PP logo


Headphone amplifier.



HA01 is a universal amplifier for headphones.
It works equally well with low-impedance ( 16-64 Ω ) and high-impedance ( 300-600 Ω ) headphones.
A power limiter limits the available power into low impedance loads.
Some of the design parameters are:


Headphones can play loud - very loud.
Some can cause permanent hearing loss (as in: for the rest of your life) in a matter of minutes.
The sensitivity for headphones is typically from 100 dBSPL / mW to 115 dBSPL / mW.
One manufacturer specifies their headphone with a sensitivity of 100 dBSPL / mW and a continous power handling of 1.5 W. That is more than 130 dBSPL!
You can not use distortion as a measure of "loud". One manufacturer specifies 1% THD @ 125 dBSPL, another <0.2% @ 100 dBSPL.
Headphones do not give you a kick in the stomach. You can not use this as a measure of loud.

Do never play louder than it is comfortable.
If your ears are ringing after using headphones you were playing too loud.
ALWAYS turn the amplifier volume down before connecting the headphone.

Table 1: Maximum allowed exposure time within any 24 hour period according to NIOSH.
Sound pressure levelPermissible exposure timePower level in a low sensitivity
( 100 dBSPL / mW ) headphone
Power level in a high sensitivity
( 115 dBSPL / mW ) headphone
Power level in a loudspeaker
( 94 dBSPL / W )
85 dB8 hours31 µW1 µW0.13 W
88 dB4 hours63 µW2 µW0.25 W
91 dB2 hours0.13 mW4 µW0.5 W
94 dB1 hours0.25 mW8 µW1 W
97 dB30 minutes0.5 mW16 µW2 W
100 dB15 minutes1 mW31 µW4 W
103 dB7½ minutes2 mW63 µW8 W
106 dB3¾ minutes4 mW0.13 mW16 W
109 dB110 seconds8 mW0.25 mW32 W
112 dB56 seconds16 mW0.5 mW64 W
115 dB28 seconds32 mW (1)1 mW128 W
118 dB14 seconds64 mW2 mW256 W
121 dB7 seconds128 mW4 mW512 W
124 dB<4 seconds256 mW8 mW1 kW
127 dB<2 seconds512 mW16 mW2 kW
130 dB<1 seconds1 W31 mW4 kW

Power level in loudspeaker is shown to put headphone sensitivity in perspective.
(1) The typical power limit for mobile devices is around 20 mW.

Headphone drive.

The IEC 61938:1996 standard recommends an output impedance of 120 Ω for headphone amplifiers.
This works very well for the (very) few headphones designed for this drive impedance.
Generally headphones have an impedance that varies 50%..300% over the audio frequency band, so for best linearity the drive impedance should be as low as possible.

A common way to design universal headphone amplifiers is to put a resistor in series with the headphone so it will deliver the same power into 2 different impedances.
The value of the series resistor is:
Rs = √Rl * Rh, where:
Rh is the high impedance load, Rl is the low impedance load and Rs is the series resistor.
For Rl = 16 Ω and Rh = 600 Ω Rs is 98 Ω
This works very well for some applications, but is useless for playing music as the high source impedance will give a very poor frequency response into low-impedance headphones.

Output power into different impedances.
Fig.1: Output power into different impedances.

Blue graph: 7.75 V output without current limit. This is 100 mW into 600 Ω, but 3.8 W into 16 Ω which I find unacceptable.
Red graph: 7.75 V output with a 79 mA current limit. This is 100 mW into 16 Ω and 600 Ω, and 612 mW into 98 Ω.
Lime graph: 9 V output with a 98 Ω series resistor. This is 100 mW into 16 Ω and 600 Ω, and only 207 mW into 98 Ω.

Circuits with a resistor in series with the load.
Fig.2: Circuits with a resistor in series with the load.

Fig.2a is the conventional way of adding series resistance to the load. This will give a very poor frequency response into low impedance loads.
In fig.2b the series resistance is included in the feedback loop. This will insure a low ouptut impedance, but it will probably be a challenge to get this stable into anything but a resistive load.
In fig.2c the resistors are placed in the OP-AMP supply lines. It is possible to get this stable with some OP-AMPs, but distortion is excessive due to the large voltage variation on the supplies.
Fig.2d is an OP-AMP followed by a discrete output amplifier. This does not work either. At clipping, the OP-AMP will dump its short-circuit current through the load, so it is impossible to obtain an accurate current limit.
In the end I had to make a discrete component output amplifier based on fig.2d.


Input amplifier and balance adjustment.

Input Amplifier Schematic.
Fig.3: Input Amplifier.

