opinion more elegant and predictable, and it offers a fail-safe function as well.

With the rapid rise of the B+ at power-up, the voltage regulator tubes light and start drawing current, thereby the driver and cathode follower tubes are immediately transferred from an 800V environment into the 400V environment for which the circuit was designed. As the tubes heat and start drawing current, the voltage regulator tubes will draw less and less current until the desired equilibrium point is achieved. Using 0A2 and 0B2 VR tubes in various combinations allows a fine tuning of this operating environment, between 324V (3x0B2) and 450V (3x0A2). In this manner, the tubes in the driver circuitry are not stressed at power-up, and in operation “enjoy” some voltage regulation.

The fail-safe function of the VR tubes becomes apparent in case one or both output tubes do not start (for instance, heater problems) drawing current. The VR tubes will draw at least 30mA current (actually, in such extreme conditions they may draw more current, up to probably 60mA, which should not prove damaging for a short period of time) providing some current draw to the power supply, keeping the voltage from exceeding a given maximum (in this case, it is almost impossible for the power supply as designed to exceed 1000V DC, thus the recommendation for 1000V DC caps rating, i.e. 500V in series: in practice, the VR tubes would draw up to 60mA and the B+ would not exceed 950V). It goes without saying that the VR tubes would ignite – being cold cathode tubes, they do not need anything else besides a minimum voltage potential to start (provided they are in good operating condition, and these tubes last quite a long time).

 

Heating the Output Tubes

The 813 has rather high power requirements – 10V at 5A – which is basically 50W. Compare it to the 5V 1.2A or 6W required to heat a 300B, and most problems become apparent.

Let’s start with what I see as the most important problem when heating direct heated tubes in SE amps – hum. With 50Hz (60Hz) AC, if a hummer (humdinger) pot or similar arrangement is applied to cancel the basic harmonic, we are nevertheless left with the residual 2^nd (and some higher) harmonics, i.e. 100Hz (120Hz) hum. The level of hum that can be heard depends mostly on heater voltage (increases for the square value of the heater voltage, i.e. it is much higher for 6.3V tubes like the 6B4G than for 2.5V tubes like the 2A3), and the gain of the tube. Since the gain of the tube in RH amps is decreased by the feedback loop between output tube and driver tube, we can enjoy the luxury of AC heaters even with a 300B, probably even on high efficiency speakers. But hum gradually increases with the increase in heater voltage and tube gain, and with a 2E22 it becomes difficult to ignore it completely, at least in the vicinity of the speakers. The 813 requires 10V and has rather high gain, thus 100Hz hum is virtually impossible to avoid, even in an RH type amp, and with relatively inefficient speakers: while it may not be disturbing at the listening position, the presence of hum probably influences the sound adversely when it comes to fine details in the music.

One obvious solution is DC heating, which should be hum free. I personally dislike the effect DC heating has on the sound, a specific type of detail muffling and bass tone exaggeration, and am aware of the other problems inherent with DC heating, like the difference in heater potential from one end of the heater to the other, or the need to switch current direction (polarity) at least from time to time in order to keep the filaments in best operating condition. Of course, the diatribe between AC or DC for the heaters of directly heated tubes seems to be never ending – both sides can rely on some advantages, and since DC is by far more complicated, the supporters of the latter may be inclined to last longer in this dispute, either due to stakes held (for instance, offering solutions for sale), or due to investments performed (buyers of such solutions). But at the cost of being accused of engineering (as opposed to audiophile) mentality, I would like to avoid the AC vs. DC sound issue, and stick to what matters: feasibility.

5A of current draw means that at least 9A will be drawn if rectification is applied. Adding some voltage losses in the process, we are safe at estimating that at least 100W of power is required for each heater – which is literally the double of what we should need with AC. Just rectifying with brute force, like high valued caps, does not present the best solution, neither as current inrush nor as final result (ripple). Thus some form of regulation is needed to filter out the ripple without investing in heater chokes and/or high valued caps. The most basic form would be voltage regulation, which is as well more energy efficient than current regulation (lower voltage drop). If we assume using very low drop Schottky rectifier diodes, and very low drop regulators capable of 5A, like the LM1084, we might be able to do it with the least energy loss. Still, at 1.5V drop and 5A current draw, we are facing at least 7.5W of heat for each regulator, which in turn requires using a hefty heat-sink, or a heat-sink with fan (a good heat-sink for older Athlon processors will make short work of the heat dissipation).

