by Helmut Ahammer
[Warning: Like other projects using tubes for amplification, the circuits described in this article contain high voltages, and, therefore, the risk of a lethal accident is evident. Neither the author nor HeadWize is responsible for any damage or harm resulting from the construction of this project. DIYers should be familiar with and follow high-voltage safety precautions when building this amplifier.]
Tube amplifiers designed for headphones have the principal property that they can be used as preamplifiers too. In most cases, the output impedance of a tube headphone amplifier is (or should be) less than the output impedance of a tube preamp. Given this advantage, the use of a tube headphone amplifier for preamplification is not so critical concerning the input impedance of the power amp or cable impedance and cable capacitance. This project expands on the basic amplifier, with a slow turn-on for tube heater and plate voltages and input section using high quality relays.
The main intention of the project was the minimal use of elements in the audio path and the creation of a fully-featured amplifier with high audio impact. Therefore, additional discrete semiconductors and digital (logical) integrated circuits were brought in, serving only as helping devices. Only one tube in the audio path amplifies the audio signal. Here I have to say that I prefer the pure Class A topology, using only a single amplification stage with a parallel connected-output triode without overall feedback. This topology is not new and is for instance also mentioned in the Top-Level OTL Tube Headphone Amplifier and No-Compromise Tube Headphone Amplifier by Andrea Ciuffoli.
This amplifier features relay-based input and power switching. The stepped and slow turn-on for the power supplies result in less stress for the tubes and other components and reduces turn-on thumps that could damage headphones. Most tubes fail at turn-on. Without the slow turn-on, if the tube heater filament is cold, the resistance is lower and the in-rush current of a cold filament could be very high and cause the filament to break. And if the filament is not hot, it is better that the plate voltage is not applied. Applying the plate voltage with cold filaments can reduce the lifetime of the tubes.
I prefer relay input switching over input switching with a rotary switch for two main reasons. First, it is very hard to get a really good rotary switch, and when available, it would be very expensive. In contrast, there are many relays on the market, the contact material is normally very good, and they are relatively cheap. Second, it is possible to locate each input relay very near to the corresponding input jacks. Then there is the big advantage, that the wiring from the input to the switch is very short, and only one stereo cable must be routed through the amplifier. If the tape loop is used, then there is also a very short wiring for the audio path.
The circuits in this article are presented more or less as independent from each other, providing the possibility of an easy change if desired. The headphone amplifier could be built without the preamp sections (input relays and so on) or any part of the power supply could be changed (for example, valve rectified) or a preferred amplifier topology not presented here could be extended with the preamp sections.
CIRCUIT DESCRIPTION
Headphone Amplifier/Preamp
Figure 1
Figure 1 shows the circuit of the tube amplifier designed without global feedback. It is a pure Class A OTL design with a triode-connected pentode output section. The input section was built with the double triode E83CC as a differential amplifier. The big advantage of this circuit is that the output signal can be taken from the first or from the second triode. The output at the plate of the second triode isn’t phase inverting providing that the whole amplifier isn’t phase inverting. The E83CC is very often mentioned as a tube with excellent audio properties. There are selected types and different brands on the market. The E83CC is the high-grade type of the ECC83. It has a very high gain (µ = 100) and is therefore well suited for building up an amplifier with only one amplifying device.
To circumvent having a high voltage negative power supply for the current source (simply R4) of the differential amplifier and for improving the linearity, it is necessary to lift the cathode voltage and the grid voltage of V1 above zero volts. This is done by R7 and R8 acting as a voltage divider. This voltage divider could be built two-fold, each for one of the triodes, but since tubes have a negligible input current (gate current), the same divider gives the gate voltage for the first triode too through R6. With a gate-cathode voltage of about -1.2V and R4 = 4.7K Ohm, the plate current for each triode is about 1mA. This current increases the lifespan of the tube and gives very good results. The resistors R2 and R3 are the plate resistors and provide a plate voltage of about 170V (with a 280V power supply).
