Sunday, November 17, 2013

The RH300B Story

The RH300B project was the first envisaged in the 2nd generation series, but the last to be developed. The development of this amplifier was long stalled by material reasons, mostly because I had no 300B tubes to build and breadboard it. Thanks to the generosity of a DIY friend who sent me two pairs of these DHT tubes, I was finally able to breadboard, troubleshoot, and finalize the project.
While an initial circuitry was published on some forums and sent to some DIY-ers back in 2005, when the time came to finalize the project some design choices were to be reconsidered. Thus the RH300B introduces a different driver stage in respect to all previous RH amplifiers, while simplification was introduced in the power supply section.

For the first time, I was hesitant and reluctant to publish the schematics of this amplifier. A lot of work, knowledge, and insight have gone into the design of the amplifier, and the various difficulties encountered on the way clearly show that these projects are not to be taken lightly. Good ideas and lots of knowledge are not enough; building an amp definitely requires some material means. RH amplifiers have met the criticism of all those who felt threatened by the publicity gained in particular by the RH84, and while many DIY-ers have built and enjoyed one of the several designs, the RH series has also spawned lots of copy-cats, and word is out that some are building RH amps for their local clients – mostly DIY-ers without the time at hand, or the experience required to build their own amps: while the clients should not be judged for feeling challenged by the prospect of building a tube amplifier, since it is among other issues dangerous as well (these voltages could kill you), the “hired builders” should at least give some credit to the original designer and avoid changing important parts just to make it look different at first site. There is a site (I will not paste shortcuts, but you can easily find it based on the description) showcasing a parallel SE amplifier with 807 tubes which is basically a copy of the original RH807 adopting 2nd generation solutions taken from the RH88, while the only difference is fixed biasing with negative voltage instead of a cathode resistor under the output tubes; the site owner has reluctantly agreed to mention my work by adding a blurred statement how his amp is in fact inspired by work of a compatriot of his, whose triode connected 807 amp (sic!) was in turn inspired by one of my designs… is there any shame left in people? Thus this amplifier represents the last in the 2nd generation series, published for all DIY-ers.

The Schematics

As it is evident from the schematics, the driver section now has a cathode follower directly coupled to the driver itself. While the driver is a rather standard choice for RH amplifiers – an ECC81 family tube, with anode-to-anode feedback connection to the output tube – the cathode follower portion of the driver is not. The cathode resistor of the driver tube is split in two, of which one resistor is bypassed with a capacitor – which is also a first in RH amps. This particular split is necessary in order to maintain a correct biasing of the driver tube while increasing the gain, in order to keep the input sensitivity below 2V RMS for full output (as declared at 1% distortion).
The cathode follower is not strictly necessary, as the driver tube is well capable of driving the output tube on its own. This particular detail appeared as a necessity while contemplating biasing methods for the output tube, since fixed bias requires the adoption of a grid resistor lower than 50k. With such a load, the driver was not capable to perform its task adequately, and the cathode follower represents the best way to solve the problem, since it does not invert the signal nor add any gain, while it is possible to couple it directly to the driver tube – avoiding additional capacitors in the signal path.
In the end, fixed bias was discarded, since the only apparent benefit would have been a lower B+ voltage. On the other hand, the negative bias power supply represents and additional cost, complication, and a source of potential trouble for the DIY-er. Furthermore, it would require checking the bias voltage of the tubes from time to time, as well as adjustments every time the output tubes are changed. All the amplifiers I have designed so far were easy to build and straight-forward in use, which to me represents an important feature – thus fixed bias (i.e. bias by negative voltage applied to the grid) was discarded as it brings more potential problems than eventual benefits.
Lowering the output impedance and increasing the current draw has enabled lowering the B+ voltage which mostly compensates for the only possible gain from fixed bias – as a result, DC RMS voltages are never higher than 470V (with 5U4 rectifiers) which helps keep costs down while the second cap in the power supply remains safely in the 500V WKG category (of course, with both channels operating, i.e. drawing current).
Finally, the cathode follower has remained in the schematics, since it obviously helps the driver tube perform its task, and keeps distortions at even lower levels. While the driver tube should be an ECC81 family member (12AT7, ECC801, 6201, CV4024, etc.), the cathode follower can be almost any small signal triode tube – almost, because once you choose the socket and the basing connections, there is little variation possible – 5687 is one example - and of course you have to keep in mind the operating conditions for this tube, which preclude some choices (most ECC88 tubes cannot be used, and even E88CC types are not advisable). But if we stick to the more common noval tubes, it could be an ECC82/12AU7, ECC81/12AT7, even an ECC83/12AX7 – while in the octal domain, a 6SN7 would do the job perfectly well (12SN7 would suit just as well, but you will have to provide 12.6V for the heaters, and connect the heaters of the driver tube in series mode instead of parallel). Whichever tube you choose, as the cathode follower section is directly coupled to the driver tube it will be biased to approximately the same conditions. What will vary is the bias voltage of the CF tube (grid to cathode differential) as well as the input impedance (approximately ranging from 2meg to 5meg) and the gain (in CF slightly lower than 1, depending on the mu of the tube used). It is thus not unimportant which tube is chosen or used – as it will have an impact on the sound of the amplifier.
The cathode follower section requires referencing the heaters supply for the driver tubes at approximately 20% B+ voltage. Most datasheets recommend not exceeding the heater-cathode differential of tubes by more of 100V DC, thus it would not be a good idea to have the cathode of the CF tube at approximately 160V DC without doing something about the potential of the heaters. The best option is to reference the heaters to a potential between the cathodes of the two driver tubes, and a 220k/47k voltage divider will perform this task perfectly well.
Of course, having two triodes per channel helps employ one double triode per channel in a mono-block amplifier configuration… there is no reason why you could not use the ECC81 as the CF tube as well, and as a matter of fact it provides the highest input impedance among all the proposed tubes.

