A Push-Pull Amplifier,
featuring the 6CM5/EL36 and the Altronics-M1115 transformer.
Grant Wills, May 2003.
Grant Wills wrote to me in early 2003; "I surfed into your website recently and saw your design for a push-pull valve amplifier using the Altronics M1115 speaker line transformers. The exact same idea had occurred to me several years ago with that particular transformer in mind. Like you, I had many questions, but at least on paper the concept looked viable. I too calculated an 8k:8 Ohm impedance ratio, based on the spec's for the transformer in its intended speaker-output application.
One thought that I had re the unresolved puzzle of the current carrying capacity of the different primary taps, was that it should be possible to confirm that the whole winding uses the same gauge of wire by measuring the DC resistance of the centre-tap to each anode - they should be the same if the wire gauge is uniform".
Myself, I recalled that the winding resistance measurements were broadly as I expected them to be, but I didn't think I'd be able to pick a change in wire gauge so didn't look for it in the way Grant now suggested. It seemed to me that the change in resistance would be too subtle for the type of resistance measurement equipment I certainly had available - i.e. my DVM! The principle of Grant's idea seemed quite right to me though, and with care and better test equipment possibly quite practical.
Four months went by, and I heard nothing more than that single initial contact from Grant. Then in May, completely unheralded, came Grant's amazing amplifier, clearly the product of much research and hard effort.
The following is Grant's circuit explanation.
My new amplifier, based on the M1115 transformers, works a treat! The final amp topology used the front end of the Leak Stereo 20 amp, as I had heaps of twin triode valves in the junk box. The output stage initially used 6L6's with a B+ supply of 300V, as again I had some on hand. But after quite a bit of experimentation with other output valves I used 6CM5's, based on the large amount of information now available on these dime-a-dozen valves. As is well understood, they are an odd valve to attempt to apply to audio applications, and I had lots of unsuccessful attempts to use them.
I thought initially that I'd feed the screen grids with 1/2 B+, as that is how all data sheets suggest you should do (and incidentally how similar line output valves such as 6DQ6 are configured in old guitar amp's). However, any attempt to use them in this way caused fairly violent oscillation at ultrasonic frequencies.
In addition, I wanted to use them in ultralinear configuration and tying the screens at 1/2 supply did not permit this. The ratings indicate a maximum screen supply of +200V and so I hesitated to use the ultralinear mode. However, after trying several configurations which either oscillated, distorted or otherwise misbehaved, I tried them in ultralinear mode with a fairly high amount of bias (-50V), and they worked really well! They were by far the most linear of any valve I tried and worked well with a fairly low standing current (approx 25mA each) and put out the maximum power (17W)! This seemed to justify their rather high heater power requirement. So I don't know if it is okay to run these valves with such high screen voltages, but mine works really well.
Grant Wills lives in Gosford, NSW having moved there from SA 6 years ago.
Starting his working career as a technician at the then SA Institute of Technology, at the time he played guitar in local bands and built PA gear. He then shifted into small business, establishing a domestic electronics service company. Over the years Grant has dabbled with Amateur Television and holds a Limited Amateur Radio license, and has built a number of guitar amp's, PA amp's and mixers.
He also has an interest in home made test equipment with a couple of circuits published in Electronics Australia! Grant commented that his recent return to valve electronics has become an absorbing hobby fuelled by masses of info now available through the internet.
You can email Grant directly, here.
(Max power of 17W was measured by applying a 1kHz sinewave to the amp and measuring the peak to peak voltage across an 8 Ohm dummy load at the point of clipping. The RMS volts = pp volts divided by 2.88 (twice the square root of 2). Power then = Vrms2/8 Ohm. Note that the max power will not be sustained at very low (30Hz) frequencies).
The power supply used an old Philips valve power transformer, which was up to the fairly high heater current load of (4 x 1.25A) + (3 x 0.3A) = 5.9A, and had a B+ winding of 110Vac which, when applied to a voltage doubler provided 300V B+. (The doubler generates twice 110 = 220Vac, which after full-wave rectification comes to roughly 220 x 1.414 which is a tad over 300Vdc).
The schematic of Grant's amplifier.
Click on the image to enlarge it.
I decided to use fixed-bias as it allows slightly higher anode-to-cathode voltage for the output valves than cathode bias. This requires a bias network with a negative voltage DC supply. But the technique you can see in the power supply section of the circuit, for also getting a negative voltage supply from a voltage doubler, is an old trick from guitar amplifier designs.
I added a 10 Ohm resistor in each 6CM5 cathode earth return to monitor cathode current (by Ohms Law the easily measured voltage across this resistor, divided by 10, yields the value of standing cathode current through the valve, this in turn being the sum of principally the anode current plus a lesser amount of screen current) and act as a fuse under overload conditions.
Note that the circuit doesn't have any provision for individual adjustment of output valve standing currents. This was because I used 4 Radiotron 6CM5's from the same batch with very similar characteristics. A couple of other 6CM5's that I tried varied a bit in standing current. It may be helpful to modify the bias supply with 4 x 50kOhm trimpots in parallel and then in series with an 82kOhm resistor to ground across the bias supply in place of the one 100kOhm trimpot. With each trimpot wiper independently feeding each output valve this arrangement could then provide individual bias control over each valve. If the bias voltage needs to be increased, the value of the 0.1µF capacitor feeding the bias rectifier can be increased.
The front end of the circuit is conventional. The Mullard 5-10 amplifier uses the same phase splitter as I have, and the text refers to it as "a cathode-coupled phase splitter". A footnote refers to an article in Wireless World May 1948 by W T Cocking entitled "Push Pull Input Circuits." It's a tried-and-true topology, recognised even in solid-state circuits as the "long-tailed-pair", and works well.
Incidentally, the output transformers seem to be fairly indestructible! A couple of times I inadvertently shorted the anode connections to ground by not realising that 6CM5's have an internal connection between pin 3 (an apparently unattributed pin) and pin 8 (which is cathode and suppressor grid), and I had pin 3 connected to the top cap (the anode). This effectively shorted B+ to ground via only the transformer "primary" side winding, which DC-wise is near zero resistance! However the transformer survived fine with such a gross overload!
Construction was fairly easy as I took a leaf out of many current designs which works well. I simply made up a square plinth of timber which has a 12" x 12" aluminium sheet (from Dick Smith or Jaycar) fixed to the top. Easy to make and it makes a great furniture piece!
I confirmed the frequency response figures were consistent with those experienced with the 6AN8/Altronics-M1115 design - the amplifier has a wide frequency response with sustained bass and sounds really nice - even my wife immediately noticed a dramatic difference to the previous solid-state amp.
The S/N ratio for this amp was outstanding - maybe the C-core power transformer, or the star earthing, or the provision of a centre-tap on the heater winding, or my 100µF overkill filter capacitors on the small signal HT??? The noise and hum are just about inaudible listening directly to the drivers of my quite sensitive speakers - certainly better than other amp's I have used over the years, and pretty amazing for a valve amp!
Overall, a lot of fun to build, and a most satisfactory result!