A 4GK5 Headphone Amplifier Design.

My original design intention had evolved from considering making a microphone pre-amp, to take advantage of a valve's greater tolerance for overloads, soft compression on overloads and euphonic distortion characteristic, in relation to vocal mic signals in particular, than perhaps might be expected of a solid-state pre-amp. This intent progressively changed and the design first became a stereo-paired voltage amp for line amplifier use, and then with the addition of a second valve to each channel amplifier as a mu-follower, it finally became a stereo headphone amp!  Have a look at the circuit of the mu-follwer and associated B+ and heater power supplies.

This was the first time I'd ever really tried to design something like this, and I was also very new to working with valves. So I took as guide the text from my ratty old copy of an RCA Receiving Tube Model (RC-27, 1970), which contains a relatively easily followed explanation for setting up and using a loadline on the valve's Ia/Va chart. I've quoted the text of this reference below, as I've followed it along. This manual also provided the basic valve data for the 4GK5.

10 new "National" branded 4GK5's were on "special" from Rockby, A$0.30 each; the ten 7 pin miniature sockets were $2 each, from Truscotts.

  1. "Locate the zero-signal bias point P by determining the zero-signal bias Vc0 from the formula;

    Zero-signal bias (Vc0) = -(0.68 x Vb)/mu

    where Vb is the chosen value in Volts of DC anode voltage at which the valve is to be operated, and mu is the amplification factor of the valve. This quantity is shown as negative to indicate that a negative bias is used."

    Set Vb = 135V from typical operational "Characteristics" and mu = Amplification Factor = 78, so;

    Vc0 = -(0.68 x Vb)/mu = -(0.68x135)/78 = -1.2V

    Point P was then plotted graphically, for Vb = 135 and Vc0 = -1.2V on the anode-current versus anode-voltage diagram.

  1. "Locate the value of zero-signal anode current I0, corresponding to point P". I0 was then plotted graphically and determined as 8.3mA.

  2. "Locate the point 2I0, which is twice the value of I0 and corresponds to the value of the maximum-signal anode current Imax."

    Imax = 2 x 8.3 = 16.6mA was then plotted.

The completed headphone amplifier, warts-and-all!

  1. "Locate the point X on the DC bias curve at zero Volts, Vc = 0, corresponding to the value of Imax."

    Point X was plotted and determined to correspond to an anode voltage of 69V.

  2. "Draw a straight line XY through X and P."

    Line XY was plotted. Y was determined for Vc = -2.4V being twice Vc0 = 1.2V and so allowing for equal input voltage swing to Vc = 0V and Vc = -2.4V. The co-ordinates for Y were determined as Vmax = 171V and Imin = 3.4mA.

  3. "The load resistance in ohms is equal to (Vmax - Vmin) divided by (Imax - Imin), where V is in Volts and I is in amperes."

    Load resistance = (Vmax - Vmin)/(Imax - Imin) = (171 - 69)/(0.0166 - 0.0034) = 7.73kOhms. This suggested a preferred value load resistance of 8.2kOhms. Maximum power dissipation in this resistance would be at point X, and correspond to Ir = 16.6mA, giving Pdiss(8k2) = (0.0166)2 x 8200 = 2.26W, suggesting use of a 5W rated resistor.

  4. "In a class A amplifier under zero-signal conditions, the anode dissipation is equal to the power input, i.e., the product of the DC anode voltage V0 and the zero-signal DC anode current Io."

    V0 was assumed to mean previous Vb, although this wasn't entirely clear. So, Vb = 135V, and I0 = 8.3mA, gave Pdiss(anode) = 135 x 0.0083 = 1.12W. This value compared favourably with the rated maximum anode dissipation of 2.5W so the Vc0 bias calculated was deemed acceptable.

  5. It was noted from the RCA text that the calculation of the load resistance line allows for a distortion level of around 5%. Increasing the load resistance appears to lower the distortion level, presumably at the expense of increased output impedance of the amplifier stage.