In the end, we get very low (but still audible, at least I can always hear traces of it) hum or buzz (I like to call it solid-state regulator buzz), and DC on the heaters (which I consider sub-optimal in operational terms, regardless of the sound quality discourse). A lot of energy lost in heating, a lot of complication and effort – and that is why I mention the residual hum or buzz: after going at that long a distance, I expect to get “no hum whatsoever”. Of course, even more elaborate alternatives are possible (several high-quality DC heater solutions are known, advocated, and discussed all over various forums – which pre-made solutions I have not tried due to obvious reasons), and you can always choose the car battery way… but even the car battery adds something audible, some (incredible but true) hum or buzz persists even with battery operated filaments (chemical buzz?).

Another solution is high-frequency AC, better known as “electronic transformer for halogen bulbs”. You can use adequately sized (minimum-maximum power) electronic transformers straight out of the box, and drop the voltage with resistors (0.15 ohm per heater leg will do the trick), but this solution is actually almost identical to what we started with in the first place – mains AC. The output of the electronic transformer is high frequency AC (usually between 30 and 60 kHz), but it is “modulated” in 100Hz format. Once your humdinger circuitry cancels the basic harmonic, you are left with 200Hz hum, which is slightly less audible than the standard 100Hz, but to my ears it is also more annoying (it actually resembles the electronic buzz). In order to be used for this application, the electronic transformer needs to be modified.

There are two basic modifications to be performed: filtering the initial 100Hz ripple, and setting the correct output voltage. It is rather easy to filter the initial ripple by placing a 270 to 330uF/400V cap (for European 230-240V AC mains) between the positive and negative pole of the bridge that rectifies the mains voltage at the input of the electronic transformer circuitry – but this will in turn cause the output voltage to rise. The rise of the RMS output voltage is proportional to the increase in DC voltage after the bridge (due to the decrease in 100Hz ripple, the average and RMS voltage will be much higher), and with the cap values mentioned is approximately 1.49x the original RMS voltage of the unit. While this excess voltage can be burnt with resistors, due to the high current draw this discourse borders with impractical: we are talking about some 35.5W dissipated in resistors, per heater (per tube). A modification of the output transformer (ferrite core, usually toroid) is necessary, lowering the number of turns in the secondary winding. This is rather easy to do with most electronic transformers.

What is not so easy is getting measurements. In order to measure HF AC, you will need a True RMS DMM capable of measuring at least 50kHz – or a scope. The output of the electronic transformer is not a sine wave, but a square wave, and since this square wave is not perfect, the RMS value is lower than that of a perfect square wave (which would be equal to the peak to peak value). Thus either you have an adequate True RMS DMM, or a scope – or you do a lot of estimation and calculus in order to get the correct value at the output.

Due to the complexity of the measurement and calculation, this rather easy procedure requires a lot of additional explanation. Since I am quite taken by the results achievable with HF AC for heating direct heated tubes, I intend to post about it in the future with more details. For those willing to try it, and not particularly afraid of learning by trial and error, the whole procedure consists in adding a cap, and unwinding some turns from the secondary of the ferrite core output transformer. The reward for using HF AC will be hum-free sound (less hum than with DC) with all the dynamic qualities of AC on the heaters. The best of both worlds: This might sound like repetition of what I already stated about other solutions in tube audio, but that exemplifies very well what I am after – finding better ways to use what is literally in front of our eyes, but remains unseen or misunderstood.

 

The Output Transformers

As already explained, I have chosen approximately 5k as the optimum load for this amplifier based on a combination of sound and performance. While 10k could offer approximately the same output power with slightly different parameters, 5k sounds better – more natural and true to life.

The problem with output transformers for an SE amplifier this powerful is just that – output power. While most SE transformers might be safe at the voltages involved in this project (the B+ is well below 1kV, and even paper insulation should be safe up to 1.5kV if properly applied), and the total DC current in the primary is not that high as to present a particular problem in terms of size – most transformers on the market just cannot provide adequate performance in terms of bandwidth at high output power. The main or most evident difference between an SE amplifier with 300B tubes, and an amplifier with 813s is output power – we are talking about maybe 10W against 35W. Even if safe, an output transformer for the usual SE amps will not be able to perform correctly at high power levels. This fact, and the availability of 813 tubes, has been keeping me from developing the project for quite some time.