The output tube, as you already know, is the 300B. NOS 300B tubes are probably extinct by now, or too expensive, and I cannot expect that DIY-ers – people who are not ready (willing, or able, choose your game) to spend 5 figures in USD or EUR on Hi-End amplifiers (Hi-Cost no doubt, the Hi-End sound is in the ear of the beholder) – would spend 4 figures on the output tubes alone. Thus the amplifier was designed, bread-boarded, and finalized keeping in mind the current production 300B tubes.
The biasing of the output tube is automatic as customary in RH amplifiers, set by a current source (or sink, if you like) made by a simple LM317 and current setting resistor. Regardless of the tube inserted in the socket, the current draw will always remain the same, and since this is a triode, there is no second grid to take into account – all current drawn is “anode current”. The current is set in such manner as to be near the upper limit for 300B tubes, while the voltage drop across the tube, deriving from the combination of current draw and B+ voltage, puts the anode dissipation between 30.5W and 33W, depending on the chosen rectifier tube (5R4 or 5U4 types). The anode dissipation is on the upper conservative level, which basically means that you will not unnecessarily shorten tube life while enjoying the best sound those tubes can provide. This relatively low anode dissipation is certainly adequate for the mesh anode types, either “mesh anode” or “plate with holes”. Those seeking additional thrills can always use a GZ34 (5AR4) rectifier for higher B+ voltages resulting in higher anode dissipation (about 36W) – this is feasible but not at all necessary.
As for heat and tube life, I have measured the temperature on the hottest point on the glass of the globe “mesh” 300B tubes after several hours of operation at approximately 33W dissipation, and it does not exceed 100⁰C. Real datasheets for current production 300B varieties do not present the same richness of data as the datasheets from the golden age of tubes, but I guess that an output tube in an SE amplifier, operating at temperatures lower than that of the rectifier tube in the same amplifier, where the rectifier tube is definitely running inside the safe envelope of operation, means that the output tube is not stressed and will not encounter a premature end – at least due to anode dissipation – regardless of what some tube dealers would like you to believe. On the other hand, filament defects (breakage) are definitely not caused by the ability of the anode to disperse heat, and should be addressed in a different manner.
Last but not least, the output tube cathode circuitry is virtually the same as the configuration already explained for the RH307A amplifier. Since the LM317 can only handle up to 35V input, the inserted resistor keeps the environment safe for the current regulator. Regardless of 300B type, or rectifier used, the voltage across the regulator cannot exceed 30V. Another solution would be the TL783, which is good for up to 125V input – but besides being relatively difficult to source, it would have to dissipate up to 7W, requiring a very serious heat-sink (temperatures inside the amplifier should be taken into account as well). The voltage dropping resistor should be powerful enough to withstand up to 5W dissipation for a long period of time (11W types and higher recommended).

The heating circuitry is AC, but 50Hz hum is absolutely low and completely inaudible on my speakers, due to the cathode circuitry solution and the general schematics of the amplifier. Even on more efficient speakers the AC heaters hum should not be prominent enough to annoy the listener. While the question which heaters power solution provides the best sound – AC, DC, unregulated, voltage regulated, or current regulated – it is my opinion the AC probably sounds best due to the fact that the whole filament is kept at the same potential level. But leaving this discussion aside, since each and every amplifier should be regarded as a whole, another great advantage of AC powered heaters is the simplicity of the solution. If the inevitable mains frequency hum is not an issue (in this amplifier it should definitely not be a problem), AC is the best, cheapest, and most practicable solution by far.