The higher the power supply voltage the higher the obtainable gain. With a power supply of 280V, the gain of the input section was measured to be about 22dB. R1 and R5 are gate resistors commonly used to reduce RF oscillations. Because the differential pair only uses one input, the other input must be grounded. C2 grounds the second input at audio frequencies. The gate is at a lifted potential for DC voltages, but is at ground for AC voltages. Like the input decoupling capacitor C1 and the interstage coupling capacitor C3, the gate capacitor C2 should be high performance too. Industrial standard MKP types (e.g. Epcos or Wima) are recommended.
The output stage is a cathode follower (gain < 1) with a triode-connected pentode EL84 instead of paralled triodes. If I had used triodes with high enough current (30mA or more), the plate voltage would have been, say, about 100V-120V. Then the voltage at the cathode resistor would have been in the range of 160V-180V. As the maximum cathode heater voltage must be less than 100V or in practical terms less the 80V, the ground plane of the heater supply would have to be lifted by 80V-100V. But lifting the ground plane with all the implemented control circuits (relay control, CMOS circuit, etc.) is not good. Furthermore, I dislike high voltages at the output capacitor. A fault of this capacitor could pose a lethal risk. Therefore I decided that the cathode resistor voltage should be not more than 80V. But what to do with the remaining 200V! There are no noval socket triodes (relatively cheap triodes) that can withstand this high plate voltage.
The solution I found was the triode-connected EL84 with a 210V plate voltage and about 35mA current – permitting a cathode resistor with lower wattage and lower cathode voltage! The drawback is the higher output impedance compared to paralleled triodes. The triode mode of the EL84 has lower output impedance and distortion compared to the pentode mode. R11 connects the second grid to the anode for the triode mode. The cathode current of 35.5mA is set with the cathode resistor R14 and the biasing resistors R12 and R13. The paralleled biasing resistors could be changed to one resistor with a value of about 180 Ohm and with a higher power handling capacity.
The output impedance of the cathode follower is approximately calculated with Rout = 1/S (S is the transconductance of the triode connected pentode). From the data sheet plate current/ gate voltage graphs in triode mode a value of S= 16mA/V, slightly higher than 12mA/V for the pentode mode is arrived and therefore Rout calculate to about 60 Ohms. This value works fine with headphones of 300 ohms impedance concerning the damping factor. By using headphones with about 30 Ohms, not only there is no damping, but furthermore the gain of the stage would be reduced. If gain isn’t that of importance low impedance headphones could give good results too, despite of the lack of damping.
Possible changes:
The resistor R4 with an actual value of 4.7K Ohms is not a real current source as the impedance of a current source should be much higher (about 1M Ohm). Implementing a better current source, it is possible to replace R4 by a choke or a FET current source. Using a BF245 with a resistor connected between Source and Gate could be replaced with no change for the rest of the circuit. A potentiometer could make the current adjustable. Bypassing R2 by a capacitor increases the gain of the input stage and would lower noise of the circuit. With the experience of the used power supply, which is very stable, noise was no problem and the gain without this capacity is well enough for driving 300 Ohm headphones.
With low impedance headphones, say 32 ohms, the output tube could be used with a current of up to 49mA instead of 35.5mA but then the voltage of the tube, the cathode voltage or the power supply voltage have to be changed. The impedance at the output is connected in AC terms parallel to the cathode resistor and the gain is reduced. Therefore, the maximum voltage swing is limited, and the current through the load is less. Adding an output transformer with a high primary impedance and a low secondary impedance could solve, but it would be better to integrate a transformer directly to the anode or cathode of the tube. Personally I prefer the transformerless approach, because these transformers need to be in the most cases very special ones and therefore are very expensive.
I thought of a using a real Class A MOSFET Follower instead of a tube output stage. Combining the gain (and the sound) of the E83CC tube with a MOSFET (low impedance) output is very interesting but actually I haven’t built such a circuit.
Heater DC Supply with Slow Turn-On
Figure 2
To minimize the possibility of hum problems (arising from AC heating of the tubes), a regulated power supply is used. Furthermore a slow turn-on was implemented (which could be found in similar form in many textbooks about tube circuitry) to limit the big in-rush current (figure 2). This slow turn-on increases the lifespan of the tubes. C1, C2, C3 and C4 reduce spikes from the rectifier diodes. Instead of the four capacitors (C5, C6, C7 and C8), one big capacitor could be used but several capacitors decrease the ohmic losses. The adjustable voltage stabilizer LM317 is used with the diodes D5 and D6, which are safety diodes in cases where negative voltages occur (mainly from turning off the amplifier).
