A Low-Voltage Class-A Tube Headphone Amplifier.

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.

This amplifier (which I call the “VR2”) is the follow-up of my previous HeadWize project (Tube Headphone Amplifier/Preamp with Relay-Based Switching) and has been designed with following conditions having in mind:

1) Pure Class-A triode OTL design and only one tube for amplification.
2) The plate voltages should be low.
3) The output impedance should be as low as possible and the maximum output current should be as high as possible.

This amplifier is a stand alone Headphone amplifier without the property of input selection as mentioned in my previous project but could be expanded if desired.

ad 1) Preferring the Class-A OTL design I decided again to implement a long tailed pair for the input section and a parallel connected cathode follower for the output section.
ad 2) The voltages should be as low as possible. This condition was taken into account mainly because the risk of any high voltages at the output of the amplifier, if any damage occurs, should be kept as low as possible. This could be the case if the output capacitor shortens and then the full cathode voltage would be connected direct to the output, damaging the headphone or there would be the risk of touching lethal voltages! In my opinion it’s better that this voltage is 40V and not 150V or even higher.
ad 3) Low output impedance and a relatively high current for the output sections are needed to drive headphones and in connection with the condition 1) and 2) only a few tube types are suitable.

These conditions could be fulfilled very good with the 6DJ8, ECC88 tube type family. An operating point with a plate voltage of only 80V is well between the limits and the gain µ of 33 is well enough. For the parallel connected cathode follower the E288CC is well suited, delivering higher currents and an output impedance of 25 Ohm. The voltage at the output coupling capacitor could be with this tube about 40V.

Circuit Description

Figure 1

Figure 1 shows the whole circuit of the Headphone amplifier for one channel. Even the high voltage decoupling resistors (1k5 and 330 Ohm) and capacitors (100µF/400V and 220µF/400V) are implemented for each channel separately.

The audio input is connected to a volume potentiometer (47K log, ALPS, Noble or Panasonic) and then coupled with a 220nF MKP capacitor to a long tailed pair with the E88CC triodes . The long tailed pair delivers a signal which is not phase inverted and therefore the whole amplifier is not phase inverted. Pin 7 is connected to ground, because the amplifier is designed for asymmetrical input signals. If an upgrade to symmetrical signals is desired pin 7 could be connected to the second input signal. The 1M resistor connects the grid of the first triode to ground and the 100 Ohm resistor blocks RF oscillation of the circuit. The common cathode resistor with 180 Ohm sets the operating current of 5mA for each triode and the plate resistors 7k5 set the plate voltages of about 80V. The anode of the second triode is connected to the 220nF coupling capacitor. The output section is a parallel connected E288CC cathode follower. The 470K, 33 Ohm and 1K/3W resistors set the operating point of the tubes (about 76V plate voltage and 21mA plate current). The cathodes are connected to the output coupling capacitors 470µF/400V and 1µF/400V. The output resistor 4k7/1W pulls the output to ground if no load is connected.

High Voltage DC Supply

Figure 2

The supply was build up for each channel separately, only the power transformer is used for both channels (Figure 2). To save money I decided to serial connect the secondaries of two easy gettable toroidal transformers. The primaries are connected in parallel. Two smaller transformers instead of one big transformer have the additional advantage that they can be placed very space saving into the chassis. The serial connection of 2x18V and 2x55V transformers gave under load an AC voltage of about 167V.

I used a tube rectifier for this project, because the rectifier tube implements a slow turn-on characteristic for the amplifier tubes and eliminates turn-on cracks. In my first project, I used a lot of electronics to do this job. When using tube rectification (EZ80) with only one supply voltage it is necessary to implement full wave rectification with two additional diodes (1N4007). The cathode of the rectifier tube is connected to the capacitor 47µF /450V and a 20Hy choke with at least 50mA current specification. The 47µF capacitor should be of high quality and rated at least with 400V because the periodic current loads are the highest for this capacitor. The output capacitors are a combination of electrolytic (1000µF/400V) and foil (1,5µF and 100nF MKP) types. The 220K/2W resistor should be soldered near and directly to the big electrolytic capacitor to discharge the capacitor when there is no load.

The supply has a slow turn on characteristic of about a half a minute because the rectifier tube gets conducting slowly accordingly to the warm up of the heater filament. The output voltage is about 135V under the specific load of 52mA. Especially the voltage drop of the choke and the tube rectifier is dependent on the load current. If the choke with an actual resistance of 750 Ohm is replaced by a choke with an other resistance the secondary of the power transformers has eventually to be changed accordingly. Using relatively cheap toroidal transformers, it will be not very cost intensive. Independent of the secondary voltage, the minimum VA rating should be chosen so, that there is a maximal current rating of 1A.

Heater AC Supply

Figure 3

The heater supply is simply a 6.3V 40VA transformer with two 100 Ohm/2W resistors connected to ground for hum reduction (Figure 3). Additionally for saving money, I bought a cheap 12.6V 60VA transformer for halogen lamps. Then I rewinded from the secondary so much turns that there were the desired 6.3V under load. Doing so, one have to keep in mind that the VA rating of the transformer is halved if conservatively calculated. The rewinded transformer has at least the same current rating of 60[VA]/12.6[V] = 4.76A and therefore a VA rating of at least 6.3[V]x4.76[A] = 30VA. The actual current draw of all tubes is 2.75A. Therefore there is some margin for the case of using other tubes with higher heater current demands.



The tube for the input section is the double triode ECC88 or 6DJ8. Better versions are the E88CC, CCa or 6922 or even with less noise the E188CC or 7308. There are many brands NOS and from current productions available. I have good experience with Philips ECC88 NOS and JAN-Philips 6922. This type of tube has µ = 33, S=12.5mA/V and Ri=2k6. Nearly equivalent is the Russian 6N1P which has slightly differing specifications.

The tube for the output section is the double triode E288CC or 8223. This tube is sometimes falsely referred as a replacement for the E88CC type of tubes. Despite that this tube has only µ = 25 the operating current must be much higher. Furthermore the inner resistance Ri = 1k25 is less. The plate voltage is about the same as for the E88CC type. For the headphone amplifier output section this tube is very well suited. Relatively low plate voltage, a current of 20mA for each triode is far below the maximum power limit and with S = 20mA/V the output impedance of the parallel connected cathode follower is 1/(2 S) = 25 Ohm. This tube is a special quality double triode with a tested life time of 10 000 hours, gold pins and with a noval socket. I used Siemens NOS and the operating point matched instantly the data sheet very closely. I can really recommend this tube.

Figure 4

If there is no possibility to get this tube, the Russian 6H30 could be used with slightly changing the circuit. Comparing the data sheets it is possible to implement this tube by changing the 33 Ohm resistor. As this triode needs about -3.4V instead of -1.4V grid voltage for an operating point of 80V and 20mA per triode the resistor should be changed to about 80 Ohms for a common cathode current of 42mA. The exact value depends on the actual tube and has to be examined by a view trials. As S=18mA/V for this tube, the output impedance calculates to 1/(2S) = 27.8 Ohm. The higher heater current demand of this tube should be mentioned too. An other possibility is the use of two E88CC tubes and therefore four triodes connected parallel together. The same current of about 40mA could be achieved and with four triodes the output impedance is calculated to 1/(4 S) = 20 Ohm. I tried it and replaced the two E288CC triodes with four E88CC triodes yielding quite the same operating point. But on the long run I think it is better to split the cathode resistors that there is again one resistor for two triodes to ensure a better equivalent current distribution through the tubes. This alternative output section is shown in Figure 4.