The Power Supply

As already mentioned, the power supply is simplified in respect to previous RH amplifiers. This is a normal cap input power supply, where the rectifier is combined with solid state diodes in a hybrid bridge. The hybrid bridge is more efficient than the standard full wave with central tap rectifier, and puts less stress on the rectifier tube. It is also by far less complicated to set and use than an all tube bridge rectifier, which would require more tubes and additional secondary windings. The sound (contribution to the sound of the amplifier) of the hybrid bridge is virtually identical to the sound of the same rectifier tube used in a classic central tap arrangement, providing a win-win combination. Rectifiers can be switched easily, either to tune the sound of the amplifier, or to increase (or decrease) anode dissipation. Since output tube biasing is automatic and always the same (set by CCS), there is no need to adjust anything when changing either output tubes or the rectifier tube.
The power supply is cap input, and this first cap should be a 600V poly or oil type – a good choice would be motor run caps rated at least 400V AC. The second cap should be a 500V type, and it can be a good quality electrolytic. There is no need to bypass any of the two caps with smaller caps. The choke is not critical as the only important element is DC current handling – a 200mA type is the minimum requirement. As for inductance, 5H or more is recommended for a good level of filtering.

The Transformers

The RH300B can be built with just one power transformer, or with several, depending on the size of the enclosure and the willingness of the DIY-er to order custom wound transformers. I have built mine employing a 150VA custom wound toroid for the high tension and rectifier heaters, and another smaller custom wound 30VA toroid for the heaters of the output tubes and the drivers. The two cannot be squeezed in a 150VA core, but could be wound on a 200VA core which would probably be a more expensive solution due to the large number of secondary windings. This even larger toroid would also require more space, both as sheer physical size, and probably due to the more complicated placing of the secondary windings output wiring. There is absolutely no need for a more powerful high tension secondary than provided by the schematics: while the DC current draw of the B+ is slightly above 210mA, the AC current draw as measured on the HT secondary is 270mA – less than foreseen by PSUD2 or calculated using the usual formulas. The temperature of the transformers in operation is up to 60⁰C and 40⁰C, respectively, in the box, after several hours of operation – thus it can be regarded as absolutely recommended.

While the schematic requires 5V secondary windings for the output tubes, it is advisable to adopt 6.3V secondary windings and drop the voltage to 5V across 0.56 ohm resistors on each leg. This will limit the current drawn by the cold filaments at power-up and definitely increase the life of the tubes, since filament failure through breakage at power-up is a common problem with many current production direct heated tubes.
A word about the output transformers – as shown in the schematics, those should be 2.5k primary impedance types. The secondary, of course, should be adequate to the nominal impedance of the speakers used. As for size, while size does matter, manufacturers tend to rate their amplifiers as output power, which is a relative point of view. You could build a 2A3 amp and a 300B amp using the same output transformer, as both tubes would work into 2.5k loads – but while a classic SE 2A3 amp will provide approximately 3W, the 300B will provide approximately 8.5W, thanks to its higher anode dissipation, and in particular due to the higher current draw.
Thus the most important part when choosing output transformers for the RH300B should be whether they are built to operate with 100mA DC current across the primary winding. If that request is met, all the remaining details (physical size, DC resistance of the primary, etc.) will work towards achieving a more or less extended and defined output, introducing additional distortion and loss in a more or less pronounced manner. While the RH300B thanks to the circuitry will have an excellent damping factor (DF), and will sound perfectly well on most output transformers that are fit for the current, primary impedance, and output power – since this is really a top-notch project, you should try not to spare on the output transformer. What is true of the RH84, re-built by many with high quality parts and output transformers to enjoy its qualities even more, relates even more to this amplifier, as the parts quality will be rewarded with increased sound quality.