The first pair of output transformers used in this project, courtesy of Paul Barker, are huge and potted units, designed for use with 845 and similar tubes – weighing over 7kg each! I have used these transformers in the initial phase of the project, and they are definitely up to the task – lots of bass and no constraints on output power. Due to multiple taps for the primary (5k, 7k) and the secondary (4, 8, and 16 ohms) – those transformers are easy to use as 5k, 7k, 10k – and even 2.5k and 3.5k if needed.

The second pair of output transformers used in this project come from Sweden, courtesy of Per Lundahl. While looking like the usual Lundahl transformers – C core, dual bobbin – they are actually twice the size and weight of the LL1623 I have used in the RH2A3/RH-TTA project. Those transformers - LL1688 are intended for 845 and similar transmitting triodes, and can easily be set for 5.5k, 9.5k, or 19k primary – with 4, 8, and 16 ohm speaker options. Since this time I was after just one setting, there is no wiring harness or switches – just 5.5k primary into 8 ohms.

While the huge potted transformers performed very well indeed, with great bass extention and lots of volume – I was surprised how much of a difference could a different type of transformer make in the same circuit, and even one which due to feedback mechanisms has lower output impedance and is therefore less demanding than no-feedback types. The improvement in mids and highs, the general liquidity of the sound, and sound-staging was something I did expect – but the LL1688 is better even in the definition and perceived depth of the bass notes. Overall, a clear winner.

 

Spice Models and Performance

I must admit having started this project without a proper model for the 813 output tubes. After all, some basic driver details are already known, and all I need are the curves – to draw some loadlines. Well, it seems that this is the sub-optimal way of designing amplifiers. Having a loadline is not enough, since the operation of the amplifier is represented by the interaction of the output tubes and the driver section – besides graphic representations of output power and distortion, and the eventual feedback applied, the distortion cancellation effect is very difficult to estimate…

While my 813 model is far from perfect, it is better than most models that can be found for pentodes and tetrodes. The difficulty is getting the “kink” in the lower part of the curves. As for the driver and cathode follower tubes, I have improved all the spice models for triodes that I am currently using to Koren 8 parameter models. Curves from the ECC81 datasheet are shown superimposed with curves obtained from the respective spice model. At this point, what the simulation shows is definitely how tubes should behave…

 

The Sound and General Impressions

I am always reluctant when it comes to discussing the sound of my designs – I prefer to hear from others how pleased they are with the various RH amps they have built. While I tend to explain in great detail the technical solutions, their eventual elegance and feasibility – the perception of music reproduction is much more subjective.

This amplifier differs from my other designs in complexity and power. Power was a design goal, while the added complexity was necessary in order to keep it user and DIY-er friendly as the previous designs. This basically means that once built and checked, it can be used for hours, day by day, without thinking about it – or playing a little bit with the combinations of driver tubes and operating points made possible by the adoption of VR tubes.

But the sound is really worth mentioning. While the general signature is the same as in other RH amps, due to the nature of the feedback and distortion cancellation applied – this amplifier shows some additional characteristics besides what can be ascribed to the intrinsic quality of the output tubes, more as a consequence of the high power available.

Audiophile DIY-ers with high efficiency speakers are probably fully satisfied with the power a 300B, or even 2A3 amp can offer – anything between 5 and 10W is probably enough. But with “the usual” lower or medium efficiency speakers most people own – each and every W counts. What made me take the good sound of an EL84 in triode mode and achieve it in pentode mode with at least twice the power (the RH84 that has started it all), finds its final gratification in this amplifier. We are not talking here about 5W vs. 2W, or whether a 300B can give you 8W or maybe 13W… this is power above 30W, and it seems very easy and so logical – once it is done. But the journey that got me there was a long one.

With more than 30W at disposal, we are talking about “commercial” power, something easily achievable with PP amps in the world of shops and magazines. A simple PP amp with EL34 is capable of 30W per channel… but there is a huge difference in sound. Once power is not the bottleneck in music reproduction, we can get to enjoy what was offered or hinted by lower powered amps. A nicely defined bass note becomes not only nicely defined and precise, but a loud, fast, and almost brutal attack! A well-defined piano note that had texture or vibrancy now has power and attack; it resonates in air without being distorted, almost regardless of listening volume. All of it by far exceeds what can be had from commercial PP amps, and explains why there is so much hype about the sound of SE amplifiers.