There are many types of the ECC88/E88CC type tubes available with probably different electrical characteristics. Therefore, if the current at the 1K /3W resistor is not about 40mA it is possible to set the operating point by changing both 68 Ohm resistors simultaneously. Increased resistor values will cause a decreased current and decreased resistor values will cause an increased current. Paralleling tubes causes an increase of the input capacitance and therefore paralleling is not very often recommended. Nevertheless I had no bad experience by paralleling tubes, there is no audible high frequency roll of or so. The MOSFET has a higher input capacitance too and therefore there are problems in high speed switching applications, but there are a lot of excellent audio circuits around.


The rectifier tube EZ80 could be replaced with the 6V4 or any other tube with the same or even higher current range. The EZ80 is rated for a maximum current of 90mA. Therefore there is a margin when using two tubes, one for each channel respectively (the current demand for one channel is 52mA) . With less lifetime there could be used one tube for both channels which I would not recommend. Using one tube for both channels the better solution would be to use the EZ81 or 6CA4 with a maximum current of 150mA. Using other tubes than the EZ80 would lead probably to an other voltage drop at the rectifier diodes and therefore the plate voltage would change or this has to be taken into account and the secondary voltage of the transformer would have to be adapted correspondingly.


Any serial combination of transformers could be used and as mentioned above the current rating for both transformers should be at least 1A. As transformers have lower voltages with the load connected, the right combination should be worked out by experimentation.


All capacitors with values up to 1.5µF should be MKP foil types. For this types I recommend the Epcos MKP capacitors. The electrolytic high capacitance capacitors should be industrial grade types. I can recommend Aerovox BHC electrolytic capacitors. They have larger dimensions as compared to standard brand types but have less resistance and a very long life time. The high voltage ratings for the capacitors are needed because the high voltage power supply is not regulated and with less load current the voltage raises. For instance, if one output tube is disconnected the voltage increases to over 200V!


All resistors should be of metal film type with a power handling capacity of about 0.5W or higher which is then specified in the circuits.


At least 20Hy are recommended and the current rating should be at least or in the best case 50mA. If a choke is used with an actual current which is less than the nominal current the inductivity is decreased. The resistance of the choke determines the voltage drop when used with a specific current. If the choke has an other resistance as about 750Ohm, the high DC voltage would be changed or to circumvent this, the secondary voltage of the power transformers should be changed accordingly. I bought the chokes at the German transformer and tube gear seller Welter Electronic located in Ulmen/Eifel (Type: Dr.7 20Hy, 50mA, 720 Ohm). Sowter (www.sowter.co.uk) sells the type CB25 with 20Hy, 50mA and 451Ohm. Hammond (www.hammondmfg.com) sells the type 193C with 20Hy, 100mA and 181 Ohm.



The pictures shows the final realisation with a similar design as my previous amplifier. As I stated in my previous article I like to see the tubes but don’t like to see the transformers and capacitors. Therefore all the transformers and big capacitors are packed into the back part of the chassis. The chassis measures 435mm x 319mm x 122mm and is made from 1.5mm aluminium plates with enlarged hardness. The single aluminium plates are mounted together with L-shaped aluminium profiles and screws. The top of the actual version of the amplifier is built from three pieces of birch plywood which where glued together and sprayed with the same colour as the front panel.


One picture shows the same amplifier with an alternative top made of transparent lacquered beech plywood. The front panel is again made from white clay and has a thickness of 15mm. I drilled the holes after the drying period of about 6 weeks and fired it at about 1000°C in a kiln. After lacquering I engraved the letters with a high rotation mini drilling machine. Because of this ceramic front panel the amplifier is really heavy.


Using good components this headphone amplifier is a high grade amplifier incorporating a really Class-A operation with all it’s sonic benefits. The use of only one amplifying device strengthen this approach. The sonic quality of this amplifiers lies in the excellent reproduction of the music especially with moderate volume levels. All the details are there without being overwhelmed by any special sonic property. Therefore this amplifier is very neutral but without any insistent behaviour. Long term listening with the Sennheiser HD580 or the Beyerdynamic DT770 Pro (250-Ohm) headphones is really a pleasure.

The overall gain = 12dB and with modern CD-players, which have outputs of 1V up to 2V the gain is far enough to get really loud volume levels. With my Philips CD-Player I turn most of the time the volume at the position 1/4 from maximum and the position 1/2 is really loud. I can’t hear with the maximum setting because this is far too loud and ear splitting for me. If really more gain is necessary there is the possibility to connect a capacitor (10µF 200V) to the 120V connection and the pin 1 of the E88CC triode. This capacitor shortens the 7k5 resistor for AC voltages and increases the gain of the first triode circuit which is principally a cathode follower. This capacitor should lead to a gain of 18dB. More gain is even achievable if the E88CC is not wired as a long tailed pair. Instead of the long tailed pair it is possible to build up two common anode amplifiers using separate cathode and anode resistors and an additional coupling capacitor. In this case the phase isn’t inverted too.

Compared to my first project I can say, that the bass control is increased (mainly because of the lower output impedance of 25 Ohm). Especially the bass with the DT770pro is more defined and accurate, but this behaviour is less pronounced with the HD580. Overall the tonal behaviour is very neutral and balanced with a slight bit of warmth and it is not the type of amplifier which is clinically detailed. Compared to the headphone output of the Philips CD-player, the tonal presentation is full of life with a very clear representation.


Finally here are some values from measurements I have done so far:
Frequency response: 10Hz-100kHz
Overall Gain: 4
Output Resistance: 25 Ohm (24.8 Ohm and 26.7 Ohm for both channels respectively)
Maximum Output Current: 300 Ohm or 30 Ohm load: 23.3mA rms, 33mA peak
Maximum Output Power:
300 Ohm load: 163mW rms, 327mW peak
30 Ohm load: 16.3mW rms, 32.7mW peak

c. 2003 Helmut Ahammer.

A 50mW Class-A Headphone Amplifier.

by Rohit Balkishan


This amplifier was born out of a need to use two sets headphones with my computer’s sound-card. The design presented here is a 50mW power amplifier meant for phones with impedances of 32 Ohms and greater. I chose this class A topology, because it offers very good distortion figures without a lot of complexity. A simple common emitter amplifier, for example, is not very linear and the overall gain is very much device dependent. In the case of my amp, it uses a voltage feedback (VFB) topology, and the gain is dependent only on the ratio of 2 resistors. Plus, the amp has very good power supply and common mode rejection on account of the differential input pair and the current source used to bias it.

The Circuit


With a simple common emitter amplifier, the gain is given by the formula (assuming hfe >> 1) Av = hfe * Rc / [(hfe * Re + hie)]. hfe is the transistor current gain and hie is its base-to-emitter resistance. Rc is the collector (or load) resistor which biases the transistor and Re is the emitter resistor which provides bias stabilization and local -ve feedback. hie is device dependent and highly non-linear. It varies with the collector (or emitter) current and causes the gain to vary with the collector current, resulting in distortion. If Re is large enough to make hie negligible, then Rc will also need to be large and the amplifier won’t be able to source/sink current into low-impedance loads.

