The completed amplifier.
The HT generator is fixed to the output transformer.
Music has been my companion whilst working/studying since my teenage years. Then a transistor radio receiving pirate radio served. Over the years many different systems have been employed in my home office, including a Quad 33 and 303 feeding a pair of Danish Jamo satellite speakers. For several years a small 8 Watt PWM brick amplifier fed the Jamo speakers on the window sill in front of my desk. Recently a pair of Wharfdale Diamond 9.1 speakers were purchased and the bass has been restored to the study area.
Why the change now? Simply put, my hearing is failing. A recent hearing test revealed just how much loss there is on my right side. Fortunately my daughter has exceptional hearing and she is my 'standard listener'. Also - why not have a valve amplifier.
An experimental mono amplifier based on the QQV03-20A double tetrode has been on trial in the 'Shack' next door for some years. The power is low > 3 Watts but the sound is very good and more than fills the museum office. An 832 has been substituted with an improvement in sound quality and this was a surprise. HT for this experimental amplifier comes from my Heathkit IP-17 stabilised variable HT power supply. This uses a pair of 6L6s as series pass elements and produces a clean DC output. The 12.6 Volt AC supplies the heaters.
The amplifier has a small level of mains hum with this arrangement but still less than that produced by an authentic 3-3 amplifier. Transferring the heaters to a DC supply has eliminated all of the hum.
Space in the museum office is at a premium and this is especially true of the desk area. There is little space for a valve amplifier but just enough if a separate power supply is used. The hum reduction achieved by using DC on the heaters and a fully stabilised HT supply is valuable in an amplifier that drives speakers just over a metre away from the listener's ears. This is much appreciated on quiet passages.
What about driving a valve amplifier from an old laptop power supply? Finding G A French's push-pull EL91 amplifier set me thinking.
A stock of LM2596 based DC-DC converters sourced from ebay has been invaluable for many small projects. For about £1.00 a 2 Amp module can be had.
A 2 Amp step-down DC-DC converter.
When stepping down from 15 Volts to 12.6 Volts and supplying just under 2 Amps the module runs cool. Heater supply for the EL91 amplifier is 700 mA per channel at 6.3 Volts. Also available are non-isolated DC-DC converters originally designed for battery charging and these have presets for both Voltage and current. The the constant limit was set to 900mA to reduce the inrush current when the heaters are cold.
Searching ebay reveals DC-DC supplies for driving nixie tubes for clock projects. These are not ideal but one supplier vfdclock produces a 10W HT module with a high switching frequency (120KHz) for audio applications and can supply 11.4 Watts at 250 Volts from a nominal 12 Volt supply - the 45 mA required for the EL91 amplifier.
The HV10W 10 Watt HT module.
A bonus from this dual DC-DC converter arrangement arrangement is that the load on the main power supply is sensibly constant. Initially the power feeds the cold heaters and the HT line draws no current. As the heaters reach operating temperature the HT starts to flow and the heater requirement is less. At 13.8 Volts the current drain for the stereo pair was 1.8 Amps from the main power supply.
A 2 Amp step-down DC-DC converter with adjustable current.
Fibreglass strip-board was chosen as the build platform and the PCB valve-holders required a small amount of hole enlarging to fit correctly. The earth returns for the amplifier are all made to the input socket to minimise loops. The DC input power return to the HT and LT modules is run on separate tracks to keep the return currents away from the signal tracks.
The DC-DC converters are mounted to the side cheeks of the output transformer. The transformer is a Variable Voltage Transformers (VVT) type VTP12186-1100 and the link is to the data-sheet. This company has now closed due to retirement and an off-line datasheet can be found here. This transformer has ultra-linear taps at 43% and 20% and so the screen grids of the EL91's are connected to the 43% taps and not returned to the HT rail as in the original design.
The completed mono-bloc amplifier.
The design was made-up to the original circuit diagram with the exception of the input to the first grid. The input is via a 1 μF capacitor and the grid resistor has been reduced to 47kΩ given that the input was to be from the computer sound card. The on-board HT capacitors are both 47 μF 450 Volt as these were available and do not stress the HT board.
All connections and track breaks were inspected twice. The heater circuit functioned as expected but the HT rail dropped well below the 250 Volt level set on the HT module. To begin with the main DC input was supplied from the bench power supply and set to 12.5 Volts. When the heaters were at operating temperature, the 12.6 Volt input supply was running at two Amps - much more than expected and the HT rail was under 200 Volts. Everything was re-checked and the voltage drop across the cathode resistor was checked. This drop was 40 Volts indicating some 70 mA of HT being drawn. The culprit was the circuit diagram in the original publication, quite unusual for The Radio Constructor, the output valve grid resistors were shown connected to the cathode and thus no bias was being generated and the valves were turned hard-on. The correct place for the lower end of the grid resistors was to ground. The circuit diagram has been altered to show this correction. With the circuit re-wired the supply was applied once more and the total current drawn from the input 12.6 Volt supply was at just over one amp. The amplifier worked perfectly.
The capacitor across the output transformer primary has been omitted and initial tests show the response essentially flat from 80 Hz to over 20 kHz.
Frequency response at 1.5 Watts output.
After successful testing of the first amplifier a second one was built. This is also on glass fibre strip-board. These boards measure 100 x 160 mm. The common point for all the earth returns is the signal input phono socket as is standard amplifier practice.
The complete stereo pair operating on my desk.
A 4.7 Ω resistor was connected as the output load to the 3-4 Ω output lugs on the transformer. With a sine wave input at 1 kHz the amplifier output was viewed on the oscilloscope. The input signal was increased until slight waveform distortion could be seen. The input signal level was then backed-off. When the display was switched to the spectrum analyser, there was a strong peak at both fundamental and second harmonic. As push-pull is expected to cancel out the second harmonic it was clear that the valves were not operating equally. With four EL91s available some substitutions were made and a pair were found that were well matched as indicated by the second harmonic peak reducing to a small value. The second pair are reasonably matched but will be replaced when a matched pair are found.
Listening trials with a range of music types are very favourable. My wife and daughter both agree that the sound is clear and comfortable on the ear. Lyrics are easily resolved when listening at a distance more easily than with the PWM brick amplifier previously employed.
It is satisfying to report that the sound is perfectly adequate for the area it has to fill even though the amplifier is driving a pair of modern loudspeakers rather than the high efficiency types normally associated with the valve age. I have a pair of 1974 Leak Sandwich 250 speakers that have a maximum input of 18 Watts and the subjective output loudness is roughly the same. The Leak speakers are far too large to sit on the window sill.
I have a stock of output transformers with which to make several more low power stereo systems including single ended 6V6 and ECL86s. The latter transformers come from a pair of Grundig TK14 tape recorders.
Over time I had acquired a pair of Pye Black Box amplifiers but initial testing revealed that both have half the primary open circuit. Both distressing and an expensive reminder of caveat emptor.