The combination of R1, R2 and P1 set the output voltage. The combination of R4, C9 and Q1 gives the temporal behavior of the circuit at turn on. At the first moment C9 isn’t charged. Charging the capacitor through R4 gives a relative high voltage at this resistor and therefore Q1 is fully switched on. Q1 gives a very low parallel resistance and a very low overall resistance for the LM317 and therefore a very low supply output voltage. By charging C9 the overall resistance for the LM317 and the output supply voltage are raised. The output voltage before diode D7 is about 13.4V (the actual value is not critical) and is for the relay and control sections. The heaters of tube V1 (E83CC) of the left and right channel are parallel-connected. The heaters of tube V2 (EL84) for the left and right channel are connected in series, because they work only with 6.3V.
High Voltage DC Supply with Stepped and Slow Turn-On
Figure 3
The high voltage supply is turned on after a 40-second delay through a current-limiting resistor for slowing down the voltage ramp-up. For the first 40 seconds after turning on the amplifier, only the heater and the relay and control sections are powered. There is no voltage at all at the plate of the tubes (e.g., the output voltage of this supply is 0). After this 40 seconds, the control circuit turns on the relay Rly1 and the power supply for the plates begins to work (figure 3). As the MOSFET (Q2) configuration has also a slow turn-on function, the plate voltage rises slowly and, therefore, the tubes are not stressed. After 60 seconds, Rly2 shorts R1 and the voltage increase is accelerated. R1 and the contact of Rly2 could be placed at the primary of the transformer to limit the in-rush current of the transformer too.
The transformer is an easy-to-get 1:1 or 1:2 (depending on the primary voltage) transformer and should have a power rating of at least 100VA. R2 and R3 effectively increase the impedance of the rectifier diodes. This mimics in a minor manner a tube rectifier because tubes have a higher intrinsic impedance when compared to rectifier diodes. C1 and C2 minimize voltage spikes originating from the rectifier. C3 is actually built from six 150µF capacitors providing fewer losses. R4 and R10 are directly and closely connected to the capacitors to minimize the danger of charged capacitors if there is a fault (e.g., loosened connections to other circuit parts). Q1, D3, D4, R5 and DZ1 set the reference voltage of 285V, which is further smoothed by C4, R7 and C5.
Connecting the headphones (300 Ohms), there is absolute no audible hum or noise, except the volume pot is at the position 9 of 10. At this position it is able to hear the amplified input fluctuations but nevertheless no hum from the power supply. R9 is a safety resistor, if the connection to the output capacitors C6 is broken.
Relay-Based Input-Output Switching Section
Figure 4
The input-output switching section with input resistors for decoupling, a tape loop function (Rly7), the volume pot P1 (high quality recommended), the amplifier itself, an output relay Rly8, decoupled line outputs and the headphone output are shown in figure 4.
It is not necessary to buffer the inputs. There are different resistor values for the decoupling resistors mainly because today’s CD Players have 1-2Vrms outputs and many other inputs (like tuners, etc.) have outputs with less voltage. Therefore, I decided to halve the output voltage of the CD player, as then there is only a small change of volume when the inputs are switched between the CD player and an other units. The resistor values for the CD-input are relatively low because the output impedance of a CD player is normally relatively low too. The resistor values for the rest of the inputs are relatively high (the resistors to ground) because many tube amp owners have other tube gear to be amplified. For this case, the resistors should be not too small as the output impedances are normally higher. Principally, high resistances have the drawback of producing more noise, and the actual values of the decoupling resistors do not have to be exact.
The tape loop could be used to implement a signal processing circuit like EQ or crossfeed when using the amplifier for headphones. The input relays are switched by darlington transistors (BC517). The electrolytic capacitors (C7 – C10) minimize switching pops or crackles.