Endowed by the same sonic signature found in other amplifiers of the RH series while apparently being free of limitations due to its high power, this is definitely the “flagship” RH amplifier – with all the inherent characteristics. Some might ask whether it has the finesse of a 2A3, and the question would be a logical and frequently found one – after all, since the “Ongaku” conquered the world of commercial High End by blitz, there have been so many attempts to scale down the concept. But it is not a fair fight: at low listening levels the RH813 is just as nice, liquid, and enjoyable – as the RH2A3, to name one example. But what it can do transcends the “sound finesse” issue: where it can go, the RH2A3 cannot follow. I guess the question is not absolute, and needs to be rephrased – how efficient are your speakers? Because, if you own very efficient speakers, you probably do not need to go the length necessary, both in financial or psychological terms (higher voltages involved, and higher power consumption/heat dissipation – if that is of any concern, not to mention the fact that bigger is usually more expensive), to involve yourself in an affair with such a behemoth amplifier. On the other hand, the only reason why would you content yourself with a lower power amplifier is either not being able to use it (the neighbors might complain), or not being able to overcome the mentioned financial or psychological issues involved.

 

An Adequate Box and Other Issues

Starting with the size of the output tubes, and the maximum dissipation (CCS) possible, as well as the heater requirements in terms of power necessary – it is obvious that an amplifier like this is not easy to accommodate in a box, or on a shelf.

The design is relatively efficient – as SE amps can be. By adopting a toroid power transformer a great reduction in size, heat, and radiated field is achieved. While the toroid transformer could be a 300VA unit (output tubes heaters not operated by this transformer), it is rather difficult to wind the 750V secondary on a 300VA core. Therefore a 500VA core is probably necessary – increasing both cost and size (but slightly) – but as an aside the power transformer will be colder in operation. As a matter of fact, the one used in my RH813 does not exceed 50⁰C – even after 6 hours of operation on a hot summer day.

The power supply chokes have very low DCR, thus they dissipate very little heat – and the same is true of output transformers. This would leave us with the system necessary for the heaters, but instead of 100W per tube – between transformers and heat-sinks for the regulators and transistors – there is just a rather small box containing the modified electronic transformer… and this box, besides being small and efficient, does not exceed 45⁰C after 6 hours of operation on a hot summer day. Ah, the beauty of power-efficient solutions.

Thus we are left with the various resistors in the tube circuit, and the tubes themselves, as the most important source of heat. While some of the ceramic resistors get quite hot in operation (70-80⁰C), they do not represent an important source of heat, and the same is true of the driver and voltage regulation tubes; this leaves us with the output tubes. At 75W dissipation, and 50W for the heater, the 813s get quite hot and radiate a lot of power: due to the various heat and power optimizations, they are the most important source of heat in this amplifier.

Putting the amplifier in a standard size and dimension box is very feasible – but the height of the output tubes, and the heat they radiate, means that the amplifier is not suitable to be placed on a shelf. The heat radiated from the output tubes goes a long way towards burning the shelf above… thus with an amplifier that is radiating mainly from the tubes, and is not suitable for shelf placement, should probably be built in a tall and narrow box. The box should be organized in two distinct levels, the bottom level containing the power transformer and the output transformers, while on top of this level lie the chokes and caps and other elements of the power supply. At the highest level, the opening of the box supports the audio circuit itself, where tubes protrude from a top cover which is supported high enough to enable air circulation. This is basically what I intend to build as a box for this amplifier – to be placed on the floor, beside the shelf with other amps and audio equipment. Although tall and narrow, the box will have a low center of gravity, keeping it stable.

RH2A3 – An Amplifier With a Twist!

During the final development stages of the RH300B amplifier a similar schematics for use with 2A3 tubes has imposed itself as quite feasible and logical. The 2A3 is a DHT tube just like the 300B, of similar general characteristics: mu is around 4 in both cases, while plate resistance is between 700 and 800 ohms. The main difference between these two tubes, heaters voltage aside, is maximum ratings – dissipation and voltage. Thus slight downscaling of operating voltage and current to create operating conditions acceptable for 2A3 tubes has been done before with several popular schematics.

This would be the general RH2A3 schematics – almost a mini-RH300B. With lower B+ and current consumption, cheaper to build due to lower power requirements, different heater voltage for the output tubes, and lower power with the same or similar output transformers. Just as in the case of the RH300B, at 5.1W and 1% distortion, output power exceeds expectations – and will probably be discussed and questioned in the future by all who have not built the amplifier themselves and thus ascertained whether the promise has been fulfilled.