My headphone amplifier is a conventional VFB type employing commonly available parts. Let’s consider a typical VFB setup using BJT transistors. We have a differential pair input stage, a voltage gain stage, an output (current gain) stage and a -ve feedback network. For the explanation, I will not include the current gain stage since it has no role to play as far as the voltage gain is concerned. The voltage gain stage is a CE stage with a constant current source (CCS) for the collector resistor and theoretically has an infinite voltage gain as per the above formula (the output impedance of a true CCS is infinite). Also note that the presence of a CCS minimises the variation in gain due to hie, which becomes extremely small and can be neglected, and also makes the use of an emitter resistor unnecessary.

Figure 1

Although the non-linearity aspect of the CE stage is minimized, the infinite (or very large) gain needs to be brought down to a more useful level. This is where the differential input stage and -ve feedback network come in. The differential stage operates by way of comparing the signals between it’s inverting and non-inverting inputs and tries to make the difference equal to zero. The signal to be amplified is applied to the non-inverting input (this is determined by where the CE stage gets it’s input from) and the -ve feedback network applies a part of the output of the CE stage to the inverting input.


Since the differential pair inputs are actually the bases of the transistors forming the differential pair, both are high impedance inputs and for practical purposes we can neglect the currents in them, if offset voltages are not an issue. Due to the infinite open loop gain, and the fact the the voltage difference between the inverting and non-inverting inputs is always zero, the closed loop gain is only dependent on the -ve feedback network, which in this case is simply a voltage divider connected across the output of the amp, and the feedback voltage taken from the point where the resistors are connected to each other.

Another thing to be noted is that due to the VFB topology, any change in temperature or supply voltage appears as a common-mode signal at the amp’s inputs and is suppressed (of course the temperature range must reasonably within the operating range of the individual devices). So this type of amplifier has an extremely stable gain which can be controlled simply with two resistors. Note that the VFB amp (without the feedback network – this is external) is in fact a discrete form of an op-amp, though not as good as an IC op-amp.

In the circuit in figure 1, Q1 and Q2 form the differential pair input stage with Q8 and Q9 as active loads. Current source Q3 biases the input stage at about 520uA. Q4 is the gain stage, biased at about 520uA by the current source Q5.


The output stage is comprised of the compound emitter follower Q6/Q10 and the compound current source Q7/Q11 which biases the emitter follower at about 124mA, which is about 10% higher than the peak output current (110mA, measured at the transistor side of C5). Q7/Q11 form the CCS along with D1 and R11/R12. This amp does not use any push-pull arrangement for the output stage.

CCS operation: The voltage accross R11/R12 is the zener diode voltage less the Vbe drop of Q7 (about 2V). This means that if the diode voltage is constant, the current thru’ R11/R12 will be constant => current in the emitter terminal of compound pair Q7/Q11 is constant => since, Ic = Ie (neglecting base current) the current in the load (Q6/Q10) connected to the CCS will be constant. R5 provides the diode current. Note that in the prototype, I have used LEDs in place of zeners. This is fine, except that a slight reduction in the o/p stage bias current will occur and the LEDs must be connected in a direction opposite to that of the zener (pointing downwards instead of upwards, in the schematic).

The closed-loop gain (excluding the input attenuator formed by R1, R2 and R3) is 1 + R7 / R6, as it is used as in the non-inverting mode. Here, R6 and R7 set the gain at 12dB. The gain is measured as the ratio of voltages (RMS) at the collector of Q10 and the base of Q1. R1 and R3 form an attenuator to decrease the input sensitivity of the amplifier to about 850mV for an output power of 50mW. To increase the sensitivity, R1 can be reduced. Changing the gain of the amplifier is possible by changing the feedback resistors R6 and R7. In the case of my amp, the feedback resistors have already fixed the gain to a value higher than needed and the input signal is attenuated to compensate for this. So to change the overall gain either the attenuator resistors can be changed or the -ve feedback resistors can be changed. If the attenuator is changed then the input caps must also be changed to maintain the lower and higher -3db frequency points.


D1 (2.7V, 0.25W) is the single voltage reference used by all three current sources (this can be replaced with a standard red LED, with a small reduction in bias currents). R5 sets the zener current to about 7mA. Q10 dissipates about 530mW and Q11 dissipates about 400mW and these need to be mounted on small heat-sinks.

The amp’s output impedance, Zo (as seen looking into R14 or R15, with both phones connected [assumed to be resistive, not inductive]), is 5.6 Ohms at frequencies > 100Hz, rising to 6 Ohms at 20Hz. This is as indicated by my simulator.

Figure 2

The ±5V power supply is a typical dual power topology (figure 2). Filtering is provided by the 2200uF/25V electrolytic capacitors. The voltage available at the caps is applied to the regulators as shown and the outputs of these are bypassed by 100nF ceramic caps. Note that the regulators don’t have any reverse voltage protection, since the PSU is meant to be assembled as part of the amplifier without any provision for connecting/disconnecting with the amp. Assembly is again not very critical except that the regulators must be fed from the caps and not the bridge and insulation must be proper to avoid shock hazards.

Here are the specifications (as simulated in SIMetrix):

Power: 2 x 50mW, 32 Ohm headphones.
Distortion: < 0.5%, 20 Hz to 20 KHz.
Input sensitivity: 850mV for rated power.
Frequency response (-1 dB): 5.5 Hz – 159 KHz.


The amp uses commonly available parts and can be constructed at a very low cost. It cost me about Rs. 200 (INR) (that’s about 4$ US) including everything (PSU, perf board, jacks, box, volume control, components, etc.). Note that this particular design is meant to be used as a distribution amp for 2 sets of 32 Ohm headphones of 50mW power each. Construction is not at all critical. I simply wired the circuit as it appears in the schematic. It is extremely stable and will not oscillate unless there is something seriously wrong with the wiring. I have mounted the components on a perforated board and connected them using ordinary stranded wires.


All resistors must be 1/4W, and 1% metal film types for minimum noise. The volume control, which is a 47K dual audio pot, is connected in a conventional manner as a variable voltage divider (wiper going to the amp’s input).

C2, C3 and C5 must ideally be non-polar types (rated at least 12V) although standard polarised electrolytics can be used. C2 and C3 can be obtained as non-polar units by simply connecting like poles of two 100uF capacitors (+ to + or – to -). The overall voltage and capacitance of the combined pair will be Ctot = 1 / (1 / C1 + 1 / C2), and the voltage rating will be the sum of the voltage ratings of the individual caps. For this amp though, polarised caps can be used without any problems since the voltage across the caps is < 50mV.

All electrolytics must preferably be bypassed with 100nF capacitors (not shown in the schematic). The 100nF caps are ceramic types, but film caps may be used (I really doubt if it’ll have any audible improvement though). C5 is used to prevent any DC from reaching the phones in the event of o/p device failure. Although a failure is very improbable, I have included it as a fail-safe, and can be omitted if a small amount of DC offset is acceptable. With C5 in place, the lower -3dB frequency becomes about 3 Hz with both A and B phones connected (1.8 Hz with either A or B). If C5 if not used, then C2 may be chosen to get a lower -3dB frequency of about 3 to 5 Hz. The amplifier may be used with headphones having an impedance of 32 Ohms or higher.

The transistors need not be very critically matched. For the power devices, the range of hfes allowable is 50 to the maximum available for BDxxx devices and the matching has to be done to ensure that hfes are within 20% to 30% of each other. As for the BCxxx devices the same above rules apply. The minimum hfe being 100. It may be noted that the output stage is a compound pair and as such, the combined hfe (of Q6/Q10 and Q7/Q11) is very high, making the absolute values and matching less critical.