  6. "The value of the resistance for cathode-biasing a single valve can be determined from the following formula:

    Resistance (Ohms) = (desired grid-bias voltage x 1000)/(rated cathode current in mA)"

    The rated cathode current = 22mA, but what is meant is clearly the cathode current, which is the same as the anode current, at which the valve is to be operated. So R(cathode) = (1.2 x 1000)/8.3 = 144.6 Ohms. The nearest preferred value is 150 Ohms. At the bias point power dissipation will be (0.0083)2x 150 = 10mW, suggesting a standard 1/4W resistor will be more than suitable.

    Leaving the cathode resistor unbypassed will minimise distortion, at the expense of gain and "power sensitivity". It was decided to leave the cathode resistor unbypassed for the cathode degenerative negative feedback effect, therefore. A note elsewhere indicated that hum may be a problem with an unbypassed emitter resistor.

  7. The valve data gives a "Maximum Circuit Value: Grid-Circuit Resistance, for cathode bias operation .. 1MOhm". It would appear therefore that a pull down resistor from the grid to ground is necessary, and that it is desirable to maximise its value (presumably to avoid loading the input to the circuit), but that value mustn't exceed 1MOhm. Here the input to the circuit will be a low impedance microphone, so there seems no point having a high value grid resistance at all. Rg was therefore set to an arbitrary low preferred value, 100kOhm. I assumed a 1/4W resistor rating to be more than adequate, without calculation. A note elsewhere indicated that high grid resistances should be avoided to minimise hum.

  8. The heater supply required was 4Vdc, 300mA. I considered using a 5V, 500mA regulator IC followed by one or two series diodes to reduce the voltage to 4V, as was having a regulated supply at 4V x 4 (the number of 4GK5's in series) = 16V. I then decided to dispense with the regulation since the load was basically fixed and constant. A separate transformer to that used for the main supply, with a single 14Vac secondary, gave about 19Vdc after a bridge rectifier and 2200uF/25V electro. A (19 - 16)/0.3 = 10 Ohm, 3W resistor gave almost exactly 16V to put across the 4 heaters in series. A second electro was added after the resistor to give AC decoupling.

    The heater supply was initially completely floating relative to the main supply - the two "grounds" were not tied together. The cascode arrangement would have the voltage amp valve cathodes near 0V, the current amp, in the anode circuit of the former, would have the cathodes between 150V and 200V. The 4GK5 max. heater cathode~heater voltage rating was +/-100V. I felt that tying the heater supply to main supply ground would conceivably therefore risk arcing between the current amp cathodes and heaters.

    It later became apparent that this lack of any reference between the heater and main supplies was contributing significant hum to the amplifier output. This became evident when I hooked up a CRO probe to the heater supply while investigating the apparently high hum levels, and most of the hum immediately disappeared. I consequently added a 1MOhm resistor between grounds to maintain the reference fortuitously added by the CRO probe. Decreasing the value below 1MOhm made no further difference, so I left the value at that.

Detail of the Power Supply Section

  1. The anode power supply voltage needed to be able to supply I0 = 8.3mA, while anode voltage was at 135V. So, 8.3mA dropped across 8.2kOhms is 0.0083 x 8200 = 68V, so the power supply needed to be 135 + 68 = 203V, insofar as the voltage amp was concerned. When the second valve was added to each stage to form the cascode, an additional 135V across this valve was needed; a total of 338V.

    I used a single 116Vac secondary transformer. It had about a 20~30VA rating, judging by its physical size. Two ERD28-08 (800V PIV, 1.5A If(av)) diodes and two 470uF/180V electro's were used as a voltage doubler, giving about 323Vdc off-load. An 820 Ohm , 10W resistor, (a 1/2W type burnt up on switch-on!), then fed a 220uF/400V electro then a choke coil, (- the largest to hand but of unknown inductance value), and then another 220uF/400V. The final supply to the current amp valve's anode was 310V - a little lower than the desired 338V, but as near as was practicable at the time.