As the schematics is strikingly similar to the RH300B, I shall not repeat all considerations related to the schematics itself – the only appoint that should be made, or rather repeated, is that the operating point of the cathode follower tube is determined by the directly coupled grounded cathode driver tube. Thus it will draw the same current regardless of tube type used – and this time I have chosen to adopt an octal socket, accommodating 6SN7 tubes, but 6SL7, ECC35, and other similar tubes can be used as well – you can go as far as 6BL7, 6BX7… even 6AS7 (or 5998). While it might elude logic to employ 6AS7, 6080, or 5998 (WE421 by TungSol) due to the relatively low current draw, and keeping in mind that these tubes have very high heater current requirements which would have to be taken into account – the sonic results obtainable with these tubes reflects their known sonic characteristics, and are quite interesting in the scope of this amplifier.

The twist

The RH2A3 can be built with regular 2A3 tubes, either NOS or current production: it seems that just like the current production of 300B, modern 2A3 tubes are good sounding, reliable, and meet the “single anode” construction so much sought after by some DIY-ers and audiophiles. And of course, that is the most logical way to go…

But I have built mine with “unusual” 2A3 tubes – on octal base. While the 2A3 tube was manufactured as well with 6.3V heaters and UX4 base (6A3), and further developed into a version with 6.3V heaters and octal base known as 6B4G – I am referring to “special edition” 2A3 tubes, with 2.5V heaters, mounted on octal base. The most notable application of these tubes have been the Audio Innovations First and Second amplifiers, as it seems that 2A3 tubes were custom ordered on octal base due to a lack of UX4 sockets at the time. In my personal opinion, while this is a plausible answer, I guess that commercial concerns were important as well, channeling the supply of spare tubes for the amplifiers.

Anyway, a different base just needs a particular socket, and besides complicating life a little bit trying to source 2A3 tubes on octal bases, this would not be much of a twist. In keeping with the inherent universality of my designs, at this point the 1619 enters the stage. The 1619 is a direct heated beam tetrode tube on octal socket. It has 2.5V heaters just like the 2A3, and is pin-compatible with the octal socket used for octal 2A3 tubes – pins 2 and 7 for the filamentary cathode, 3 for the anode, and 5 for the control grid.

So far so good - but what about the second grid and the beam forming plates? While the 1619 is used by some in triode configuration, with g2 connected to the anode – this application is almost reserved for the old radio collectors market. The 1619 is seen as a cheap replacement for expensive 45 tubes, and relatively elaborate adapters are being made or sold to allow plugging the 1619 in old radio receivers. In the case of the RH2A3/1619 amplifier, the second grid on pin 4 is connected via a zener diode to B+ (56V 5W zener recommended, setting the g2 voltage at approximately 250V), while the beam forming plates (pin 8) are connected to the shield (pin 1) and directly to ground. Furthermore, pin 6 – which is missing on 1619 tubes – is used as the “virtual cathode” point. Pins 4, 6, 8, and 1 are NC on 2A3 octal base tubes, just like they are NC (not connected) on 6B4G tubes (as a matter of fact, the octal base of the 2A3 octal is identical to the 6B4G base). Thus when a 2A3 is plugged into the socket, it is indifferent to the connection of these pins.

How to make it all work?

Both tubes have 2.5V heaters with similar current consumption, and 15W anode dissipation ratings. But the bias voltage of the 2A3, i.e. the potential of the cathode in respect to ground, is much higher than that of the 1619. At the foreseen operating condition of approximately 250V across the tube and (exactly) 60mA of current draw, the bias voltage of the 2A3 exceeds the ratings of the LM317 regulator. While this could be solved by adopting a TL783 regulator which is safe for input voltages of up to 125V (the main and probably only relevant difference between these two regulators is the adoption of a DMOS output transistor in the TL783) – the bias voltage times the current draw will require a relatively large heat-sink and lead to a potential overheating problem. The resistor shown in the schematics absorbs most of the voltage differential and lowers heat-sinking requirements for the active element (LM317), leading to more stable operation. I just somehow prefer to run solid state components as cool as possible, and leave the heat to the high power ceramic resistors. In order to operate the 1619 tube, the resistor must be excluded from the circuit, which is easily done bypassing it with a switch.

