The Valve Side of Canberra
" ‘ere, Bob, come see what these buggers ‘ave done to yer organ".
My friend Michael and I stood awkwardly near the little shed that served as the sales office for recycled goods at the Canberra tip. This was a fascinating place; not what you expect at all in the National Capital. Mike clutched the valve power amp and I lugged the huge power supply we had literally just unearthed. The boss of this recycling enterprise looked forbidding, so we did our best to look forlorn. I shuffled my feet and stared distractedly at the dust. Doubtless he and worker Bob harboured high hopes of selling the huge old keyboard intact at a ridiculous profit. This was privatisation in full flight, and in just the right town, too.
"Not good, fellers, not good." I quietly explained that someone had got in first, and all four KT88’s and pocketfuls of 12AX7’s had already been pilfered before we arrived. We were on the scene long after the horse had bolted.
"Awright", pause, "$20". Very final. No argument on this one.
I cried "Done!" much too loudly and enthusiastically, letting the cat out of the bag. We scampered away, as fast as our load would permit. I spent the afternoon carefully cleaning out Mozzarella cheese from part of the amplifier chassis. At least, I hoped it was Mozzarella cheese. The hot Canberra sun had not helped.
Days later, back home in Melbourne and reflective after this quiet detour from a work trip, I managed to clean the amp and power supply thoroughly and get them working again. The power supply was enormous, with strangely long, thin transformers and chokes, and weighed about 40Kgs! I removed a lot of components – it had included valve rectification and a substantial regulator. I was left with two 6.3V, centre tapped, 10A filament supplies, and a 500~600V B+ supply. I changed the HT supply to 1N4007 diodes and added two pre-loved 330m F/400V electro’s in series, with equalising resistors strapped across them. A large choke and a non-polarised, brick-sized 10m F/600V capacitor, probably an oiler, now comprises the main 500V B+ supply. I was on the way to building a hifi valve amplifier...
I checked my RCA valve data book and the amplifier’s output stage and couldn’t believe my luck; pin for pin compatible with EL34’s as outputs! Only later did I find out that several well known output valves, KT66, KT88 and others, shared the same octal pin arrangement – the original valves could have been almost anything of that vintage. These standards, compared to the profusion of proprietary solutions we find today, reminded me that this was a very mature technology in the fifties.
I used two pairs of EL34’s from my used collection, inherited a year or so back from Mike. He had bought them in an auction pallet from a guitar amp repairer. They were all the left-overs when output pairs or quads were replaced collectively, but most were not faulty or spent. Earlier I had borrowed a big old AVO valve tester from Hugh Dean, figured out how to drive it, and had two matched quads and several matched pairs. Hugh loaned me this beautiful old instrument with merely a quick finger point at one of many sockets and the words; "Don’t touch that one there – it’s at several hundred volts and it’s bitey".
The small signal valves were obviously 12AX7’s – the organ was replete with them, and I had been able to acquire quite a number from the Canberra tip. So the amp ran. Output was a little over 30W/8W per channel at clipping, from an ultra-linear push-pull output stage. The fixed output stage cathode bias ran the EL34’s at about 80mA. This is very hefty plate current for an EL34, perhaps indicating that the original outputs were a different valve after all, so I changed all four cathode resistors to back each valve off to about 38mA or so. I was mightily pleased with myself at this point. Ready to crow, I thought I’d check frequency response with the signal generator as input, having grafted on some RCA input sockets, a volume control pot and robust speaker terminals.
The low –3dB point was lower than the 20Hz my little signal generator could drive. You beauty! I was on a roll! Upper –3dB point . . . O ‘ell! I measured it again. The upper frequency roll-off point was just 5kHz!! I was stunned. Measuring it repeatedly failed to budge it at all. I promptly did what I always do at the first sign of adversity. I sulked.
I poured out my horrific discovery to Hugh a week or so later, and his response gave me new determination. Within minutes we had a plan; 6SL7 voltage amplifier into a 6SN7 concertina phase splitter, capacitor coupled to the EL34’s, "strapped" as triodes. I was actually beginning to think I knew enough to hook up these stages myself, but Hugh had a couple of schematics in front of me before I could say "Radiotron Designers Handbook, Fritz Langford-Smith!"
The 6SN7 and 6SL7 are sonically very well respected octal, dual triodes, and one of the few valves I had heard about. On a previous trip I bought a few of each from David Crittle at Retrovox in Wagga Wagga. The 6SL7 is a high mu, low current beast for voltage amplification duties in amp front ends; the 6SN7 is medium mu, medium current, high voltage valve, often used as a driver. It turns out there is a well established design convention to follow one with the other in audio amplifiers.