End Results and Remarks

This long awaited amplifier, while being the first to be designed in the 2nd generation series, was the last to be finalized and built. The RH amps started as an alternative to classic DHT SE amps, succeeding in offering better sound and at least similar output power, step by step, DHT by DHT. The RH Universal easily over-powers classic 300B amps while providing better bass and definition, while the RH307A offers approximately the same power of classic SE 300B amps from a less expensive tube (while providing better bass and definition), and adds the thrill of direct heating, the particular sound of direct heated tubes… What is to be expected of the RH300B, since this is the tube mostly addressed by previous RH amplifiers with pentode (and beam tetrode) output tubes?
All RH amplifiers sound quite similar – since they have been designed by the same person adopting the same scheme type, which is quite rigorous on the tubes, imparting a particular type of sound. The main difference, beside output power, is the output tube itself, as the intrinsic sound quality of the tube will define the overall sonic quality of the amplifier. Direct heated tubes definitely have some interesting edge when it comes to sound, and the 300B is probably the most coveted of all DHTs – due to its intrinsic sound qualities, and the relatively high power it can provide, which is necessary in our world of relatively inefficient speakers spawned by several decades of powerful transistor amplifiers rule on the commercial Hi-Fi market.

Well, this amplifier is virtually a nail in the coffin of classic SE DHT amps. Whatever the 300B tube used (since there is currently a relative multitude of current production 300B types, and of course the almost extinct NOS WE) would provide in a classic SE design, it will do better and provide more in the RH300B amplifier. This is not any more the case of comparing pentodes with the 300B – but using the same “weaponry”. As a matter of fact, this was an easy win.

The simulation clearly shows 12W output at 1% distortion, with almost 2V RMS input sensitivity. I do not have any equipment that I could use to provide measurements, so DIY-ers will have to believe my word – or the simulation presented. Nevertheless, what matters is the sound, and the relative loudness that the amplifier can achieve in a system. In practice, the RH300B goes just as loud as the RH Universal does – on my normal real-world 88dB/W/m speakers – quite loud by anyone’s standards. There is no hint of the usual syrupy sound of SE 300B amps, which is so good for taming wild high efficiency horn speakers… instead, the sound is extremely detailed and controlled, with great bass definition and depth. Such statements are not to be taken lightly, but those who have successfully built any of the previous RH amplifiers probably know that their expectations will be met and exceeded.
If you ask me, my only regret is the fact that the 300B, being a triode, does not have a second grid that I could use to my advantage, like a lever to increase performance even further… It is interesting how many attempts have been made to copy the 300B, including indirect heated prototypes, curve clones, that were made out of pentodes in which the second and third grid were missing, not wound on their supports during manufacturing. All this trouble to construct a good triode, when there is so much to be had from pentodes – only because the knowledge required for that task has been lost or misplaced. Humans sometimes find it easier to replace an object, or resource, than to learn how to use it properly.

What Next?

This amplifier represents in a way the end of the 2nd generation series. The next, 3rd generation of RH amps will strive to go further, including different and maybe unexpected output tubes, and the circuitry solutions necessary to make these tubes work at their best.
Designing these amps takes a lot of time and thought, as I try to see the eventual problems in advance and design including simpler and more effective solutions, never straying from the path of excellence: the sound comes first, since those amplifiers are designed to enable enjoyment when listening to music, not winning awards. Bread-boarding an amplifier requires resources which are neither cheap nor easy to source (particularly in my country). The RH300B exemplifies this more than any of my previous amplifiers, as it was not possible to bread-board and build the finalized amp without the output tubes, and the lack of other resources, like output transformers and chokes, hit me more than ever before: for instance, the output transformers had to be extracted from the RH307A (those are 2.5k into 4/8 ohm transformers, when connected to 8 ohm speakers they present a 5k load to the output tube – the same ones pictured in the RH84 PPE prototype), and I had to find a way to design a good sounding power supply using only one choke in a cap input configuration…

There are no publicity banners, nor a Pay-Pal button on this blog. My efforts have always been totally non-commercial. But without help from DIY-ers who respect my work, I am not going to be able to continue designing and sharing amplifiers. All those who would like to help with output transformers, chokes, tubes, sockets, caps… are welcome to contact me.

Sunday, September 15, 2013

RH-307A v2 "Super"

The RH-307A has undergone further development, with the aim to improve output power and explore the possibilities of the circuit, as well as to improve reliability - achieving the same project quality other RH amps are known for.