The difference in voltage across the tube will be dealt with naturally by the difference in tube type: since the 1619 is a tetrode, the total current draw is the sum of anode and screen grid current. Thus anode current will not be 60mA as with the triode 2A3, but about 54mA – as approximately 6mA will be drawn by the second grid. Changing tubes and excluding the voltage dropping resistor by switching it off will lead to a different set of operating conditions which suits the 1619 just as well as the other set of operating conditions (with voltage dropping resistor) suits the 2A3.

The output transformer

Another important difference between 2A3 and 1619 is the optimal anode load: while the 2A3 triode will perform at its best at values between 2.5k to 3k, the 1619 beam tetrode would require a load between 5k and 6k. To achieve this goal, we need a very versatile output transformer which would allow us to change anode loading at the flip of a switch (or switches).

The Lundahl LL1623 output transformer used in this amplifier, besides the highest build and materials quality, has a probably unique versatility and adaptability to various circuits, since both the primary and the secondary windings can be arranged to suit a wide range of currents and transformation ratios.

While the primary windings could be arranged in parallel to double the maximum current capability, this would lead to lowering the inductance, and since the gap is already set at the factory to suit for 60mA, 90mA, or 120mA of DC current across the primary – the best solution is to connect all primary windings in series, as shown in the datasheet. Thus the inductance of the primary will be defined by the chosen gap, i.e. required DC current value. Quite logically, the optimal gap for the RH2A3 amplifier is as set for 60mA DC current across the primary.

The secondary windings can be combined to achieve standard foreseen anode loads of 1.6k, 3k, and 5.6k – for an insertion loss that varies between 0.2dB and 0.8dB. Bearing in mind that all the above combinations are possible for 4, 8, and 16 ohm loudspeakers – it is obvious that the possible anode load is actually even more varied.

While the user might exercise his (or her) wits by finding alternative connection possibilities, an application sheet is available on the Lundahl site showing possible secondary winding connections that allow switching between “adjacent” anode loads by use of switches or bridges. Since my application is with 8 ohm speakers, I have chosen to prepare the output transformers for B/C, i.e. 5.6k and 3k loads into 8 ohm outputs.

While for the connections of the primaries I have chosen to solder pieces of wire, the much more elaborate secondary connections seem to require a different approach: hence the wire harness showed in the pictures. The three switches allow the choice between 3k anode load (suitable for 2A3) and 5.6k anode load (suitable for 1619). While the wiring harness looks complicated, and indeed requires a couple of hours to complete and check – there is no need to rush while building an amp. The building experience is indeed a source of pleasure, and once you finish building the amp, you will probably be using it for years to come: what would be the point of building it in one afternoon?

I feel confident with the solution I have adopted – but a further simplification and expedient could be a custom PCB, foreseen to be simply mounted and soldered on the pins of the secondary windings, greatly simplifying the build and avoiding the need to check and re-check.

Thus, to recapitulate: one switch to include/exclude the voltage dropping cathode resistor (2A3/1619) and three switches to select the optimal anode load for the tube (3k/5.6k). I guess that is not too much to do in order to enjoy two (slightly) different amplifiers from a single box?

Of course, the switch for the dropping resistor can be omitted if TL783 is used with an adequate heat-sink (dissipation will reach 2.5W with 2A3 tubes) – and with that touch the only possibility for error would be avoided: because, if you insert a 2A3 tube without the voltage dropping resistor, you will most probably kill your LM317 – and if you insert a 1619 tube with the voltage dropping resistor in circuit, it will not be able to draw the specified current, leading to possible overvoltage conditions for the power supply caps (which is why those are not rated for a tempting 400V that looks to be just fine, but would be easily exceeded if the output tubes do not draw current for any given reason). On the other hand, if you forget to set the correct load for the output tubes, the lack of power and early clipping will harmlessly remind you to check the switches…

How does it sound (and compare)?