The tape loop relay (Rly7) and line output relay (Rly8) could be turned on with a normal switch, or with a momentary switch in combination with a NAiS VS5-24V electronic switching circuit and a bistable relay (Rly9 and Rly10) as shown in figure 4. The VS5-24V integrated circuit and a cheap polarised 2-coil relay (from any company) has the function of an expensive and hard to find stepper relay (a relay with two stable switched positions set by pulses).
The same circuit is used for the line output relay (Rly8), but in this case the relay can be switched only after the whole turn-on delay of 70 seconds. When the Tape/Line momentary switch is pressed, the second switch contact of Rly9/Rly10 (between the coil of Rly7/Rly8 and ground) is switched permanently. The other switch contact of Rly9/Rly10 between the two coils of Rly9/Rly10 is only necessary for the proper function of the VS5-24V. LEDs on the front panel indicate the status of Rly7/Rly8.
Turn-On Delay Control
Figure 5
Figure 5 shows the circuit for the turn-on delays which control the course of the turn-on procedure. The CMOS circuit 4060 is a binary counter and Q14 is the highest bit. C3 and R3 give the frequency of the oscillator and therefore the delay duration. At the start, C1 is not charged and pin 12 (RESET) of the 4060 is HIGH and resets the counter. Charging C1 sets pin 12 LOW because of R1, preventing a fault reset. The first time Q14 goes HIGH (after 40 seconds), Q1 is turned on. If, in addition, Q13 goes high (after 60 seconds), the NAND gates IC2C and IC2B turn on Q2. And finally if, in addition to Q14 and Q13, the pin Q12 goes HIGH (after 70 seconds), the gates IC2D and IC2A turn on Q3. D1 stops the oscillator to preserve this state. Changing the oscillator frequency would change the total duration of the whole delay, but would preserve the proportional timing. The formula for setting the frequency is: f = 1/(2.3 * R3 * C3). C1/R1 are reset at power-on.
This circuit starts every time in the same manner, and all the steps are cycled through with the depicted timing, regardless of the usage of the main switch (for example, if the main power switch is turned off and on improperly).
Flashing Led Signaling Slow Turn-On Procedure During Power-On
Figure 6
The circuit in figure 6 provides a rough visual display of status of the different steps of turn-on (and therefore representing the state of the amplifier during power on). Q1/Q2, D4/D5, C1/C2 and R2/R5/R6/R9 are the common components of an astable oscillator circuit. When the amplifier is powered on, the LED (D3) flashes slowly with the time constants given by C1, C2, and mainly R5 and R6. There is no second LED in connection with R2. This asymmetry ensures that the circuit starts oscillation under all circumstances.
After 40 seconds, the turn-on delay control for Rly1 (figure 3) also turns the transistor Q4 on. Then R6 is paralleled with a low resistance and gives a shorter time constant with C2 and, therefore, a faster flashing rate for the LED. Similarly, after 60 seconds, the Rly2 (figure 3) and the transistor Q3 are turned on and the LED flashes again faster. The diodes D1 and D2 prevent the toggle circuit from incorrectly turning on the relays. Playing with the values of R4, R7, C1 and C2, it is possible to change the flashing intervals of the LED. The depicted values are given for operation where the no-light durations are smaller than the light durations.
D1 and D2 prevent the toggle circuit from incorrectly turning on the relays. A “low” signal, produced periodically by Q1/Q2 could eventually be conducted through C1/C2, Q3/Q4 and R3/R8 to the coils of Rly1 and Rly2 during the delay time. Probably the resulting currents to the coils would be far too small to really turn on the relays, but I wanted to prevent it in any case.
And finally after 70 seconds, the output relay is turned on. Q5 and R1 stop the toggle function of the circuit and the LED operates in continuous mode. Overall this gives a very convenient optical control of the start-up procedure with one LED, which is located at the very left side of the faceplate directly above the main power switch.
CONSTRUCTION
The 6BQ5 and The Russian 6P14P (6P14P) in figure 1 can be substituted for the EL84 with very good results. The decoupling output capacitor (C4) should have at least 220µF with an appropriate voltage rating. A high voltage rating with a big amount of margin for this cap lowers the possibility of a failure (70V DC at the output!) and is therefore strongly recommended. Raising the value of the capacity gives better low frequency results. Bridging with an 1µF MKP type is recommended. All resistors are of 0.4W-0.6W type if not otherwise specified in the schematic.