The Driver Circuitry

I have initially considered the 15W dissipation rating of the 307A tube a limiting factor, and the 9.1W output foreseen in the datasheet as overly optimistic. Thus the circuit was developed around the 2nd generation RH driver circuitry idea - using the feedback resistor as the only anode resistor. This circuitry inevitably leads to a limitation in power output: graphically represented, the sine wave of the driver tube is in opposite phase with the sine wave of the output tube – and the peaks of the two will inevitably collide at a given power output, unless the difference in potential between driver and output tubes is larger than the combined voltage swings. The latter is virtually impossible to achieve at common operating voltages, and represents a limitation to the output power – which does not necessarily represent an important factor if the output power cannot be much higher anyway due to other limitations, like the anode dissipation of the output tube.
The 307A has actually shown itself as an unexpectedly powerful output tube, which probably due to its direct heated nature largely surpasses the output power of other pentodes with similar maximum anode dissipation. Thus the next step was the introduction of a “classic” RH driver circuitry, albeit of the more “modern” style that can be found in the RH Universal v2. The resistor ratio (anode to Rfb) is slightly skewed, although the purpose of such values is not limited to confusing the critics or those who taught they knew everything that was to be known about the subject of anode to anode feedback… Anyhow, this type of circuitry does not limit the output power by colliding the two sine waves, thus power is only limited by the characteristics of the output tube.


Some 307A (VT225) tubes will exhibit a tendency towards screen grid arcing at voltages near the maximum operating condition as shown in the datasheet. In practice g2 might arc (particularly at power-on) if operated near 300V, and this arcing actually happens towards the nearby g3, since the suppressor grid is connected to a low potential point (the cathode or virtual cathode point).
This does not occur with all 307A tubes, nor does it manifest itself regularly in affected tubes. Most documented cases where this tube was used in amplifiers are related to triode strapped operation, thus arcing has not happened as most tie both g2 and g3 to the anode. In pentode operation, tying the suppressor grid to the screen grid, or the anode – would cause the appearance of the “tetrode kink”, with all the negative consequences, and is therefore out of the question.
A very easy solution to this problem is operating the 307A in pentode mode with lower screen grid voltages – conditions that this “filamentary pentode” was actually intended for as a transmitting tube. I have thus made the choice to set approximately 200V as the g2 operating voltage – a value totally safe from arcing in all conditions on all of the 307A tubes I have tried.
With solid state devices in the circuit, like zener diodes and voltage/current regulators, the arcing which has otherwise not damaged the tubes or other parts of the amplifier, has succeeded in killing both the zener diode and LM317 on the interested channel… as a further reliability measure, and also after reconsidering carefully the datasheet, I have decided to connect the suppressor grid (g3) directly to ground, instead of connecting it to the “virtual cathode point” as it would have been customary and usual. When g3 is grounded, the arc from g2, if it ever happens, will miss the cathode and relevant circuitry (LM317). Also, it should be taken into account that the operation of the 307A can be controlled as well by the suppressor grid, and the potential at which it is set – another good reason to connect it directly to ground (i.e. 0V).