C101 is a DC blocking capacitor, R101 sets the input impedance at low frequencies.
R102 and C102 is a HF-filter with a corner frequency of 725 kHz .. 350 kHz for source impedances from 0 to 1 kΩ.
The filter is intended to reduce HF pick-up from external wiring. If you use the amplifier inside a system where a HF filter is already present, you can skip this filter (Short R102, leave C102 out).
The input impedance for this filter is below the value of R101 at higher frequencies, so the input impedance will drop from 100 kΩ at 1 kHz to 66 kΩ at 20 kHz.
If you prefer a constant input impedance over frequency, replace R101 with 10 kΩ and C101 with a 10 µF electrolytic with the minus towards U101.
R104 and R105 allows you to add gain to the input buffer. C103 should be 22 pF to 47 pF if you add gain.
R106 and R107 is a voltage divider for the balance adjustment. R107 goes to a potmeter to GND. C104 prevents DC from U101 on the balance potmeter.
R110 and R111 compensates for the gain loss in the balance adjustment.
C105 and R108 is a 16 Hz high-pass filter.
R112 is HF-decoupling for the wire to the volume control and C107 provides DC-blocking. If you have more than 20 cm wire to the volume control, increase R112 to 22 or 47 Ω.

Table 2: Mounting options.
Balance Adjustment.High-pass filter.Modifications.
YesNoChange R108 to 100 kΩ.
NoYesShort C104, R106 and R110.
Leave C106, R107 and R111 out.
NoNoShort C104, C105, R106 and R110.
Leave C106, R107, R108 and R111 out.

Balance Adjustment.

The Balance Adjustment in this circuit is designed to adjust for level differences in the headphones, so its regulation range is limited to around ±5 dB.
It can not correct for the channel difference in poor vinyl or cassette tape recordings.
The balance potmeter can be a single-gang or a dual-gang 1 kΩ type.
The dual-gang can provide a better channel separation than the single (use a separate GND wire for each section).
A single-gang potmeter of reasonable quality will give a channel separation of around 60 dB and I find that fully sufficient for a headphone amplifier.

Balance potmeter connection Schematic.
Fig.4: Connection of single-gang or dual-gang Balance Adjustment potmeter.

Volume Control.

Volume Control Schematic.
Fig.5: Volume Control.

This is a standard active volume control. For a detailed description see [1].
The maximum gain for this circuit is R123 / R117. The addition of R118 and R119 allows the gain to be increased by 8 dB or 16 dB.
C109 and R116 keeps the Gain Switch DC free. R115 isolates the OP-AMP from capacitive loading from the cable to the Gain Switch.
R120 is not used.
C110, R122 and R123 isolates the OP-AMP from capacitive loads while still maintaining a low output impedance at audio frequencies.
The cabling to the potmeter can be a little tricky - more on this later.

Gain Switch.

The gain switch can be 2 or 3-position switch suitable for audio.

Gain Switch Schematic.
Fig.6: Gain Switch Examples. LSA, LS1, LS2 are the PCB terminal names for left channel (RSA, RS1, RS2 for right channel).

Fig.6a:Using a 2-position toggle switch. These are the most economical switches that are available with gold-contacts.
Fig.6b:Using a 3-position toggle switch. This was how I initially made it, but it is not very intuitive with the low gain in the center position.
Fig.6c:Using a 3-position non-shorting rotary switch. This works very well, but the switch is very expensive.
Fig.6d:Using a 3-position shorting rotary switch. This should work (not tested), but the switch is very expensive. The 2.2 kΩ resistor must be mounted on the switch.

See Appendix B for a Gain Switch with relays.

Output amplifier.

Output Amplifier Schematic.
Fig.7: Output Amplifier.