Q10 and Q11 (optionally) need a heatsink of about 40mm x 40mm or thereabouts of 1 or 1.5 mm thick aluminum plate. These can be avoided if the box is big enough to provide good ventilation. The transistors dissipate about 600mW each, without a heatsink. So using heat-sinks will assure complete reliability. The prototype does not use heatsinks for the transistors.

It is very much possible to substitute transistors. The following is the min. required rating for the transistors Q1 to Q9 (BCxxx ratings in brackets):

Power: >= 0.1W (0.5W)
Ic: >= 20mA (100mA)
hfe (DC): >= 80 (150 typ.)

The min. required rating for Q10 and Q11 is (BDxxx ratings in brackets):

Power: >= 3W (8W)
Ic: >= 500mA (1.5A)
hfe (DC): >= 50 (100 typ.)

Looking at these ratings, I guess the choice of transistors for substitution is quite large, but the transistors that I have used provide a very wide safety margin and I would recommend that any substitutions be with equivalents rather than based on the min. required ratings. Of course, transistors with better ratings than BCxxx & BDxxx can be substituted without any problems.

The power supply must be ±5V regulated with a minimum current rating of 300mA per channel (600mA for stereo operation). It uses a 7.5-0-7.5VAC, 600mA transformer to step down the mains to a usable level. The rectifier bridge is comprised of 1N4007 diodes. Any IC bridge can be used instead of the diodes as long as it can handle up to 1A of current. The voltage regulators are mounted on heatsinks.

I haven’t used a fuse with the PSU of the amp that I have built. In spite of this, I would recommend using a 300mA fuse for 240V (600mA for 120V) mains in series with the transformer’s primary (use the nearest standard values available).


The picture above shows the internals of the amp. At the left is the power transformer. The PCB mounts, the PSU rectifiers, capacitors and regulators (on heatsinks) are next. The output caps can be seen towards the top edge of the PCB (one in the middle and the other towards the right edge). Just below the output caps are the output transistors (not on heatsinks) and their current source transistors and LEDs. The headphone out jacks and volume control pot are towards the bottom of the image. The power switch and input jack are towards the right-most end of the box. The input wires (from the input jack to the volume control) – these need to be of shielded type so that any noise pick-up is avoided.


I have used a plastic enclosure, since it’s very easy to work with. A metal enclosure can well be used, provided it’s properly earthed with care taken to avoid a ground loop. There is no hum or noise in the amp. By now it will be quite clear that I’m not very good at mechanical workmanship. The extra holes that can be seen in the front (to the left of the volume knob) and side (to the right of the power switch) views are a result of this. smile


As for the sound, I found it to be of the same quality as the source used to drive it, such as my computer’s sound card headphone out, which can also drive headphones directly. As may be noted, the amp has an attenuator in its input that makes it work with an overall gain of just above unity. As for comparison with other systems, the sound is better than the headphone output of my Kenwood portable music system, when both the headphone amp and the system are being fed by my Sony CD walkman.

The amp does not “improve” the sound of an existing source to which it is connected. It only gives a sound that is the same as that of the source. It’s not possible for a good amp to improve the sound of a bad source, but it is very much possible for a bad amp to spoil the sound of a good source. When the source is the PC or CDP, the amp’s sound is very good, and the amp doesn’t change it except by way of allowing a greater power output and to connect 2 sets of phones. As for the Kenwood system, the headphone output (built-in) is not at all comparable to either the PC’s output or my CD player’s output – it is noisier with quite some amount of audible hiss. So, in all, if I use my Kenwood system with an external source (CDP or PC), the sound is not as good compared to the output of my headphone amp being driven by the same CDP (and PC) for the same music.

Modifications for Driving High Z Headphones like the Sennheiser HD600

The Sennheiser HD600 is a high impedance headphone (300 ohms). As per the headphone power requirements table given in Dennis Bohn’s article, the max. power handling of the HD600 is 80mW. The modifications required are as follows (I have only simulated the circuit for 300 Ohm operation):

a) R7 = 39K. To increase the gain so that the voltage swing will be about ±6.5V.

b) R11 and R12 = 56 Ohms, 1/4W each (or a single 27 Ohm, 1/2W resistor. A 1/4W resistor can be used for the single 27-Ohm but will cause the resistor to heat up slightly). Bias current is reduced. DO NOT skip this step; otherwise the output devices will be running at a bias current that’s much more than needed, causing unnecessary heating.

c) The supply voltage needs to be ±10V DC regulated, at 300mA. This is to avoid clipping. Change the 7805 and 7905 regulators to 7810 and 7910 ones for the HD600 version. Also, the power transformer needs to have a secondary rated at 12-0-12V/300mA to cater to the 10V regulators. The rest of the PSU need not be changed.

d) R5 = 680 Ohms (this is optional). To keep the zener diode current nearly the same as that of the 32 Ohms version.

e) For the sake of reliability, Q10 and Q11 must be mounted on small heatsinks (each will dissipate about 650mW). Although the dissipation has not changed a lot from the 32 Ohm version, still I am suggesting the use of heat-sinks as I have only simulated the circuit for 300-Ohm operation.

The HD600 version must not be used with a low impedance phone as it is. The HD600 version can provide a larger voltage swing but less current. This is fine for the impedance and power requirements of HD600. The 32 Ohm headphone will (assuming a desired output of 50mW) will need greater current, and if connected will cause clipping (and possible overheating).

To make the amp work with both headphone types, the bias current can be increased to that required by 32 Ohm phones and the gain resistor R7 can be switched to different values depending on headphone impedance. To increase the bias current R11 and R12 need to have the older value of 33 Ohms each and R7 needs to be: 39K for 300 Ohm operation 10K for 32 Ohm operation Note that this will result in a dissipation of about 1.3W in the output devices, so good heat-sinking IS A MUST. The design basically remains that for 300 Ohm operation but the output stage is then biased at a higher current. So, the power supply needs to be ±10V at 300mA.

Figure 3

One way to achieve the gain switching would be to have R7 as a parallel combination of 2 resistors R7A and R7B such that – R7A is a 39K resistor and R7B is connected across R7A via a switch (figure 3). R7B needs to be about 15K. Close the switch for 32 Ohms and open it for 300 Ohms operation.

c. 2003 Rohit Balkishan.

A Pure Class A Dynamic Headphone Amplifier.

by Kevin Gilmore


Grado headphones, for me, are terribly uncomfortable, but my amp sure makes them sound nice. (The ones in the picture are the SR225; I also have a pair of HP-1.) They are the only low impedance headphones out there that I know of. Most if not all of the standard opamp-only headphone amps have trouble driving Grados with any decent bass, because they want a high current output. This amplifier can output up to 0.5W into a 32-ohm load – which is unbelievably loud for the Grados.

I built this headphone amplifier for dynamic headphones based on my rules of proper audio design. People who know my designs will realize that this amplifer is much more than just a headphone amp. It is a pure class A design containing a new never-before-seen servo loop that is not part of the audio signal chain in any way. I should patent it, but in any case, it is “copyright 2001 Kevin Gilmore.”