Additional Effects

Operating the screen grid at lower voltage means that the anode will draw less current with the same cathode to control grid voltage differential. If this amplifier had a cathode resistor, the value of the resistor would have to be adjusted. But since the current draw is controlled by a current setting device (LM317 with current setting resistor), the result will be a lower cathode to control grid voltage differential, about 15V instead of the almost 30V of the original version where g2 was operated at 300V.
Needless to say, loosing 15V less in the cathode circuitry means drastically cutting on the dissipation for the LM317, with all the theoretical and practical reliability improvements. This also means having 15V more across the tube (anode to cathode) which leads to an increase of anode dissipation since the current is fixed. While some have reported operating the 307A tubes at 22 or 25W dissipation, and 80mA current, I have chosen to stick to the values given in the datasheet – 15W maximum dissipation and 60mA maximum current (set to approximately 43mA anode current). Since the 307A/VT225 is not a tube in current production, and the stock is going to dwindle in years to come, although I think that life is too short to be squandered with sub-optimal solutions – there is no need to burn your (rare) tubes too quickly.
Since the tube is forced to conduct a set current, this means that the characteristics change – a lower input signal will lead to the same output power – the sensitivity of the amplifier increases almost twofold. The effect of this change is obviously audible – the amplifier is more dynamic sounding than the original version: the percussion attacks are more pronounced, it seems as if the amplifier has gained speed. If “syrupy” is how many would define classic 2A3 and particularly 300B SE amps, this is quite the opposite.
The change in screen grid operating point is achieved by increasing the value of the zener diodes. While it could be easily achieved using one 150V zener diode, two 75V 5W zeners are a far better choice. It goes without saying that the power of the zeners needs to be quite high, since their power rating is greatly derated with temperature (and tube amps tend to be warm or even hot). Still, a 150V 5W zener should do, but its dynamic resistance is much higher than the dynamic resistance of two 75V 5W zeners in series – this dynamic resistance is obviosly very important to achieve the particular sound in amplifiers where zeners are used to drop voltage and set the g2 operating point.
The zeners used to drop voltage to screen grids (instead of resistors) seem to be susceptible to performance degradation, and the result is a relatively fast decay in bass (low frequency) performance. This is a topic which, unfortunately, is totally un-documented elsewhere on the net! The results of zener performance degradation can be easily experienced experimentally by connecting a grid stopper resistor between g2 and zener diode – the bass will be filtered and bass levels lowered, as if some RC filter was introduced. Removing the resistor restores low frequency extension, and the same effect can be experienced when a degraded (but otherwise normal in operation) zener diode is changed for a new zener diode of the same type.
Of course, one way to avoid this issue would be supplying the screen grids from a regulated source, or even an additional power supply… but besides being large and complicated, this solution would also miss one important issue – the specific sound achieved when screen grids are connected to the B+ via zener diodes. Thus the “free lunch” solution would be using zener diodes of much higher power, possibly achieving the desired voltage drop by putting them in series, and, last but not least, keeping them as cool as possible by physically separating the zener diodes from sources of heat – for instance, not soldering the diodes directly to the g2 pin…
The series of two 75V 5W zener diodes, placed separately and not soldered directly to the sockets or to the anode resistors, definitely solves this issue with the least of cost and complication – while keeping the particular sound character.
The classical driver circuitry removing swing limitations allows to obtain full output power, but also more freedom in the choice of driver tubes. In this case, it meant getting back to the ECC81 family of tubes, and in particular to the 6201. It is more than obvious that an amplifier will work as foreseen by the simulation or mathematical calculations based on the schematics – but there is more to sound than simulation or mathematics. While circuit simulation with good models allows setting the best values for resistors and estimating frequency response, power output, and distortions – the quality and intrinsic characteristics of the tubes used will have an important influence on the sound, which cannot be simulated. Just like the 6AU6 pentodes are no match sonically for the 6201, or the E180CC, the E88CC is also not playing in the same league. While my choice of 12AU6 was relatively limited (although RCA black anode always means high quality in my dictionary), I had a lot of various ECC88 family member to play with… even the famed CCa does not come close to the sonic performance of the 6201 as a driver in RH amplifiers. Getting back to the ECC81 family, and in particular to the 6201 – is like the “return of the King”.


Well, after solving the reliability issues, improving sensitivity and speed, and changing to a preferred driver, I became aware that the small output transformers are probably a limiting factor, since the amplifier is already capable of sound volumes way higher than the RH84 (the SE version). Thus the small output transformers capable of maybe 5W output have been swapped for larger units (E108 size lamination).
Larger than necessary output transformers are regarded as a trade-off, most are afraid of bandwidth losses (high frequency loss due to higher parasitic capacities). Not in this case… the output transformers used are 5k into 8 ohms, with primaries foreseen for 100mA DC current: if output transformers could be estimated by power, these would probably be rated around 25W. While this looks as a total overkill, the results are awesome with this amplifier: not limited by the small output transformers, the bandwidth extension is nothing short of astonishing, and output power is almost at RH-Universal v2 levels. This should come as no surprise, since the WE datasheet states 9.1W output power at 300V across the tube into a 4.5k load. With 350V across the tube, anode to anode feedback loop, approximately 43mA anode current draw… if the datasheet is of any relevance, no wonder there are about 9W of undistorted output into a 5k load. Thus the “Super” in the name of the amplifier – if built with adequately sized transformers, it will undoubtedly outperform classical 300B SE amplifiers both in sound and power output.

Monday, August 5, 2013

RH84 PPE – "Parallel Pentode Edition"

The RH84 seems to be my most famous amplifier – and probably the favorite EL84 DIY amplifier. Recent builds of the RH84 revision 2 confirm the good sound that many DIY audiophiles have enjoyed so far. Thus, the RH84 remains a good vehicle to further new ideas and improvements.
Critics have mostly criticized the choice of the driver tube – ECC81 – claiming that the application of a pentode driver tube would yield more power and/or less distortion. Still, to most using a pentode as driver actually means using it in triode mode (g2 and g3 strapped to the anode) – and while this is a valid if meaningless approach, it should not be mistaken for a pentode used in pentode mode as driver, which is what the critics claimed as superior, but have never shown in practice, either as one of the numerous copycats, or as an RH inspired improvement.

