GoodSoundClub - Romy the Cat's Audio Site

In the Forum: Melquiades Amplifier
In the Thread: Single-stage Melquiades vs. DHT amps

Post Subject: Some feedback on designing with the 45 and 2A3 DHTs... might prove interesting.Posted by floobydust on: 1/19/2009

This thread has been pretty interesting overall. I've read most of it and would at least like to provide some feedback on the technical side of designing with the 45 and 2A3 DHTs... at least from my personal experience. A few years back I decided to do a clean-sheet design for a DHT amplifier after building a Shishido 2A3 schematic using a 45 instead (and proper circuit changes). I wasn't overly impressed with it despite using very good quality parts and OPTs, hence the clean sheet design.

I'm old school... so I started with tube specs and working from old published plate curves. After much searching I eventually found Hashimoto transformers and purchased a set for testing... (great choice I've never regretted). I built a breadboard for and open build and easy change and started with the power supply and the output section. I'm lukcy to have a large stash of NOS 45 and 2A3 tubes along with some 6B4G... so I started with the 6B4G. After much testing and measuring (again, just the output stage), several things became clear:

1- AC filaments on the 6B4G are a disaster for output hum. The reason became clear... the 6B4G is essentially a dual-section 2A3 where the filament sections are wired in series and the grids plates are paralleled. This makes it impossible to balance the hum out with an AC filament. A simple full-wave DC supply was added and things quieted down so I could resume design work.

2- Recommended plate load impedances vary quite a bit with operating points. Finding an optimum load and staying within recommended ratings was desirable, so a fair amount of time was spent optimizing the output stage. When I say optimize, it is to ensure the best possible output power with the minimum amount of distortion, symmetrical clipping (i.e., operating point balanced against the transformer's magnetic saturation and field collapse) and be assured that the tube will have adequate life in actual operating conditions.

3- Input capacitance has a much greater affect on the driver. As is common with triodes, the input capacitance is greater as is the required input voltage swing compared to most beam-power tubes (for similar output power) this places a much greater demand on the driver circuit to ensure proper frequency bandwidth and slew rate for a crisp and clean rise/fall time.

4- Fixed versus cathode biasing. More of a personal choice here. In general I prefer cathode biasing for single-ended designs and fixed bias for push-pull designs. For this design I chose cathode bias. More on this later.

For the 6B4G (and 2A3) I found that optimal loading impedance was 3500 ohms. Note: the same testing on the 45 resulted in 5000 ohms. This yields the best overall performance and lowest distortion by a measurable amount. Yes... I use some simple and proven techniques to measure with. A clean sine/square wave generator, dual-trace scope and accurate AC RMS meters and distortion analyzer (all HP and Tek gear). Whlie this is not same as listening, I still believe that each stage of a good design should be able to pass some simple measurements and get them right, otherwise you wind up with a cascading of stages which each adds it's own coloration (i.e., distortion) and it's own short-comings in being capable of driving the next stage.

As the Hashimoto SE OPTs invert polarity I eventually opted for a 3-stage design as I wanted to maintain the phase and AC grounds on the transformer. I spent considerable time on designing the input/driver, again using some previous experience of various dual-triodes I've used over the decades. Initially I started with a 6FQ7 which is (essentially) a 6SN7 in a 9-pin miniature size (many vintage tube manuals refer to the 6SN7 plate curves). While this did work quite well, I wasn't completely happy with the performance and noise levels despite having a large stock of NOS tubes to pick from... I even picked up some additional ones without any significant change.

I later swapped to a 5814A (rugged version of the 12AU7) and again started with the plate curves and re-biased the same topology. Success! Output noise was lower, gain was about the same (140-150 volts/volt) and measured distortion (at the 6B4G grid) was also lower and at the same or lower drive current. Further testing and some slight value changes resulted in very consistent performance with a wide range of 5814A, 12AU7, ECC82 and 5963 tubes of different brands. This is always a design goal to have a circuit which will result in expected performance with unit-to-unit consistency as parts tolerances can skew things out the window. I also swapped many tubes thru the output stage with repeated measurements to prove the same.

After much effort, the design firmed up and I eventually built final versions using the 45 and 2A3. I ditched the 6B4G with it's DC supply and went on to develop a better balancing arrangement to ensure the output stage is quiet with AC filaments. By definition, the 45 can be the quietest as it is always a single triode section with a simple "M" filament design. The 2A3 can be almost as quiet, but ONLY in the dual-section paralleled types. The original single-plate triode has some advantages but also has disadvantages which makes an AC filament impossible. After a while, I devised a simple technique which I've never seen on any design before (and I've seen hundreds or more over the past 40 years). Someone mentioned he had seen it, ws nothing new and he also used it... but he could never point me to anything to confirm... so who knows... in any case, I adopted a split balance technique. It's ability to acheive better hum balance is unquestionabe... it works with all good quality 45 triodes and dual-section 2A3 triodes.