Detail of the amplifier itself.

  1. The 220uF/400V electro's can't have been used for quite some time; one persistently arced internally for the first 10 minutes or so of operation before gradually starting to behave normally. The "crack" and the effect of the main supply being dragged down a couple of hundred volts were readily apparent.

  2. In the final circuit both valves in the cascode pair had 100 Ohm grid stopper resistors, and 180 Ohm cathode resistors, across which was developed about 1.3Vdc of bias for 7.2mA anode current. These were later reduced to 150 Ohm resistors for 1.2Vdc of bias and 8mA of anode current.

Circuit Description Summary

  1. Input/Voltage Amp

    Input audio across a 100kOhm pot, wiper via 100 Ohm to voltage amp grid. 180 Ohm, (later 150 Ohm) from cathode to ground. 8.2kOhm/5W anode resistor to output coupling capacitor junction. 1uF capacitor from anode to current amp grid via 100 Ohm.

  2. Current Amp

    180 Ohm, (later 150 Ohm) cathode bias resistor from cathode to output coupling capacitor junction. 220kOhm grid resistor to this point as well.

  3. Output Coupling

    10uF electro with 1uF polycarbonate in parallel to primary of the output transformer which would look like about 1.15kOhm AC impedance with an 8 Ohm load across the secondary. The secondaries were then connected to a panel mount stereo headphone socket.

The first version of the main power supply. The same 116V sec. transformer as the final supply but with just bridge rectification and no voltage doubler, and a series resistor giving just 135Vdc final output, @ 60mA. With Class A amplifiers the load current is constant, so no voltage regulation is needed (insofar as needing to respond to a continuously varying load current dur to audio signals); the output voltage for a given current is set simply by the value of the series resistor in the RC then CLC filter.

The transformers were out of old solid state Philips televisions where their use had permitted the use of the main 155Vdc supply rail in the audio output stage, while only developing a few Watts output. They had a voltage ratio of about 12:1 and so an impedance transforming ratio of this squared; 144:1 and a power rating of about 1 or 2W.

Measured Results

Measured results were as follows;

  1. DC bias conditions;
    • Current amp anode: 310Vdc
    • Across 180 Ohm cathode resistor: 1.295Vdc (so Ia = 7.2mA)
    • Output Capacitor Junction: 185Vdc
    • Across current amp valve: 123Vdc (310 - 1.295 - 185)
    • Across anode resistor: 58Vdc (so Ia = 7.1mA)
    • Voltage amp anode: 127Vdc
    • Across 150 Ohm cathode resistor: 1.3Vdc approx.
    • Across voltage amp valve: 125Vdc (= 127 - 1.3)
  2. Performance Measured
    • Gain: 500mVpp 400Hz input gave 2.2Vpp output; so Av = x 4.3, or +12.6dB
    • Input overload occurred at input 2.1Vpp, (740mVrms, -0.4dB), 400Hz, output 9Vpp, (3.2Vrms, +12.3dB, corresponding to 1.25W in 8 Ohms).
    • Frequency response didn't appear to change at all, (say +/-1dB), from 40Hz to 20kHz. Below 40Hz however, even with only a medium amplitude level test signal, distortion was becoming evident in the sinewave test signal and this was quite substantial below about 30Hz, even though overall amplitude still was about the same as mid-band.
    • Hum remained a serious problem. Grounding one side of the output transformer secondary and the cases of the two power transformers helped. The 1MOhm added to link the "grounds" of the main and heater supplies also made a big improvement. Rechecking that all power supply wiring was physically distanced from the signal wiring, and that the heater supply wires were as short as possible and didn't run parallel to any other wiring, all made no difference at all. About 20mVpp 100Hz hum remained across the headphones output.

Later I did try a DC heater supply in my efforts to reduce the hum still further. The change from AC heater supply to DC supply had no effect at all on the hum level, but I've included the schematic for it here anyway!