The Pentode Edition in the name of the amplifier stands for “all pentode”. The driver chosen for this amplifier is the 6AU6 (actually, 12AU6 – since I have got no 6AU6s, and the difference is confined to the heater voltage). At 6.2mA current draw it’s operating point is just where most would put it (if they knew how)…

For my taste, or more precisely, speakers – the RH84 could be improved by having more power. 5W is barely enough for 88dB/W/m speakers, if you like listening to music at moderate and high levels: this is obviously not an issue for fortunate owners of the various large Klipsch, Lowther, Tannoy, Altec, or JBL speakers, with efficiencies higher than 94dB/W/m. While many of those who built the RH84 over the years own such speakers, those that have lower efficiency speakers have stuck to this amp due to its good sound… and eventually gone further in the search of more power. Unfortunately, the power output is a limitation of the output tube – with 12W dissipation, an indirectly heated pentode cannot yield more than 5-6W. Since the EL84 cannot be driven into class A2, no additional power can be had above it physical limits – and there is no driver than can change that – as it should be clear to everyone with enough common sense and some electronics knowledge.

The only way to have more power from the cheap and affordable EL84 is the parallel approach: two EL84 in parallel have a combined anode dissipation of 24W, and output power is comparable to, if not slightly exceeding, the EL34. The parallel approach has been criticized by purists due to the fact that identical tubes do not exist and even matching cannot solve all the issues…

OK, so you cannot have everything from life, thus improvisation and compromise are necessary – this is not a perfect world. With the right design choices and approach, the most important issues related to parallel output tube implementation can be overcome and minimized.

Let’s start with the output tubes: as shown in the schematics, the two devices share the anode connection to the primary of the output transformer, and the zener diode connection to the B+ (22V of drop, slightly more than the drop across the output transformer, keeping the g2 at a constantly lower potential than the anode). The lower half of the tubes is actually where the separation occurs – each tube is controlled by a separate current sink, matched to equal current. The two grid stopper resistors connect to the common grid resistor for both output tubes. If the tubes are perfectly matched, the bias voltage across the current sink will be identical – but eventual mismatches will be automatically addressed as an equalizing voltage across the grid resistor: in practice, with reasonably similar tubes (not matched) this voltage is in the 0.02V range, meaning a grid current of 6.06x10-8 A (sorry if it’s more than the eclectic taste and knowledge of some allows). Even quite mismatched tubes (worn and almost new tube) will not lead to more than 0.4V, i.e. grid current of 1.2121x10-6 A… thus it can be concluded that the two tubes tend to balance without problematic repercussions.

The current sink device chosen for this version of the RH84 is the lowly and cheap 7805. Besides the need to provide some ideas and guidance to the DIYer, this was also done to keep costs lower: the 7805 is 33% cheaper than the LM317 (it that is of any importance) and if 10 pcs from the same batch are acquired, it should be easy and feasible to match accurately for output voltage two pairs of devices. Because, the output voltage is the reference voltage that will define and limit the current draw of each current sink, and thus each EL84. Furthermore, since the output voltage is 5V, approximately half the bias of the EL84 in this design, the resistor will perform half of the dissipation (approximately 250mW) leaving the remaining 1/4W to the 7805 which can handle up to 600mW without heatsink. If that was not enough, the screw hole is connected to pin 2 or ground, and thus is at ground level – meaning that it can be simply bolted onto the metallic chassis of the amplifier (if it is cold enough and not heated by the transformers and tubes…). The current setting resistor can be a 1/2W unit, and I have used 0.6W metal film precision resistors for this position.
Each EL84 cathode is separately bypassed to ground by means of approximately 100uF valued capacitors – I have chosen to place in parallel two 47uF/22V ROE tantalum caps, but any cap type above the 22V rating would do.

The driver in the RH84 PPE obviously is a pentode, used as a pentode (in pentode connection). While small signal pentode tubes can always be used as triodes by strapping g2 and g3 (if not already connected to the cathode inside the tube) to the anode – I see very little reason to pursue that direction, since there are many types of triodes readily available with a wide range of gain and current draw characteristics, thus an adequate driver can always be found without resorting to triode strapping pentode tubes. On the other hand, using a small signal pentode in pentode mode can bring some advantages over triodes, mostly in the available gain “department”… of course, using a pentode as driver in the RH amps (or any other similarly conceived amplifier which does not operate in class A2, or with such output tubes that are not suitable for class A2 operation, provided the pentode used as driver could drive the grid of the output tube with the current necessary… to be precise) will improve neither power nor distortion: it’s used to show that it can be done, and how it should be done properly, with excellent sonic results.