I used to have for quite a long time a 6B4G SE “no feedback” amp that I enjoyed quite a lot and used it as a “reality check” at the time of the initial RH designs – RH84 and RH807. The amp was very straightforward and classic, I had conceived it as a two stage design with 6SN7 (grounded cathode, bypassed cathode resistor, resistor loaded anode, 6mA current) driver and 6B4G output (automatic bias via bypassed cathode resistor, 300V across the tube, 50mA current draw, 3.5k load). While the “no feedback” designation is a nonsense per se – as particularly in triodes, plate current is dependent on the plate to cathode voltage differential: as the signal tries to pull the control grid positive it causes a rise in anode current, but the anode to cathode voltage differential decreases at the same time, trying to pull the anode current down. Since anode and control grid are pulling in opposite directions, this is negative feedback by definition… (to put it as straightforward as possible, though simplified to the point of easy criticism) While the behavior of the triode suggests an “inbuilt” feedback mechanism, the term “no feedback” is more of a common lingo expression reflecting the fact that cathode degeneration was “avoided” throughout by adoption of bypass caps on cathode resistors (avoided, that is, from the high pass frequency on), as well as the lack of NFB either local or overall. While the feedback nonsense should be left aside, the reality check comparison is something I frequently do, and can only suggest it to everyone – as a means of avoiding the proud of ownership (or craftsmanship, or own design) syndrome and illusions about the perceived quality or quantity. The mentioned amp was a classic 3.5W 2A3 class amp (with good quality black anode 6B4G tubes) and represents a valid statement to compare the performance of the RH2A3 with.

One important detail is the hum inherent with AC heating operation of direct heated tubes: the hum level is proportional to the heater voltage, and the gain of the tube. Thus the most obvious advantage of the 2A3 in comparison with the 6B4G is AC related hum. In RH amplifiers, hum is obviously reduced by the application of feedback – so much so that it can be barely heard on the RH300B (5V heaters), and it becomes virtually inaudible on 88dB/W/m speakers with 2A3 tubes (2.5V heaters). On the other hand, the hum of a “no-feedback” 6B4G amp can be easily heard on the same speakers, although it might not be so distracting at the listening position (“humming pot” or not, it could be heard between songs during quiet night hours). If a 6B4G is used instead of 2A3 in this circuit, AC hum would be more apparent, but much less pronounced than in a classic “no-feedback” circuit.

First of all, the power output of this amp is nothing short of awesome, from a 2A3 amp – that is. While 5W will not rock most houses (unless those houses are equipped with very efficient speakers), it definitely (re)produces music much louder than a classic 2A3 amp – to the point that, depending on your listening taste, you would probably not need anything more powerful 95% of the time, even with 88dB/W/m loudspeakers. Obviously, the RH300B can go much louder – and does so effortlessly after the point where the RH2A3 clips – showing that there is a definite difference in output power, quite adequate to the 5W vs. 12W comparison.

This gap in power is reduced when 1619 tubes are used. While lacking good models for the 1619 I can only estimate the power to be about 7.5W – and the best comparison probably is the RH307A Super. While the curves of the 307A look more promising than those of the 1619, both are 15W anode dissipation tubes, and the difference in curves probably stems as well from the difference in internal construction – while the 307A is a direct heated pentode, the 1619 is a direct heated beam tetrode. The direct comparison shows similar power for both amplifiers… which is quite logical to me. Thus the option to use 1619 tubes does extend the possibility to enjoy music with this amplifier.

In operation, what makes this amplifier so nice and user friendly is current consumption and heat. If you sum up the requirements for the various secondary windings on the power transformer, it becomes apparent that a nice 100VA unit will be just fine for the task – which translates into nice and small, particularly in the realm of toroids. Not to mention further size decrease if you choose to wind two transformers, of which one for the tube heaters (allowing easy and cheap modification to use i.e. 6B4G output tubes). Current draw is slightly less than 140mA, allowing you to employ a nice smallish choke… while the schematics calls for a 10H 150mA choke, I have used a 5H 150mA unit without any ill effects (like hum). Cold and relaxed in operation, particularly when compared to larger amps, you can just forget about it once you have powered it up, and concentrate on the music (or whatever else you are doing that is just being complemented by the music).

How about the sound? While the amplifier quite predictably sounds remarkably similar to other RH amplifiers, it does present some unique characteristics due to the very high quality of the output transformers used. While the socket play is an interesting feature, in a way the amplifier is built around the LL1623 output transformers – allowing for easy adaptation to the optimal load for each tube used. The low insertion loss and high intrinsic quality of the transformer are clearly felt in the sound, compared to other good quality and large size, both EI laminations and double C (4C in fact – it should be called quadruple C, I guess) core output transformers used in the RH300B and RH307A, respectively.