Kevin’s Rules of Proper Audio Design

  1. Capacitors in the audio signal path are BAD. Even the best silver-mica or poly caps exhibit non-linearities at low voltage levels. Capacitors belong in power supply sections and nowhere else. Capacitors used to compensate an amplifer generally mean that the amplifier is otherwise unstable, with poles in the right half plane and is therefore a bad design.
  2. Transformers in the audio signal path are even worse: non-linearities in the gain structure, parasitic capacitance between windings, impedance problems…. Transformers belong in linear power supplies and nowhere else.
  3. Ultra high open loop gain: REAL, REAL BAD!!! That basically means anything with an opamp in it. Opamp circuits with open loop gains of 10,000 or more require large amounts of feedback to make them usable. While this reduces THD, the intermodulation products, and especially the transient intermodulation products are much higher than they should be.
  4. Servo loops MUST NOT be in the audio feedback loop. This rule is also very important. Two of my favorite high-end audio electronics manufacturers put servo loops into the minus input of their amplifiers. Most other manufacturers that use servo loops do the same thing. opamps used for servo loops do not have an output impedance low enough to make them suitable for this purpose. Furthermore the dynamic output impedance of opamps adds non-linearities to the audio when put in series with the gain resistor on the minus input.

Kind of makes designing ultra high quality audio stuff tough. My design goals in this amplifier were: keep the gain per stage down, keep all stages in class A, keep the differential front end from coming even close to clipping by the use of a current source. Because the amplifier has a low overall gain and little feedback, a servo helps to prevent DC voltages at the output. In general, if the open loop gain is kept down to eliminate all or most of transient intermodulation distortion, then the amplifier circuitry has to be extremely linear and low distortion. Otherwise, you end up with something that measures and performs like crap.

The Circuit

Figure 1

The schematic for the standard (non-bridged output) version of the headphone amplifier is shown in figure 1. The open loop gain of the amplifer is about 35. Even with the feedback removed, the THD is less than .01%. That’s important because the more linear an amplifier is without feedback, the more the THD, IM and TIM distortions are reduced to unmeasurable levels with feedback added.

Stage 1 is a dual FET fully-differential fully-balanced front-end. The idling current is 2mA total per dual FET (1mA per FET) and 4mA for the complete front-end stage which consists of both dual FETs. The dual FETs generate the bias which runs the second stage, and keeps it and the resulting output section in class A at all times. The FETs are ultra low noise dual units specifically designed for audio uses. The total voltage gain of the first stage is 50.

Stage 2 is the driver stage. It is a standard class A voltage amplifier – in this case used as a voltage shifter. The voltage gain is 0.5 and the idling current 4.3mA.

The push-pull class A output stage is a series of paralleled emitter follower, current buffers. The voltage gain is 0.9 and the current gain is 75. The idling current is 15mA per transistor (or 60mA for the 8 transistors off the +16VDC rail and 60mA for the 8 transistors off the -16VDC rail). I have designed the output section to run at what I have determined is the sweet spot for these transistors, which is 15mA each. Yeah, it gets hot; its supposed to get hot (but not hot enough to require heatsinks). It’s not possible to make an amplifier with an output impedance less than 0.1 ohm without throwing around a fair amount of current.

The servo circuit is new: most of the servo designs (Mark Levinson and Krell, for example) put the output of the DC servo back into the – leg of the amplifier. I just do not like this. That puts the noise and non-linearities of the opamp inside the audio loop.

My servo feeds back to the current sources for the dual FETs in stage 1. Like all servos, it is an integrator. Due to the large (relatively) integration capacitor and the 1 meg resistor, the frequency of this filter is 0.05 Hz. With even a decent opamp, the servo’s noise is in the tens of microvolts, and does not affect the operation of the current sources significantly.

The servo opamp in this amplifier measures the DC at the output, if any, integrates it and applies it to the midpoint of the two LEDs. The LEDs do have a slight change in voltage with respect to current, about 3 or 4%, and that is enough to make the servo work. Notice that if the transistors or the resistors are very poorly matched, the servo will not work because its total control range is at most 10%. Most standard servos (such as the Mark Levinson or Krell servos) have a much wider range.

For high impedance headphones, a little DC will not hurt the phones. With the low impedance Grados, even 0.1VDC over a long period of time will definitely damage and/or change the sound. If all the parts are hand matched, the power supplies are exactly the same and all the resistors are really good quality, the amp should be stable and should not drift. In that case, the servo could be omitted or replaced with a 20K trimmer pot wired from +16VDC to -16VDC, with the wiper going to the DC adjust pin. The prototype uses 0.05% tolerance resistors, and I hand-matched the transistors. The output DC is less than 6mV and has stayed absolutely stable for the few months I have had the unit.

Figure 2

The amp gives about 0.5W into 32 ohms. In the classical definition of class A, the top transistors would be sourcing 120mA and the bottom transistors sinking 0mA. However the two 3k resistors in the second stage actually prevent the total shutdown of either transistor bank by keeping the opposing stage at an absolute minimum of 5mA. In any case, 0.5W into Grados is unbelieveably loud.

The balanced bridge output version of the amplifier (figure 2) is for those headphones that can be wired as dual mono (see the addendum for instructions on converting a pair of standard Grado SR-80 headphones into dual mono headphones). It has twice the voltage swing, twice the slew rate and 4 times the output power (competes with the $2600 HeadRoom balanced Max amplifier).

Figure 3

The ultra-regulation of the power supply (figure 3) is so over the top and unnecessary that most, if not all, people building this amplifier would not even notice the difference. However there are a number of benefits to this design. First its a dual tracking design. Because the open loop gain of the amp is low, its common mode rejection due to the power supply is not great. However if both the + and – voltage rails move up and down the same amount, there is no bias drift.

Because of the pre-regulators, the total line/load variations are under 0.0001%. The fast capactors allow the power supply to react rapidly and in a controlled manner with highly reactive loads like the Grados. The opamp outputs are active in both directions; they can push or pull to keep the power supply at exactly the right point. The power supply design was an attempt to come up with the absolutely best power supply I could. It is even quieter than batteries.



Download high resolution images of PCB pattern and layout 1
Download high resolution images of PCB pattern and layout 2

The prototype was built with point to point wiring, keeping it tight and tiny. The layout is exactly as shown on the schematic, which is also why the board is only one layer. Although the amplifier is easy to build without a PC board, I designed one for the amplifier. Each board (you need 2 for the standard amplifier or 4 for the bridged version) is 3.3″ x 3.8″. In a few months, I may redo the board in one of the systems where I can ship the file over the internet and get boards back (like expresspcb.com).

The servo’s total control range is at most 10%, so the some of the servo-related parts must be closely matched for optimal operation of the servo. The 500 ohm resistors, 1.6V LEDs, bias transistors and second stage transistors must be matched to within 0.5% or even 0.25% for optimal operation of the servo (the dual FETs are already matched). To match the LEDs, put one in series with a 10k resistor, hook it to 15VDC and measure the voltage across it. Do it to a few of them and pick the closest match.

Figure 4

I used a Tektronix curve tracer to match the transistors. Figure 4 shows two circuits for matching the beta (gain) of the NPN and PNP transistors using just a voltmeter. Simply measure (and match) the voltage of each transistor’s collector with respect to ground. It should be in the range of 10V to 11V for the NPNs or 5V to 6V for the PNPs.

There are no substitutes for the FETs. The PNP and NPN transistors in the prototype were the MPS8099 and MPS8599. Onsemi has discontinued them, but there are still lots available. I am buying all my transistors now over the web from MCM Electronics. I have had too much trouble with everyone else. The 2SA1015 PNPs are $0.46 each; the 2SC1815 NPNs are $0.41 each. The 2SJ109 p-channel dual FET is about $6.50; its n-channel counterpart is $5.90. By the time you add it all up, it’s still under the minimum MCM order, so you’ve got to buy something else.