Small signal pentodes are used in a slightly different manner than output pentodes – the g2 resistor in small signal pentodes has a very important function in setting both current draw and gain, and the screen grid should be adequately bypassed to ground, since the combination of g2 resistor and cap forms an RC filter: in this case, 10uF to 20uF caps should suffice. Another detail to keep in mind is the voltage value of this bypass cap – while in operation the g2 will be below anode level and roughly at ½B+, at power-up the grid will not draw any relevant current and the cap should have the same rating as all the other power supply caps!

The 6AU6 was chosen as driver tube due to the fact that it is not some exotic hard to find tube, it should be relatively cheap – and because I had some 12AU6, mostly CEI (gray anode) and RCA (black anode), at hand. The 7 pin sockets are similar in looks to the common noval sockets and are easily available. I am not aware whether there are any xAU6 tubes from current production, but the NOS stock of those tubes does not seem to be dwindling yet.

Another small signal pentode I would use for the task is the EF86 – if I am correct, those can be sourced from current production as well. They require standard noval sockets and are therefore even more easily applicable to the driver task. I guess prices are slightly higher… anyway, I did not use it because I have got none, but here’s a simulated schematics with resistor values for those who would like to try the EF86 in pentode mode as RH84 drivers.

What about the old faithful, ECC81? Of course, it can be used as driver for the parallel RH84 as well, and the results of the simulation are, of course, quite in line with the results of the 6AU6 and EF86 versions – attached is a resistor configuration similar to the RH Universal which is actually optimized for use with ECC81. As a design issue, the ECC81 is a double triode, thus requires only one socket for two units (and half the space in the amplifier, not to mention less heater wiring). Furthermore, there are no additional g2 resistors and bypass caps – less parts equates not only to less cost, but more importantly less complexity and potential problems.

When it comes to sound, I guess we are on very subjective ground. Ultimately, higher quality tubes, drivers in particular, will lead to better sound – that is beyond subjectivity. The PPE is designed as to sound just as good with a garden variety xAU6 as the RH84 sounds with a garden variety ECC81 – just like the EL84 in the RH84 amps sounds in pentode mode just as good as in triode mode... But the power part of this well-known sentence does not hold water here – while in pentode mode the RH84 has (much) more power than in triode mode, the pentode driver does not add to the power, nor improve on the distortion values. On the other hand, there are several great ECC81 family members out there (6201, ECC801, ECC801S, etc.) that are quite tough to beat (in my book, and in the books of many others). Frankly, the RCA black anode 12AU6 are hardly a match for the Philips SQ 6201 – for my taste. As I already stated, this is subjective opinion terrain, and everyone should attempt to find his own best combination of tubes. To someone owning a box full of black anode RCA 6AU6 the acquisition of some Philips or Valvo 6201 (particularly at current prices, which inevitably increase in time), probably represents a nonsense – unless they are absolutely necessary.

It is quite obvious that the RH84-PPE lacks “universality”. You can use the xAU6 chosen, the output tubes are EL84, and the power supply as drawn requires a GZ34/5AR4 rectifier. But if your output transformers have 8 and 4 ohm taps, you can always play the game of “power vs finesse”! Provided your loudspeakers are a nominal 8 ohm load, if connected to the 8 ohm tap the anode load will be 3k, just as foreseen in the schematics. But if you connect them to the 4 ohm tap, the anode load will be 6k, just what is needed for a classic RH84. The schematics foresees the possibility to operate the amplifier with just one EL84 per side into 6k – with half the power: just unplug one EL84 per side before powering up, and you can check whether two tubes in parallel necessarily lead to a loss of focus, as some like to define it. This is basically the same game as “triode vs pentode RH mode” some used to play at the time the original RH84 was introduced. The output transformers will probably be quite large (3k at 100mA DC current across the primary) and thus represent an overkill for a single EL84 – if nothing else, power bandwidth should be at its best, and expected power output is just above 5W per channel. You could do it as well to consume less energy – up to 50W per hour less, as your contribution to a greener planet.

Operating the amplifier with just one EL84 per side will decrease the current draw for approximately 100mA, thus the B+ will be too high at approximately 360V. Therefore a change of rectifier will be necessary, and a 5R4 will provide the correct B+ value. Since the current draw will not exceed 125mA with one EL84 per side, a 5Y3GT can be used as well.

Last but not least, a few words about the power supply. It is kept deliberately simple to keep costs as low as possible, thus a simple C-L-C filter is applied. The first cap should be oil for the best results, while the second cap can be electrolytic. Any choke capable of approximately 200mA DC would do, preferably a 10H unit. Since the power supply is cap first, the choke will not be particularly stressed, and does not need to be particularly large or stiff (potted etc.).