This is a discrete design as I could not find a suitable integrated circuit.
The circuit is based on Douglas Self's "Blameless amplifier" design described in [2].
Q151..Q153 are 2 mA current sources. D152 keeps the current sources off until the voltage between VCC and VEE has reached approximately 20 V.
This prevents a large turn-on thump in the headphones if the amplifier is powered from a PSU with a voltage doubler rectifier where the positive and negative supply comes up at different times.
It also prevents large DC voltages on the output in case the negative or positive supply is missing (<100 mV into 600 Ω, <10 mV into 15 Ω).
Q158 and Q159 is a differential amplifier. They should be thermally coupled for low offset drift due to air-circulation.
Thermal coupling is easiest done by putting the transistors into a piece of 6 mm heat-shrink tubing.
Q163, Q166..Q167 is a current mirror. Q166 and Q167 can be thermally coupled. I was just watching the output with a scope, but I think it made a slight improvement.
Q164 and Q165 is the voltage amplifier and Q154..Q156, Q160..Q162 is the output stage.
Q168 limits the collector current in Q165 in case of overload. Without Q168, the full saturation current of Q165 would go into the load through the current limiter.
C158, C159 and R179 is a 2-pole compensation network. This gives a peak of 0.2 dB at 340 kHz, but increase the 20 kHz loop gain by 10 dB.
If you prefer a flat frequency response (with an increase of distortion at higher frequencies), short C158, change C159 to 330 pF and leave R179 out. I have not tested this.
D153 reduce the recovery time in case of overload.
Q157 is the bias generator and this should be thermally coupled to Q156 for best distortion performance.
All "transistor-pairs" (Q158-Q159, Q166-Q167, Q156-Q157, Q154-Q155, Q161-Q162) should preferably be from the same batch. If you buy them on (cut) tape they will be (and they fit directly into the PCB without lead-forming).
R153, R154 and R181, R182 is the power limiter described below.

Power Limiter Schematic.
Fig.8: Power Limiter.

The circuit has a large value resistor in the transistor collectors to limit the output power. This dissipates a lot of power and is only useful at low power levels.
When the output voltage and current increase, the voltage on QA's collector will fall. At some point QA will go into saturation and DA will clamp the drive current for the output stage into the load.

The values of RC and RD are:
RC = RD = √Rl * Rh, where:
Rh is the high impedance load and Rl is the low impedance load.
RC = RD = √16 * 600 = 98 Ω.
R157 // R158 + R164 + R167 ( Fig.8) should be subtracted from this value, giving a value of 91 Ω.
As it is difficult to get 91 Ω 3 W resistors in small quantities, I first tested the circuit with 100 Ω, but distortion at 100 mW into 15 Ω is too high.
Adding a 1 kΩ resistor across RC brings the value down to 91 Ω.

Maximum power dissipation ( PDRC ) in RC and RD is approximately:
PDRC = 15² / 100 = 2.3 W for the 100 Ω wire-wound resistor.
PDRC = 15² / 1000 = 0.23 W for the 1 kΩ film resistor.

Maximum power dissipation ( PDQA ) in QA and QB is approximately:
PDQA = ( VCC / 2 )² / RC
PDQA = ( 15 / 2 )² / 91 = 0.62 W.

Maximum current ( ICQA ) in QA and QB is approximately:
ICQA = 15 / 91 = 165 mA.
The power and current calculations above are absolutely worst case DC values into a short-circuit. For AC operation they can be divided by 2.
The transistor power dissipation and collector current is shared by 2 transistors in the actual circuit ( Q154-Q155, Q161-Q162 in Fig.8 ).

Common components.

Common Components Schematic.
Fig.9: Components common to both channels.

D1 and D2 are optional reverse voltage protection diodes.
H1..H4 are mounting holes.
C1..C4 are optional HF decoupling capacitors between the PCB and the chassis.
C5 and C6 are PSU decoupling. The point between C5 and C6 is the reference ground for the PCB.

OP-AMP decoupling.

OP-AMP decoupling Schematic.
Fig.10: OP-AMP decoupling.

The electrolytics and the ceramic capacitors are shown with 2 ground symbols.
The electrolytics goes to the midpoint between C5 and C6 while the ceramics goes into the ground-plane.

Bias adjustment.

This is best done using a THD analyzer, but can be done with a voltmeter.

Using a THD analyzer:
Connect a 15 Ω resistor between the LO and LG terminals.
Apply power and wait some minutes for the circuit to settle.
Set the gain switch to minimum gain, the balance adjustment to center and the volume control to maximum.
Apply 50 kHz between the LI and IG terminals.
Set the output voltage (between the LO and OG terminals) to 1 V.
Measure THD between the LO and LG terminals with the THD analyzer set to its highest bandwidth.
Adjust R171 for minimum THD+N (around 0.015% in a 400 Hz to 300 kHz measurement bandwidth).
Repeat for the right channel.

Using a voltmeter:
Make sure nothing is connected to the output.
Apply power and wait some minutes for the circuit to settle.
Measure the voltage between the TP151 and TP152 pads.
Adjust R171 for 2.0 mV.
Repeat for the right channel.
If you do not have a voltmeter that can display mV with a reasonable accuracy, it may be possible to measure the voltage across R154 and adjust this for 90 mV.
I have tried this with 2 different battery powered voltmeters and it works well.
I also tried with a mains-powered voltmeter and this caused the amplifier to oscillate, the reading on the meter to climb and R154, R182 starts to get warm.