The high-speed 5uF capacitors in the power supply are not critical. They satisfy the “lunatic fringe” in audio. These capacitors are rated for a slew rate (dv/dt) of about 4 times a standard capacitor. They cost $7.50 from the Illinois Capacitor Company. Similarly, the opamps in the power supply are not critical. In the prototype power supply, I used the Apex PA09, which has a 400V/uS slew rate and costs $167 each (remember, I am crazy). Normal people should use the Texas Instruments OPA549 (10V/uS) that cost $11 each. The LM3xx regulators are heatsinked (dissipating at least 3W each, 6W for the bridged version of the amplifier).

The enclosure is the Mod.U.Line by Precision Fabrication Technologies Inc. (part number 03-1209-BW), available from Newark Electronics. They are about $85 a piece these days. The aluminum enclosure is easy of use, easy to punch, and keeps a decent paint job. The headphone jack is a cheap Radio Shack one. Although it works fine. I have since gotten some Neutrik connectors which I will put in at some point. In theory, there is a ground loop between the RCA jacks on the back plate and the headphone jack on the front plate. Although the 60 Hz hum is about 110 db down, the Neutrik jacks are isolated and will make this problem go away.


The amplifier’s output is voltage limited (not current limited) because the second stage runs out of voltage swing – typically about 6V rms. For a 32 ohm load, the maximum output power is 1.125W (0.1875A). For a 300 ohm load such as the Sennheiser HD600, the maximum output power is 0.125W (0.02A). To increase the maximum output voltage to, say, 10V rms, try increasing the power supply to ±20VDC and change the 500 ohm resistors in the current sources/sinks to 600 ohms (or ±24VDC and 700 ohms – but careful not to fry the output transistors).

The bridged version will output 4.5W into the 32-ohm Grados and 0.5W into the 300-ohm Sennheiser HD600. At full blast, the amp does drop out of class A, but at levels that will fry your ears anyway.

The Result

Adjustments: In the power supply, adjust the top 20K trimmer pot for +24 volts at the tap after the LM317. Then adjust the bottom 20K trimmer pot for -24V at the tap after LM337. (Note: the ±48V taps are shown for reference only; they are not actually used by the amplifier.) If the servo has been replaced with a 20K trimmer pot, adjust the trimmer so that the output measures 0VDC. Anything under 25mV is fine.

I listened to the balanced HeadRoom Max at The Home Electronics Show. The low frequency bass slam is absent from the unit. When I listened to the BlockHead ($3333) and compared it to my unbalanced/balanced amp, the same thing happened. Without the kaboom, it’s just not as much fun to listen to.

This amp gives the Grados and Etymotic Canalphones a fuller and much more upfront sound than the built-in headphone jacks on various players – the bass has a snap to it that it never used to have. And the image moves from around your head to the center of your nose. The Etymotics tend to sound kind of thin and distant with other amps. In general, I have to say that this amp produces bass that is much more solid – similar to putting the microphone in front of a bass violin instead of inside it.

c. 2001, Kevin Gilmore.
From The Homepage of Kevin Gilmore. Republished with permission.


10/25/2001: The author provides instructions for converting a standard pair of Grado SR80 headphones into dual-mono headphones for use with the bridged version of the dynamic headphone amplifier. He writes:

Dual mono Grado SR-80s have more and tighter bass when driven with the bridged dynamic amplifier, which can output twice the voltage, twice the current (not at the same time) and double the slew rate, giving better control of the diaphragm and a higher damping factor.

I cut apart a pair of Grado SR-80s. I had to take them apart anyway – one side was damaged by my toy terriers. There are two ways to wire Grados for dual mono operation: the easy way and the harder way. The easy way is to cut the wires just above the “Y” of the headphone cord. Carefully splice new mono connections to each of the wires.

The hard way is to take the headphones completely apart and rewire all the way to the transducers.

  1. Take the transducers out of the headband.
  2. Take the ear pads off.
  3. Very carefully, with a sharp jeweler’s screwdriver, pry around the side seal. It is superglued in only a couple of spots. Comes apart real easy.


  4. Buy a set of 20-ft, 1/4″ headphone jack extender cables from Radio Shack. Cut one end off at the length you want and then solder to the transducer. I use the tip and sleeve part of a stereo 1/4 jack, and leave the ground connection unconnected.
  5. On the amplifier, use 2 separate stereo 1/4″ headphone jacks (one for each channel) and wire the tip and sleeve to match the headphone plugs. The ground sleeve IS connected to ground in the headphone jacks only (actually decreases the noise level a bit).


Class A MOSFET Headphone Driver.


The circuit described in this article is a MOSFET follower for driving headphones. FET followers can supply high current, but have a voltage gain of slightly less than unity. They are most suitable in applications where the input signal does not require voltage amplification – such as the output of a preamplifier or a portable stereo. If the input signal is too low, you can add voltage gain stage (see Designing an OpAmp Headphone Amplifier).

The class A headphone driver in figure 1 is ultra simple, except it does need a regulated supply and input and output capacitors. Most of the parts are not critical. I used a pot instead of the 100K bias resistors, so I could vary the output side of the MOSFET at 1/2 the supply. This maximizes the ouput potential of this driver. R1 limits the input current. The diodes are reversed-biased protection only. They short out spikes greater than 9 volts and negative 1 volt. In the original I used a 9 volt zener diode instead, but I figured it’s harder to get zeners. The circuit will work without the diodes.

The volume control, Rp, is optional, because the driver is meant to be fed from a source with a volume control. If the volume control is not used, then it would be a good idea to put a discharging resistor in front of the input coupling capacitor C1 to ground – around 100K ohms. Sometimes the preamp is capacitor outputted, but may not have a discharging resistor. So, in the worst case, there would be capacitors back to back, and the voltage bias in between the capacitors would not be necessarily predictable (e.g., at or near zero volts to ground), or could cause some pops when connecting with the power on. A volume control accomplishes the same thing with its resistance to ground.

All capacitors should be the highest quality, especially the output capacitor. The output capacitor is an easier-to-get size capacitor of 470 microFarads, but one could at least double that for better low response. Bypass it with a quality 1 microFarad polypropylene cap for the best sound. I included the discharge resistor R5 to help eliminate pops when inserting the headphone plug. The input resistor would do the same if the input was inserted live. It might also help eliminate turn-on thumps.

The MOSFET (Q1) should be mounted on a heat sink. Almost any heatsink could be used. Q1 dissipates up to 2.5 watts. I think about 2-3 square inches surface area is minimum, and probably should be more. I just mounted them to the chassis with insulators in my model. A LITTLE silicone grease is nice. The MOSFET and R4 will get warm during operation, which is normal. With R4 = 20 ohms, the source resistors and MOSFETS will dissipate up to 7 watts in the stereo pair using a 15 volt supply. R4 could be changed to even lower values – its simply a matter of delivering enough current out of the supply and providing enough heat dissipation.

The IRF513 is an N-channel power MOSFET, rated at V(ds) = 80V, I(d) = 4.9A and r(ds) = 0.74 ohms. I chose a device with a reasonably small resistance. A higher resistance, lower-rated device could also be used. There is a large selection from which to choose. Q1 draws less than 200 ma. at 9 volts. One thing I didn’t try was to use two circuits per side, so one could use direct coupling on the output side by feeding the signal to one side, and using the other side to provide a positive but equal resting voltage creating an effective zero-DC operating point. I wanted something simple though.