These are sought after valves, which inevitably means they are not cheap, and I suffered much gnashing of teeth shelling out several dollars apiece. I should say though that David is amongst the most honourable and fair fellows I’ve ever met. His valve dealership is very successful, and I drove away with several "bargains" which continue to delight. This is fortunate, since for twenty years now I have called my audio hobby ‘Poverty Electronics’.
There are many phase splitter topologies. The more commonly used phase splitter is the long-tailed pair, since it has gain. The concertina uses just one triode, with outputs taken via capacitors from the anode and the cathode. The single triode cannot pull "up" quite as well as it pulls "down", so while simple and linear normally, the concertina has a quite severe asymmetrical characteristic at overload. However, this asymmetry generates even rather than odd-order harmonics, which are not sonically objectionable. Properly biased, the concertina phase splitter has superb AC balance and good linearity, overloading to sonically-pleasing even-order harmonics smoothly when overdriven. This makes the topology ideal for a hifi amplifier, particularly the high voltage 6SN7 which has a very broad sweet spot. After listening to my new amplifier critically for some weeks now, it is not hard to see why hifi manufacturers like Cary and Wavelength favour the 6SL7 and the 6SN7.
‘Strapping’ a pentode for triode operation improves linearity and sonics, but substantially reduces mu, transconductance and plate efficiency. It is done by joining the screen or "g2" grid to the anode (sometimes a small resistor is used here), and with the suppressor or g3 grid tied, as it usually is anyway, to the cathode, five terminals become just three.
I leave behind a solid-state, PA-oriented background, where anything less than 200W is high camp. ("I’m in Artillery; don’t care what you play, jus’ play it LOUD!"). So to see my 30W push-pull amp reduced to 10 puling watts per channel was heart wrenching. That is, until I listened to it. Absolutely magical. What on earth have I been doing all these years?
In the first run-up I measured the 3dB point out to about 7~ 8kHz; almost half as much again as the original, but still not acceptable by a long chalk. I knew now that the output transformers were the culprit; the original late 1950’s organ application just hadn’t needed top-end response, and so in the interests of economy it had not been designed for it. It is even possible the output transformers did not use sectioned windings. I had seen an EL34-based amp on Steve Bench’s huge web-site which showed quite different strapping for the EL34 screens, so I copied his approach and was rewarded with much cleaner test waveforms.
I also used Steve’s idea for treble boosting by bypassing the cathode resistor of the 6SL7 voltage amp right back at the start of the amp with a tiny, nF-rated capacitor. The AC voltage gain of the valve, amongst other things, is related to the ratio of the plate to the cathode resistor. Bypassing the cathode resistor with a tiny capacitor reduces the AC impedance at high frequency, increasing this ratio, and hence the gain, giving a worthwhile boost to the top end. Now I had frequency response past 10~11kHz, but I was still gloomy about extension to 20KHz.
Hugh called over one Sunday morning and we launched a coordinated attack on the amp. We raised the phase splitter’s quiescent current, to drive it harder. I learned that triodes sound best when their grids are shaken, not stirred. We realised that only interstage and global feedback would permit us to reach the 20KHz upper response point, and felt the 6SL7 alone could not provide the gain necessary to throw away in exchange, particularly with the EL34s wired as triodes. So I added a pre-amplifying stage, using a spare 12AX7. We were then able to throw away some of the additional gain by applying interstage feedback from the output of the 6SL7 (now the second voltage amplifying stage) back to the grid of the 12AX7, now the first amplifying stage). That sure helped. Last of all we added just a few dB of global feedback, from the secondary of the output transformer, right back to the start. Now I had a –3dB point at 22kHz. Phew. At last!
I now had a beautiful sounding amp, and listening to it was now a sheer pleasure. Hugh wants me to run the phase splitter even harder, and as soon as I score the particular resistors needed I will continue the evolution of the circuit. Using a ‘plebeian’ 12AX7 has probably compromised the full potential of the amp slightly, and maybe one day I will follow Hugh’s advice and replace it with a second 6SL7.
I have spent somewhere around $50. This project has been a huge amount of fun, and I have learned a great deal. Perhaps I should write to Bob at the Canberra tip to let him know that his organ is reborn and has found a loving home. I’d have to be careful how I phrased that though...!