This is the specification for the HA01FAA prototype.
Supply is ±15 V from a low-noise lab power supply (<-92 dBu (18 µV) noise in the 22 Hz to 22 kHz bandwidth).
The unit dBu is dB referred to 0.775 V.
Unless noted measurements are with the gain switch set to minimum gain, the volume set to maximum and the balance adjustment in center position.
More detailed measurements are in the file HA01F_Measurement.ods in the design file download.

Table 3: Specification for HA01FAA. Load impedance is 15 Ω to 600 Ω unless noted.
Supply current, no signal, no load44 mA
Supply current, 0 dBu, 20 Hz square-wave input signal, outputs shorted to GND230 mA ( 400 mA peak)
Input impedance, 1 kHz100 kΩ
Input impedance, 20 Hz..20 kHz>68 kΩ
Output impedance, 20 Hz..20 kHz1.1 Ω
Voltage gain, 1 kHz, 600 Ω load, gain switch set to low gain4.1 dB
Voltage gain, 1 kHz, 600 Ω load, gain switch set to mid gain12.7 dB
Voltage gain, 1 kHz, 600 Ω load, gain switch set to high gain20.6 dB
Balance adjustment range±5.5 dB
Volume control tracking, 1 kHz, 0 dB to -50 dB<1 dB
Volume control attenuation, 1 kHz108 dB
Linearity, 20  Hz..20 kHz, ref. 1 kHz, 50 mW into 600 Ω (Note 1)+0.05 dB / -1.9 dB
Linearity error, 20  Hz..20 kHz, ref. 1 kHz, 50 mW into 600 Ω (Note 2)<0.2 dB
Output clip level, 1 kHz, 1% THD+N into 600 Ω20.8 dBu ( 121 mW )
Output clip level, 1 kHz, 1% THD+N into 300 Ω19.7 dBu ( 187 mW )
Output clip level, 1 kHz, 1% THD+N into 120 Ω17.1 dBu ( 256 mW )
Output clip level, 1 kHz, 1% THD+N into 60 Ω13.9 dBu ( 247 mW )
Output clip level, 1 kHz, 1% THD+N into 30 Ω9.8 dBu ( 192 mW )
Output clip level, 1 kHz, 1% THD+N into 15 Ω4.9 dBu ( 123 mW )
Slew-rate>7 V/µs
Cross-talk, 20  Hz..20 kHz ( Note 3 )<-61 dB
THD+N, 1 kHz, 22 Hz..22 kHz BW,
10 mW to highest of 100 mW or 1 dB below clip level.
THD+N, 20  Hz..20 kHz, 10 Hz..80 kHz BW,
10 mW to highest of 100 mW or 1 dB below clip level.
IMD, SMPTE, 60 Hz / 7 kHz, 4:1,
10 mW to highest of 100 mW or 1 dB below clip level.
Maximum capacitive load ( Note 4 )100 nF
Output noise, 22 Hz..22 kHz, input shorted, gain switch set to low gain-102.1 dBu ( 6.1 µV )
Output noise, 22 Hz..22 kHz, input shorted, gain switch set to mid gain-96.4 dBu ( 12 µV )
Output noise, 22 Hz..22 kHz, input shorted, gain switch set to high gain-89.3 dBu ( 27 µV )
Output noise, 22 Hz..22 kHz, input shorted, gain switch set to high gain,
volume at -6 dB
-94.7 dBu ( 14 µV )
Output noise, 22 Hz..22 kHz, input shorted, gain switch set to high gain,
volume at minimum
-107.9 dBu ( 3.1 µV )
Output noise, A-wgt, input shorted, gain switch set to high gain,
volume at minimum
-110.9 dBu ( 2.2 µV )
Board size (length / width / height)115.6 mm / 119.4 mm / 15 mm

Note 1:This is the HP-filter roll-off.
Note 2:0.2 dB is at 20 Hz. The value for 50 Hz to 20 kHz is <0.05 dB.
Note 3:This is cross-talk in the balance adjustment. Without the balance adjustment this value is <78 dB (<103 dB @ 1 kHz).
Note 4:I have not found any load that will make the amplifier oscillate, but with capacitive loads between 100 nF and 390 nF there is up to 20% overshoot on a square-wave.


HA01FAA photo.
Fig.11: Photo of mounted board..

Download HA01F design files.

I have boards available for this project. See the PCBs page.

Known Issues / updates.

No known issues.

Appendix A: Board terminals and wiring.