The technique is simple and I would recommend anyone with a 45 or 2A3 AC-operated filament SE amplifier to try it. It only works with cathode bias due to the topology but it is effective. You also need a good quality filament transformer with an accurate center-tap. As usual, the filament transformer drives the filament of the 45/2A3. The center-tap is used for the biasing resistor (ONLY) to ground. This becomes a "fixed DC balance" for the directly-heated cathode. As the gauge of the filament is accurately maintained during manufacture, it is reasonable to believe that maintaining a centered DC balance provides an even DC current thru the filamentary cathode.

The next part is the AC balance which is adjustable. As the filament wire itself is coated with rare earth elements and other compounds to increase it's ability to become an effective cathode for electron emission, the coating is not 100% even. This results in the filament AC voltage being imposed in the electron flow. I used a 100-ohm wirewound potentiometer with the ends in parallel with the filament. The wiper of the balance potentiometer goes to ground thru the cathode bypass capacitor. As this circuit can not pass any DC current, it becomes effective only for AC balancing. I also pad the potentiometer with a pair of 12-ohm resistors, one from each end to the wiper. This reduces the effective resistance of the 100-ohm pot to less than 4 ohms (from the "cathode" to the bypass cap) and increases the wiper range so you don't have a "knife-edge" adjustment on the balance pot.

So, how effective is it?? With good quality 45 triodes, I've had the output noise (at the speaker output) better than 85dB referenced to 1-watt output. For an 8-ohm load, measured output voltage is less than 125 microvolts. As a result, I would not bother to use a DC filament supply on a good quality 45 or dual-section 2A3 tube. I think the AC filament on these tubes is better technically as you come closer to acheiving a similar operation to a unipotential cathode.

The final measured specifications for the 45 amp were:

Rated output: 2 watts RMS clipping at 2.25 watts RMS
Frequency response at 1-watt: 25Hz to 45KHz within 1dB
Power Bandwidth at 2-watts: 50Hz to 45KHz within 1dB
Distortion (THD): 0.35% at 1-watt, 0.75% at 2-watts (slightly higher below 100Hz at higher power, less at higher frequencies)
Signal-to-Noise: 85dB with good tubes... some samples are better 90dB below referenced to 1-watt output

The design uses all AC filaments, no regulators anywhere and zero feedback. So it is possible to have a wide-bandwidth, low distortion and quiet SET without any fancy circuits for filaments and/or plate supplies.

Some information regarding some other DHTs I've used... namely the single-plate 2A3 (original RCA/Cunningham) and the 300B (Western Electric version). Both of these tubes are NOT candidates for AC operated filaments. Both of these tube designs use what is known as a "center-tapped" filament. Yes, there are only two pins for the filament, but electrically there are two halfs to the filament arrangement and they are strapped in parallel. Western Electric will confirm their 300B (old and new manufacture... as it's the same design) uses a center-tapped filament arrangement. If you look closely inside the WE300B, you'll note the filament takes the form of a double "M" arrangement. The bottom center where the two "M"s meet is the center-tap and goes to one of the filament pins. The two far ends of the filament are tied together and goes to the other filament pin. Because you effectively have two filaments operated out of phase within a single grid and plate structure, it's impossible to balance the hum out.

The single plate 2A3 is similar... I've actually seen some where half the filament was gone where the other half (side) lights up. The arrangement is somewhat different as there are 8 runs of the filament wire and a single bar tensioned by two coil springs up top to maintain wire tension. It suffers from the same basic problem as the WE300B tube, but it's actually worse (and easily demonstrated). While the WE300B has a double-"M" filament, each loop is tensioned by a separate "fishing rod" tension wire. The single-plate 2A3 has 8 loops all on a common tension bar. As the filament heats up and each run conducts current, minor variations cause unequal expansion of the wire loops. This results in some wire loops being loose enough to to exited by small mechanical vibrations. With a scope on the amplifier output, a handclap within a few feet can set them off. They will also resonate at certain frequencies as well. I played a couple scales on a trumpet within 6 feet and certain notes drove them to extreme levels of exitation. This could, perhaps, be one reason that many people claim the single-plate version to have a warmer sound or whatever terms you'd like to apply... bottom line, you have mechanically induced distortion to the output stage. And these versions go for major dollars too!

I've not had much experience with the new range of tubes being manufactured by EML, KR, the Chinese or the Russians, etc. but I did buy a new pair of EML-45 solid-plate tubes a couple years ago. They are well made, huge (by contrast to the original 45) and butt-ugly (personal view). Two things that I don't like about them. First, they exhibit the same mechanical vibration as the single-plate 2A3 noted above. Second, they sound dull and lifeless when plugged into the same output circuit as any of my good NOS 45s. A conversation with Roger Modjeski resulted in me increasing the idle current closer to 40ma (from around 34ma). This made a huge improvement sonically but alas they are still easily exited mechanically.

So, sorry for such a long post from someone who's new at Romy's place.... but hopefully some of you find the above useful to some extent. I also have much feedback regarding 45 and 2A3 brands and internal mechanical structures but will save that for another post. Needless to say, I've narrowed my acceptable choices on which 45 and 2A3 tubes I keep and which ones I don't like. If anyone has interest, I'm happy to post those findings as well.

Regards, KM

Rerurn to Romy the Cat's Site