In figure 2, if the BC557 transistor is not available, try 500mW types such as the BC556, BC327 and even the 300mW types like the BC177 should work as well. I’m not aware that there is a direct substitute for the ZTX758 in figure 3, because few PNP types are rated at 400V. If the ZTX758 is absolutely not available, then I would recommend 300V types such as ZTX757 or MPSA92. Substitutes for the IRF840 MOSFET include the IRF740, IRF830 or BUZ40B or any other N-channel MOSFET with a rating higher or equal than 400V, 4A should work. Types with less Rdson should be preferred. The fuse labelled 75 degrees (C) is a temperature fuse mounted inside the chassis which cuts the high power supply, if the temperature of 75° is reached. There are no direct substitutes for the BC517 darlington transistors in figure 4, 5 and 6, but the BCX38C could be used. For the BC107 transistor in figure 6, any small NPN transistor like the BC337, BC546 or 2SC1815 would also work.
The MOSFET IRF840 (figure 3) needs a heatsink with a temperature resistance of less than or equal to 6K/W, and the voltage regulator LM317 (figure 2) needs a heatsink of less than or equal to 4K/W. The “hot” parts of the amplifier (tubes, heatsinks, ev. transformers) should be equally distributed in between the chassis. Mounting the heatsinks in a way that they are an outer part of the chassis should be preferred.
At this time, there is really no information on the net about the NAiS VS5-24V module. I have a single data sheet from the local distributor. RS components and some local distributors here in Europe (e.g., Schuricht Elektronik) sell it. Furthermore I bought a small quantity (2 pieces) direct from the local NAiS distributor. The price is about Euro 9.30. There is no substitute, but I think it would be not so hard to develop a discrete version by any advanced DIYer. “Aromat” is the USA brand for NAiS, but NAiS does not know if Aromat sells the module. If not, it could be purchased from NAiS in Europe. I recommend contacting Aromat about the module and the availability.
If it is impossible to get the VS5-24V module, omit the Rly9/VS5-24V combo and replace the switch SW9b with a push button single pole switch. Electrically this switch could be a cheap one, because neither audio signals nor high currents nor high voltages would be switched. The same for Rly10.
All of the relays should have 12V coils with a resistance greater than 700 ohms. Rly1 and Rly2 have to be good power relays, such as the Siemens V23092 (any reliable monostable relay rated for 230VAC, 4A-6A works well in this application but cheap products should be avoided). They need only one switching contact (SPST). The switch contacts of Rly1 and Rly2 should NOT be moved after the rectifier to switch any high voltage DC – they should switch only high voltage AC. Normal “230V” relays are not constructed to switch 230VDC. I once had a bad experience with a relay specified for 320 VDC to switch 280VDC. The contacts melted together!
The use of high quality 12V relays for Rly3 – Rly8 is highly recommended (e.g., P2 Siemens or DS Nais DPDT types). For switching audio signals, it is important that the relay has a low and constant switch contact resistance at low voltages and currents. The resistance of a relay switch contact is specified with a minimum and maximum rating (the minimum rating should be as low as possible). For example, the Siemens P2 relay has a minimum voltage rating of 100µV (modern power relays (e.g., Siemens V23092) have a minimum voltage rating of at least 5V). The bipolar relays (Rly9/Rly10) have DPST switches. They do not have to be of the same quality, because they aren’t in the audio path and don’t switch high voltages or currents.
The part number of the Siemens P2 relay is V23079. DigiKey in the U.S. sells a Potter and Brumfield V23079 relay that is the same as the Siemens P2 relay. All Siemens relays have this type of number code. Despite that there are slightly different dimension values, I think it is really the same relay, especially since the coil current is exactly the same. The bistable relays Rly9 and Rly10 can be cheap products, because they don’t switch high power and don’t switch audio signals. They must be polarized, 2-coil types and must have at least DPST contacts.
Using a high quality momentary switch for the Tape and Line controls results in a very professional feel. I used the Apem type 18535CD, which costs about Euro 5.84.