Table A1: HA01 terminals.
Terminal nameGoes toFunction
VCCPSUPositive supply ( +15 V ).
VEEPSUNegative supply ( -15 V ).
IGInputGround for input connector.
LIRIInputInput signal.
LGRGOutputGround for output connector. Use either for 3-wire connections.
LOROOutputOutput signal.
LBGRBGBalanceGround for balance potmeter. Use either or both for 3-wire connections.
LBEBalanceSignal to balance potmeter top (CW position).
RBEBalanceSignal to balance potmeter bottom (CCW position).
LVTRVTVolumeSignal to volume potmeter top (CW position).
LVARVAVolumeSignal from volume potmeter wiper.
LVBRVBVolumeSignal to volume potmeter bottom (CCW position).
LVSRVSVolumeGround. Can be used for cable shields to volume potmeter.
LSARSAGain switchSignal to gain switch wiper.
LS1RS1Gain switchSignal from gain switch for mid gain.
LS2RS2Gain switchSignal from gain switch for high gain.
LSSRSSGain switchGround. Can be used for cable shields to gain switch.

External components wiring Schematic.
Fig.A1: Wiring of external components.

Balance Adjustment: This can be wired with unshielded wire.
If you use a dual-gang potmeter, LBG and RBG should not be connected together at the potmeter.
The prototype was wired with shielded hook-up wire where the shield was used for GND.

Volume Control: The connections to LVT/RVT are not critical and can be unshielded.
The connections to LVA/RVA are high-impedance and should be made with shielded wire.
If you use unshielded wires, they should have a reasonable distance to the signals on LBE/RBE, LVT/RVT and to signals in the opposite channel.
The connections to LVB/RVB should have a reasonable low resistance for best volume attenuation.
The prototype was wired with shielded wire for LVT/RVT, LVA/RVA. The shields was used for the LVB/RVB signals.

Gain Switch: This can be wired with unshielded wire.
The prototype was wired with shielded wire for all 6 connections. The shields was connected to the LSS/RSS terminals at the PCB and left unconnected in the other end.

The shielded wire used for the prototype was single-conductor 0.06 mm² with braided shield. 25 cm long wires was used as I expect this is maximum required for any practical application.
Unshielded wire above can be 0.25 mm².. 0.5 mm².

Photo of Balance Adjustment wiring.
Fig.A2: Balance Adjustment wiring.

The potmeters here are for PCB mounting and their terminals are not as rugged as those designed for wire connections.
A piece of strip-board distributes the mechanical load from the wires over all the terminals.
The wire-shields are cut shorter than the inner conductors so they act as a strain relief.
The 2 cables are held together with a piece of heat-shrink sleeving. This sleeving is a cheap type from a web-shop. It is fine for this application, but do not use it for insulation. The break-down voltage is <500 V (the lowest setting on the insulation-tester).

Photo of Volume Control wiring.
Fig.A3: Volume Control wiring.

Photo of Gain Switch wiring.
Fig.A4: Gain Switch wiring.

Here the shields are connected to an unused terminal on the switch to act as a strain relief.

Appendix B: Gain Switch with relays.

Relays are available with excellent contact performance at a reasonable price (less than 5% of a good rotary switch). An example is TE Connectivity 3-1393789-7.
The 2 circuit below can use a any low-cost, 3-position switch and can easily be built on a piece of strip-board.
Note that these circuits has not been tested.

Relay Gain Switch Schematic.
Fig.B1: Gain Switch with relays.

The operate voltage for 24 V relays is typically around 18 V, so there is no guarantee that the switch is in the correct state when the amplifier "un-mutes" at around 20 V. This may cause a small thump in the headphones at power-on.
Even with a mechanical switch the diodes across the relay-coils are NOT optional. I have seen relays flash over from the coil to the contacts without the diode.
The zener diode prevents sudden load-changes from the PSU when the switch is operated.
The supply current is around 7.5 mA with the components shown.

Relay Gain Switch Schematic.
Fig.B2: Gain Switch with relays.

Using 12 V relays and a constant current source, the relays will be in the correct state shortly after the supply voltage has reached 10 V, so the turn-on thump should be avoided.
The supply current for this circuit is around 15 mA with the components shown.


[1] Texas Instruments: Active Volume Control for Professional Audio.
[2] Douglas Self: Audio Power Amplifier Design.
An excellent book on power amplifier design, but quite expensive.

Poul Petersen, C/Faya 14, 35120 Arguineguín, Las Palmas, Spain.
Poul Petersen home, Poul Petersen DIY index, E-mail: diy@poulpetersen.dk
Copyright © Poul Petersen 1990 - 2017. Last update: 20170729. Valid HTML!