After the success of the 813 PP amplifier, I needed to future proof the design and that meant changing the 813s as they had become expensive and hard to get. The GM70 had been an option from the beginning and now seemed the best option.

The journey is documented here.

Amplifier Circuits

The architecture of this amplifier is reminiscent of the theatre/PA amplifiers of the 30s and 40s by Western Electric and Altec. I love the simplicity of these circuits bought up to date with modern iron. It is heavily influenced by Lynn Olson’s Karna.

Starting with the power supply, there is a seperate supply for the high and low level stages. This proved to be a major improvement over the decoupled version off the main HT. There is also a lot less heat as the dropping resistors were dissipating about 40W. Each are choke input for best regulation and smooth operation of the rectifiers. I’ve used 6CL3 damper diodes now in place of the previous MV types as there was no future for them being hard to get and expensive. Similarly, I was using mercury for the first stage but had no spares going forward so used a direct heated 4V rectifier, DW4 which could use the same filament transformer. I prefer the sound to the MV rectifiers so suspect I had some RF interference now the chassis was a lot more compact. The supplies are connected to the audio chassis via 1000V rated shrouded 2mm test plugs and sockets. The double switches are not now required as there’s no need to preheat the MVs; the damper diodes and DHT rectifier ensure a slow ramp up of the HT supplies.

The LT supplies for the filaments are a voltage controlled current source (VCCS) design from the Audio-Talk forum which were suitable for the majority of lower current and voltage DHTs. They were modified to run the GM70 at higher voltage and current. I did use SMPS to feed the GM70 but always had doubts about how good they were compared to a linear supply. In the end I used the Coleman design for the basic DC supply to the regulators and am very happy I made the change and very lucky I could fit the new supplies in the existing chassis. Rod Coleman and Guido Tent have similar DHT supplies. The VCCS supplies have all the advantages of DC, quiet, without the disadvantages that many DC supplies have compared to AC.

The bias supply is more complicated than many PP amplifier bias supplies. I think there is an assumption that they don’t need to be that good as it’s low current and the PP stage cancels any noise. As the grid resistance needs to be kept under 30k, the bias supply draws 8mA so not that low in current hence the more sophisticated design. I’ve found the better I made it, the better the sound. I may not have finished in this area yet.

The Lundahl 1676 is a superb performer as a phase splitter at the input. The 2P29L configured as a triode had been reported a good line stage valve. I was apprehensive about adding another stage to provide the needed extra gain over the 2 stage design but that proved to be unfounded and was an improvement all round.

I used the 4P1L based on it’s reputation and I preferred it to my first choice 46 which was a suprise. The 4P1L is quite microphonic but that doesn’t worry me and mounting it on a more solid chassis reduced the effects.

The 4P1L, configured as triodes are transformer coupled to the GM70s. 4P1L is a small transmitter valve and can be used as finals so in effect the GM70s are being driven by a small power amp.

I used fixed bias to again, reduce the heat, have easy adjustment of the DC balance and gauge the effect of bias point on the sound quality. The DC resistance in the GM70 grid should be kept under 30k to ensure stability and prevent the GM70s from runaway. The GM70 has the advantage of no top cap keeping the high voltages inside. It is relatively easy to drive, fantastic value for money and sounds great. Apparently the copper versions are a step up in some areas but come at too great a cost for me.

I used a Sowter output transformer that was in use for the 813 amplifier. It has configurable secondaries so the a-a primary impedance could be 7k or 14k. I configured it for 14k which should be a lower distortion option.

In a rough test using an 8R dummy load, I saw about 21 to 22V out before visible distortion of the sine wave so approximately 55 to 60W out. That required about 2.5V input. The point of this amp was to provide effortless sound with huge headroom which I think it achieves.

Although the layout means the rectifiers are hidden, I need the chassis to be narrow and split in two otherwise it’s too heavy and given the packaging, seperation allows it to be worked on. It has casters so it can be moved easily. Size is about 400mm from lower deck to upper, 430mm deep and 220mm wide.

I modelled the layout in Sketchup as it is relatively compact for an amplifier of this type and the packaging is dense in places. It took ages but was well worth the time as it took the guess work out of the build.

Sketchup views of model

Top and side views, audio chassis and power supply base.

Next move is to make panels to enclose it.