As can be seen in the picture at the top of this article, this project could be built up in a relatively small case. The front plate was made of a 15mm ceramic plate. Ceramic is very heavy but absorbs vibrations very well and is therefore well suited for audio gear. The surface is manually fire lacquered giving a slight structure to the appearance. This plate is not 100% flat and uniform but therefore reflects on the other side the hand made aspect and individuality and completes for my opinion the tube approach very well.
To make the faceplate, I used white clay which deforms not so much when fired, and pressed it into a wooden form and dried it for about 4 weeks before firing. The wood form is a piece (about 40mm) of glue-pressed wood with material cut out with a router. The faceplate was fired in a kiln slightly above 1000°C for end stability (only drying the clay does not work). After firing the faceplate, I lacquered it with common spray lacquer (I didn’t use any glaze). The lettering is engraved with something like a small drilling machine and with high rotation. The white colour is the white clay under the lacquer. I tried drilling holes for the switches and pot in the faceplate both before and after firing and had different success, but actually I don’t know which method is better.
The chassis measures 435mm x 319mm x 142mm and consists of single aluminum plates mounted together simply with L-shaped aluminium profiles and screws. The thickness of the aluminium should be at least 2mm, and there should be enough holes in the chassis for cooling the components. Apart from that, many DIYers like to see the transformers and the big capacitors. I don’t, and therefore I constructed this chassis, where only the tubes are visible. The top plate extends over the tubes for protection. To mount the ceramic faceplate, I drilled little holes in the backside of the faceplate and glued little threaded bolts (threaded rods) into it. Then I mounted it with nuts to the aluminium chassis.
RESULTS
For testing, I would recommend measuring the voltages across the cathode resistors as they determine the current through the tubes (see the DC operating points shown figure 1). Also measure the heater voltage, the high power supply voltage and the voltages at the anodes. If all these voltages are OK, the amplifier should work properly. The start up procedure could be tested easily by hearing if the relays click at the right order in time. The proper function of relay Rly1 and Rly2 could be tested further by measuring the voltage right after the switch contact of Rly2 (before the rectifier). The slow turn-on of both the high voltage supply and the heater supply could be tested by measuring the voltages during the start up.
The voltage drop at the MOSFET Q2 (high voltage supply) should be in the range of 20-25V. When all the tube heaters are connected and warmed up, the output of the low voltage (heater) power supply is set to 12.6V with P1. When the power supply is turned on for the very first time, it is recommended to set the trim pot to the mid position and then adjust for 12.6V at the output. After connection of all the loads and after warm up, the voltage must be readjusted.
The heaters of V2 are connected in series across the 12.6V supply. If there are big differences in the heater resistances, it could be that one heater would run with a higher voltage. If each individual heater voltage does not exceed the specification (mostly ±10%), minor differences don´t matter. I have never had this problem, because normally I use pairs of tubes with the same “history,” and even during experimentation with different tubes (age, brand, etc) I had no problems. It could be, that even a tube from the same batch has a failure, and then the second tube would be influenced concerning tube performance or lifespan. Therefore, I recommend checking the heater voltages, when the tubes are turned on the first time.
Here are the output current measurements of the amp driving a 30-Ohm load and a 300-Ohm load:
30 Ohm resistance:
Maximum current 15.5 mA rms without clipping
Maximum current 32 mA rms with clipping
300 Ohm resistance:
Maximum current 10.5 mA rms without clipping
Maximum current 20 mA rms with clipping
I have compared this headphone amp to an old Technics receiver and to a hand-made solid state pre-amp built with a NE5534 opamp and a push-pull BD139/BD140 transistor output stage for driving headphones. The Technics receiver is the worst because of noise at the output and inferior audio quality. Mainly, there was no clear musical representation and therefore a very bad localisation of single instruments. My hand-made amps perform, in my opinion, very well. It’s hard to say why, but I like my tube amp more because of the overall tonal representation, especially during longer listening sessions. This is the reason why I presented my tube headphone amp and not my semiconductor amp. I’m using only my HD580 with this amp.
c. 2002 Helmut Ahammer.
HeadWize Memorial 