A 15V supply could also be used, but then the power rating of R4 should be increased a little. [Editor: Radio Shack sells a pre-assembled 13.8V regulated supply (RS 22-504). If this supply is used, make sure the MOSFET heatsinks can dissipate 5W.] C3 (usually a ceramic type) decouples the power supply from the circuit.

I put the amp in a Radio Shack box with vents (about 5″ x4″ x2″). Since the boxes sold change so often, I don’t know if Radio Shack still carries them. The binding posts on the front were used as auxillary drive amp for experiments. There is a 1/4 stereo phone jack on the front and RCA input jacks hidden on rear. The power supply is built-in.

This headphone driver design is simple, but it sounds pretty good to me. Using a pair of Grado SR60s, I compared the MOSFET amp to the built-in integrated phone amp in my modified Hitachi HCA 8300 preamp. The MOSFET amp was fed by the Hitachi preamp outputs. There was an immediate noticeable difference in sound. The MOSFET had a cleaner overall sound, not as muddy. Connecting the MOSFET amp directly to my Sony CD player, it sounded good and had more than adequate volume.

c. 2002 Greg J. Szekeres.


7/16/98: While not absolutely necessary, small gate resistors (Rg) can be added to prevent high frequency problems with the MOSFETs. They should be placed as close to the gate leads as possible.

6/24/99: Updated value of R3 in figure 1 to 220K ohms when using 9VDC power supply. The author adds these comments:

As shown operating with a 9 volt supply, the value of R3 should be changed to 200K ohms, or 220K, a more common value. When using 9 volts with R3 = 100K ohms, the circuit does not get very warm, and this value of R3 should be avoided. If using a 13.8 volt supply, R3 should remain at 100K.

To make sure this device is operating at optimum, regardless of the power supply voltage, the output DC voltage should measure approximately 2.75 volts in front of the output capacitor – not more than 3 volts for a 13.8 volt supply, and closer to 2.5 volts for a 9 volt supply. The value of R3 can be adjusted so the output voltage is correct, a higher value representing a higher output DC voltage set point.

These voltages are the result from a higher than expected gate-source differential. I am measuring approximately 4.5 volts. It’s possible this circuit would benefit from even higher supply voltage and higher bias potential, but power dissipation will start to multiply rapidly, and there is some hazzard of gate blowout if care is not taken.

6/27/99: Power supply range increased to 9 – 15VDC with R3 at 220K ohms. R4 is 20 ohms, 5W. If things seem to get too hot or the supply can’t deliver a good clean 1 amp, use a higher R4 value: 39 ohms or something in-between. The author adds these comments:

The output DC voltage on top of R4 will vary between 2.5 – 6 volts depending on the supply voltages from +9 to +15 volts. For those who have previously built this device to operate at the lower voltages, R3 should be changed from 100K to 220K. In fact, it is recommended that all designs to +15 volts supply also change this resistor value. A noted increase in dissipated heat will result.

7/25/99: Added image of finished project.

9/7/99: Tomohiko Ishigami built this verion of the Szekeres MOSFET headphone amp in a spacious metal enclosure that once housed a computer power supply. It features the acoustic simulator circuit by Jan Meier (see A DIY Headphone Amplifier With Natural Crossfeed). Note the “star ground” wiring (all ground wires soldered at a single point on the chassis) to avoid ground loops.

I feel it is very good idea to use modular approach. I love this approach so much, I rebuild my Class-A MosFet amplifier with the same technique. I biased each MosFet on one small board while I lumped up my large polyethylene caps on another. Some may think this will take up a lot of room, but the result was much simpler and cleaner boards less susceptible to noises caused by, I guess, interferences. This amp ended up being very pleaseant sounding headphone amp. It is milder and warmer sounding which I love.



1/7/00: For DIYers who will be using Jan Meier’s natural crossfeed filter as a front-end to the MOSFET amp, here are some tips from Jan re: selection and placement of the volume control:

It all depends on the specific circuitry. Generally it might be better to place the pot after the filter instead in front of it. The influence of impedance changes might be less pronounced. A 10 kOhm pot will certainly be too small. 50 kOhm will be a kind of minimum I think.

1/21/00: The IRF513 has been discontinued. Greg Szekeres suggests substituting a MOSFET with an input capacitance of less than 200pF, such as the IRF510, IRF610 or IRF710.

2/15/00: Tomohiko Ishigami wrote:

I am sending you the picture of the upgraded Szekeres amplifier. As you can see the size of heatsink is rather too big. I punched up the power supply voltage to 15V using the LDS-X-15 PSU from Lambda Electronics. Also, the same PSU can output 5A. The resistors that control the quiescent current is replaced with gold metal-looking resistors. (deviation 1% and takes 20W) You can also see the large capacitors. All the film caps have been replaced with Audyn caps (1.5uF) and electrolytic caps with ELNA Cerafine caps. The headphone jack has been replaced with gold plated 1/4 inch socket. It was very cheap but very tough.

The very accurate quiescent current controlling resistors will give you very ideal current supply to the Mosfets. You can use voltage regulators here, but I think those are overkill and hey! I am lazy (valid excuse!). The capacitor upgrade in the path of signal always result in improvement. Especially, the electrolytic must be a good quality. I found Black Gate and other super-exotic parts, but ELNA Cerafine was reasonable in cost and quality. Black Gate rivals or is better than ELNA Cerafine but I hear that the chemicals within have very strange characteristics. (Ok, Ok, my pocket did not like Black Gate …)

The decision as to which headphone jack I would use took a while. I obtained a Neutrik socket. But it was not realistic. The Neutrik Socket is too large and requires you to make a major chasis modification. So I went for cheaper gold plated jack. I took a reamer and enlarged the 3.5mm socket hole I made a while ago. Personally, I feel more comfortable with this cheap one. Neutrik socket does not seem much more rugged. And the cheap one can be replace without modification at any time in future.

Right now, it has reference quality. Of course, not as refined as Melos SHA-1, but considering that it is 10th of the price, it is reasonable comparison. Although it is cheap and very simple, it still is a full blown class-A amplifier. I have to admire how it sounds so good.



5/01/00: The author has submitted several updates to his design:

The first update is a change in the MOSFET gate resistor (Rg) value from 100 ohms to 220 ohms. Also, the author now recommends a power supply of at least 15VDC. A 9VDC supply may not give the amplifier enough headroom.

The next update is an optional AC-coupled voltage-gain front-end for the MOSFET amp based on the Pocket Amplifier by Chu Moy. The OPA132 opamp is configured to run from a single supply by biasing the input at 1/2 the power supply voltage:

I got a modified pocket amplifier circuit to drive the MOSFET amplifier with gain. I’m not advocating the use of the combination to replace the original, but for use by those who need the gain. The 4.7K ohm resistor from the opamp output to ground is one way of making sure the opamp output rides in class A. It provides a somewhat constant DC current flow. The 4.7uF capacitor in the feedback loop decouples DC and also causes a bass rolloff – I think at about 10 Hz. The 1.0 uF capacitor stabilizes the DC bias point in the bias circuit. The protection diodes D1 and D2 are not required in this circuit.


The AC-coupled front-end allows for maximum voltage swing because the MOSFET can be powered and biased separately from the opamp (although both are shown in the schematic being powered from the same 15V supply). With the 15V supply as shown, the opamp will be able to swing between +7.5 to -7.5V. However, the MOSFET (which is biased at 2/3Vcc) will limit the output to about +5V/-5V with a 15V power supply. A +5V/-5V maximum output voltage swing is adequate for all but possibly the highest impedance headphones (e.g., greater than 600 ohms). For driving high impedance headphones, the output voltage swing can be increased by increasing the power supply for the MOSFET. The MOSFET bias voltage should be adjusted (via R3) until Vds is 1/2Vcc.