An initial premise about the sound in the 2A3 version is dictated by the specific output tubes. Since I am not aware of other 2A3 octal base tubes beside the Shuguang, this is probably a limiting option for tube rolling and synergistic combinations. The “old type contemporary” double anode 2A3 tubes are probably inferior in quality to the best contemporary “mono-plate” 2A3 tubes, and the “fake mesh” 300B (most probably made in the same factory) which were sold under various brand names during the years, and are known for a specific sound signature favoring great highs and mids. The RH2A3 (with 2A3 tubes) sounds remarkably similar to the RH300B, albeit with lower power. Besides unexpectedly high listening volumes when compared to other 2A3 amps, the most striking points are great, well defined, and extended bass notes – as well as great mids, particularly lower mids. We could say that the amplifier does a great performance in the “melody range”. Differing from the “fake mesh” 300B, the RH2A3 sounds more similar to the sound of the RH300B used with EH300B tubes – but the bass is even better defined, the mids finer chiseled: I guess that with the extreme similarity of the schematics, and with the same or very similar drivers used, as well as the same rectifier tube, the merit clearly goes to the Lundahl LL1623 output transformers, easily outperforming the nicely wound and rather large (and heavy, too) EI105 lamination output transformers used in my RH300B. This is an assertion that I hope to check in the future by substituting the output transformers in the RH300B.

The 1619 is a direct heated beam tetrode, and beside an increase in power, few would expect it to introduce an improvement in sound quality… but as a matter of fact, it does. Compared to the Shuguang 2A3 octal tube, the black metal 1619 tubes are visually inferior, but their sound is better balanced overall. While the bass is slightly less pronounced, it is by no means inferior due to the increase in power – but the most striking difference is less pronounced mids, and an improvement in upper mid and highs quality. The mids are less in evidence, and the general tone is even more listenable and involving. Output power is on a par with the RH307A Super, and comparing to the large 4C output transformers in that amp, the LL1623 shows even greater finesse in the upper mids, a clearness of tone that becomes particularly apparent when listening to LP records. When clear and clean upper mids are mentioned, the first thing that comes to mind is a possible exaggeration of these tones when vinyl is reproduced – but it is quite the opposite, as the upper mids and highs are not exaggerated at all, while the impression is one of increased listen-ability. The upper mids and highs result extremely clear and clean, not intrusive or evidenced.

What if?

As already mentioned, the only 2A3 octal base I know of are the older production Shuguang made specifically for Audio Innovation, which one might already own, or find a means to acquire somewhere. On the other hand, the 1619 is basically a sleeper tube – visually unattractive, odd heater voltage, direct heated, and commanding relatively low prices (mostly driven by the radio collectors market). Besides the possibility to opt for building it as a classic UX4 socket amp – which is always a good solution per se, but leads to a loss of 1619 functionality – one possibility could be adopting UX4 to octal adapters, similar to those employed to make use of 5Z3 tubes in octal sockets.

Making such an adapter should be very simple and straightforward once you procure the parts necessary – after all, it is about soldering wires between the correct pins, and fastening the two parts together – and it would allow to listen to the amp with modern 2A3 tubes as well as the highly regarded mono-plates of times long passed.  I guess employing adapters would be much easier than changing the bases on 2A3 tubes (once you unsolder the pins and dilute the “glue”, it is easy to detach the base on any given DHT… all you would need to do would be fitting the wires into the proper pins on the octal base, gluing the base and soldering the pins. A good option to glue the base could be thermally stable silicone (for heating applications)… but that might mean pushing it too far, for most!

Another very simple alternative would be to build the amp for 6B4G tubes, which can still be found at low and affordable prices (Russian 6C4C – i.e. 6S4S) – and use 6L6 tubes as an alternative. By 6L6 tubes I mean specifically the metal 6L6Y, the glass ST version 6L6, and the several Russian 6П3С, i.e. 6P3S types – all those 6L6 types are rated between 15 and 19W anode dissipation, thus perfectly suitable to be used in the circuit instead of the 1619, not to mention the possibility to employ 6V6 and 6F6 types as well – and all that with just a few circuit differences (simple modifications to be performed on an already built RH2A3-1619) as shown in the schematics.

Besides 6.3V heater windings for the output tubes (6L6 types draw 0.9V which is almost identical to the 1A drawn by the 6B4G), the main difference is the cathode connection: instead of choosing between keeping or excluding the resistor in the circuit, this time it is about choosing between “virtual cathode” (made on pin 6 – which path includes the voltage dropping resistor) or indirect heated cathode (on pin 8). Pin 1 can remain connected to ground, thus the metal envelope on 6L6Y tubes will be grounded, and EL34 tubes might be used as well with their g3 correctly grounded via pin 1.

I would like to thank Mr. Per Lundahl of Lundahl transformers, Sweden, for his kind assistance, help, and support in the realization of this project.