In those cases where a +5V/-5V output voltage swing is sufficient, the author has designed this DC-coupled version of the front-end gain stage. The MOSFET’s bias voltage is provided by biasing the opamp output at 2/3Vcc.


Finally, several DIYers have asked for a DC-coupled version of the MOSFET amplifier itself. It uses a dual ±7.5V (or higher voltage) regulated supply. The MOSFET Vgs is adjusted via a 10-turn potentiometer until the output voltage is 0V. The schematic show the bias pot value to be 100K ohms, but the author suggests that it could be higher: 250K to 500K ohms. Note: the MOSFET should be allowed to “warm up” for a few minutes and the output voltage checked again and readjusted if necessary for 0V.

Note: There have been reports that this DC-coupled circuit is difficult to stabilize due to thermal drift. Benny Jørgensen’s DC servo (below) automatically compensates for DC drift.


7/15/00: Eduard Orvisky built this DC-coupled version of the Szekeres headphone amplifier. He used a ±7VDC regulated power supply, and says that the amp “achieved excellent sound” in combination with his Parasound Class A preamp. More details of this design can be found on his website.


1/26/01: Benny Jørgensen created this version of the MOSFET amp with an opamp servo that keeps the DC output of the MOSFET at 0V. He has not built it, but did test the circuit in a SPICE simulator. He writes:

The DC servo opamp keeps the MOSFET output at 0 VDC (during idle). With the opamp power supply at +18V and -7V and the MOSFET supply at ±7VDC, the servo can compensate for variations in Vgs from around -1 V to 8 V (it will normally operate in the range from 3V to 7V – the absolute maximum). You can a single set of power supply voltages for both the opamp and the MOSFET, but the servo will not have the same range. For example, if the opamp and MOSFET are both powered from ±12V, then the servo will only compensate for Vgs changes up to 5V. If the opamp can’t deliver enough voltage, there will be DC on the output. MOSFET Vgs vary from model to model and I have tried the make a general servo that will work with most N-Type MOSFETs. The opamp’s negative supply should be less than -3V in order to compensate for any negative voltage at the output.

The idle current in the circuit shown is 7V / 20 ohm = 350 mA. To increase the power supply for the MOSFET to ±9V, just make the DC servo run on +18 V and – 9 V. The DC servo needs a supply that is more positive than the MOSFET itself. Using ±9V MOSFET supply and a 22 ohm resistor (for example) result in an idle current of 9 V / 22 ohm = 409.1 mA.

The opamp should be a FET input type like the TL071 or better the OPA132/604. All capacitors should be poly types. C2 forms the integrator together with the R6 = 330 Kohm. C3 is a low pass filter with R5 = 10 Kohm. C4 makes sure that the servo does not disturb the sound at high frequencies.

The SPICE simulations below are for the version of the circuit, where both the opamp and the MOSFET have the same ±12V power supply. The color of the markers corresponds to the measurement points on the schematic. The input is 1VAC. V(M1:s) is output of the MOSFET. V(C2:1) is output voltage of the opamp – the “working zone” of the servo. V(R4:1) is the bias voltage on the MOSFET gate. The impact of the DC servo falls off above 2Hz (see the Frequency Response). The Time Response shows the servo at startup going from -3.5V to 0V in 2.0 secs.




3/20/01: Tomohiko Ishigami replaced the source resistor on his MOSFET amp with a constant current supply (current source) made from a LM317 voltage regulator. Originally, he set the MOSFET idling current at 500mA (Rref = 2.5 ohms/1W), but lowered it to 250mA (Rref = 5 ohms/0.5W) because the MOSFET and LM317 were running too hot. He writes:


Constant Current Sources are good and allow active elements like MOSFETs, transistors, or even tubes to operate better. The modified Szekeres has survived 120 hours. It sounds surprisingly good. I recommend 250mA idling current. This is done simply by removing 2 of 4 stacked resistors. I used 10 ohm, 1%, 1/4W metal film resistors, because of the very precise value and temperature-stability. Carbon is not precisely valued and not temperature-stable. Certainly I would not use wirewound; I don’t want extra inductance.

500mA was a little too much, so I brought it down to 250mA, which decreased the heat production to 1/4. The amplifier as a whole is warm but not dangerously hot. LM317s should have heatsinks rated at least 10W. MOSFETs should have heatsinks rated at least 25W. If you are insistent on 500mA, you could expose the heatsinks or you could use a fan inside chassis. The sound quality does not change in going from 500mA to 250mA. My PSU is happier as well.


All other parts values are the same EXCEPT for C2. The value of output capacitor C2 in my amplifier was calculated for the AKG K240M 600-ohm headphones. When you drive K240M with this amplifier, you will need a high output CDP or other audio source (~2vrms), easy for most non-portable CDPs. If you have lower impedance headphones like Gradoes or even the Sennheiser HD580/600, you must recalculate the capacitance with Apheared’s equation:

C2 (uF) = 1,000,000 / [2(corner frequency) Rheadphones]

The corner frequency (in Hz) is the 3dB low frequency response drop for the amplifier. So with my choice of 600 ohm phones with 10Hz corner frequency, the value of C2 = 26uF (I used 24uF + 1.5uF – all polypropylenes).

C2 is only for the AC-coupled amp. DC-coupled designs can be modified with a constant current source. Please do not forget to use large heatsinks. Using a dual power supply allows us to make MOSFET source voltage zero. If you have built DC-coupled Szkeres properly, you should need NO output capacitors.

Summary of amp upgrades:

  1. Noble Pot
  2. RCA inputs
  3. Teflon internal connections
  4. Less wiring
  5. Added 1000uF + 47uF + 1uF + 0.47uF resevoir to PSU – smaller than the original 10,000uF (it died, after being fatally wounded with a reamer during case work). Will add new 10,000uF capacitor may be later.
  6. No more crossfeed circuit inside amp chassis. I have external one instead.
  7. DIY interconnect

First stage of upgrade:


You see smaller output coupling caps and 10000uF reserver bypassed with few lower value caps. The Noble 50K pot is used and connected close to RCA inputs. Note that those metal clad resistors are Non-Inductive DALE NH series resistors. Ninety-five percent of wiring is, of course, teflon insulated cables.

Second stage of upgrade:


In the center, there are two moderately-sized heatsinks attached to T0-220 chips, close to 4 blue metal film resistors. Those are the constant current supplies that set the MOSFET idling currents to the original 500mA. Both the current supplies and the MOSFETs heat up enough to boil water drops.

Third stage of upgrade:


Now I have 24uF Solen output coupling capacitors. Please forgive the messy wiring at the output. I was barely able to fit the caps in the box. Idling current is now limited to 250mA which is enough to make heatsinks hot, but cool enough so the chassis stays lukewarm. This was done by increasing the current controlling resistors (Rref). The metal clad resistors are removed permanently.

The DIY interconnect is 3 braided 22 gauge cables. This sounds fresher and more dynamic than RCA coaxial cables. Probably due to low capacitance. Teflon insulation and 3-wire braiding makes a subtle but audible difference.

WARNING: All versions of Szekeres amp may burn your skin since MOSFETs run very hot. So be careful when constructing and handling this amplifier.