Headphone Amplifier


The use of headphones for music listening at home is completely unnatural and ridiculous. There’s no proper soundstage and what ought to be perceived as coming from front/centre seems to float above your head. It’s just plain hideous and people who go on about their wonderful “cans” simply have no idea of what good hi-fi is all about. They have a place elsewhere such as on public transport as long as used at a low levels so as not to disturb others, but don’t waste money on the kind of crap on this page. I happened to have the PCB from a “past life” in Audiofoolsville so I built this out of it rather than throwing it on the tip. It has not actually been used beyond testing:

Finished headphone amp


Junkpile chassisThis is a “re-boxing with improvements” of something I put together poorly back in 2003.  It’s a “QRV-03” supposedly ultra high performance current feedback amplifier with a transimpedance gain stage.  It’s based on a Mark Alexander circuit by a PCB designer called Per-Anders Sjöström.  →

Recently I opened the original chassis to have a look and was embarrassed by my poor chassis layout and idiotic inclusions like that dumb aluminium box encasing the cheap EI transformers for reasons I can neither recall nor validate.  There were ground loops, bad grounding practice and a dangerous fuse connection.

Fortunately the PCB had screw terminals for all connections, so it was an easy enough pull to start over.  Here it is:


It’s a seriously OTT circuit, so I built it without any modifications.  Here are the claimed specifications (adapted here for my headphones):

 Frequency response (with input filters)  1.5Hz – 338kHz  (-3dB)
 Equivalent input noise  100nV
 Signal to noise ratio  137dB
 Dynamic headroom  157dB
 Slew rate  60V/µs
 Step response  Perfect without overshoot
 Gain  4(12dB)
 Input impedance  100kΩ
 Output impedance  3mΩ
 Damping factor  192,000 (with 600Ω load)
 Output current  1A peak
 Supply voltage  ±18V

Whoopee do eh!  I can’t even hear that well and suspect than no amplifier for mere audio needs specifications like that!  A laboratory instrument maybe.  Dual power supplies, dual transformers etc., etc.  If it’s to improve channel separation I don’t get it.  The headphones share the same lead anyway!  What the heck  – I stuck with the theme since two small toroidal transformers fitted nicely enough into the small aluminium chassis that I had in mind.


The plan is to use the amplifier with a 30 year-old pair of Sennheiser HD-230 headphones:

Sennheiser HD-230

They’re a German-made 600Ω sealed design with separate high and low frequency drivers.  It seems that they’re quite rare.  These are perhaps the only photos of a pair on the internet (apart from the one in the owner’s manual archive).  They can sound pretty decent:

Behind the foam


The PCB has a jumper to short across a 100Ω series output resistor thereby providing the stated low output impedance (practically zero Ω).  The jumper is in place, but can be lifted to flab-out the bass if wanted (depending on the headphones).

Zero output Impedance risk?

With nearly zero Ω output impedance it was suggested to me that there might be a risk of damaging the amp just by inserting or retracting the headphone jack while music is playing!  After testing the jack/socket interaction, I found that it’s only the right channel that can be shorted.  The ring and sleeve terminals of the socket are momentarily shorted by the TRS jack.  The Neutrik locking type headphone socket will be a good reminder to turn the power off (or at least turn the volume down) before plugging/unplugging the headphones.  Thank you Mr John for alerting me to shorting TRS sockets!  I didn’t know that.

The Rebuild

First thing was to check all the solder joints and attach a current-limiting resistor to one of the power supplies for a front panel “power” LED:

Complex dual layer PCB

Those nuts are on plastic screws which I used to ensure that the heat sinks can’t touch the upper surface of the PCB or nearby components.  The original build had a 250V neon power indicator, but that was a bit dumb.  Here I just took unregulated DC from a +ve capacitor leg in front of a regulator of one of the power supplies.  With 18V transformers, that’s around 25.5Vdc, so a 3K3 0.75W resistor was used to feed the 2V LED with about 7mA.

Progress pics:

VT4C Chassis etc.

It’s a familiar generic aluminium chassis from VT4C in Hong Kong.  A couple of small Nuvotem resin-encapsulated transformers (15VA, 2 by 18V), a fuse-holder through the floor, an Alps 100K RK27 stereo volume pot recycled from the old chassis, the locking Neutrik headphone socket etc.  The mains EMI filter is a low current hospital type (whatever that means).

Volume control DC protection

While there are on-board series capacitors at the PCB inputs to block any DC, these do not protect the up-front volume control pot which would produce scratchy noises if there was DC on the input, so the build needed additional series capacitors.  Here they are at one end of the input harness (spare 1μF PPs from the junk box):

Input end of harness

Here’s the overall harness using Teflon coax (chosen purely for its melt-resistance to soldering and nearby heat sinks):

Whole input harness

With some tricky soldering to the pot:

Wiring to pot

All the shields are tied to that single wire extending from top left.  That’s a ground strap which connects to the star grounding terminal of the PCB.  The shields of the coax do not connect to the PCB at all.  They are cut and covered with heat shrink.  Installed:

Pot end of input harness

Input end of harness

Signal input to PCB

Coax shields unterminated this end.  Input gnd wire from the pot is attached to the middle terminal (see star grounding trace on PCB).  Similar treatment at output:

output harness

Output header

Here the coax shields are not connected to the TRS socket.  A separate ground lead runs from the Sleeve terminal to the PCB header (second from left).


Well it works and I observed the following on my scope:

Sine wave test: Apparently dead flat with headphones connected from around 15Hz to around 45kHz.  Test point taken at output headers.

Square wave test (irrelevant):  My cheap generator cannot produce proper square waves.  They have square leading edges, but rounded trailing edges (I must look into that).  However, the output of the amp looks exactly the same as the input over the same range as the sine wave test.

I might see if I can loop the thing through my PC sound card and run an impulse response measurement through it later to graph its performance.

To be used where?

It’s to be attached with long inteconnects to the “Tape” output of an old C-J valve preamp.  The preamp needn’t be turned on, but whatever is selected by its input selector is fed through.

That's where it goes


A little mod (for big safety):

During previous testing I had mistakenly thought I heard a ground loop hum.  As it turned out it was just a loose connection between the tone generator and the input RCAs.  It vanished when the plug was more firmly inserted, so there is no hum problem.  But it got me thinking about whether a headphone amp ought to have a safety ground loop breaker.  Every power amp that I have ever built has one, but none of the “line level” projects (etc.) do.

Anyway I assembled another high current safety loop breaker.  Here it is:

high current safety ground loop breaker

It’s the same as those installed in my power amps – it’s from the ESP site (the 100nF ceramic cap just suppresses RFI).  Here there is no loop problem to start with.  Moreover, the so-called “breaker” will actually create a minor loop by inserting 10Ω of resistance between the “zero volt line” (tranny centre taps/PCB gnd plane) and the chassis/mains Earth where previously there was infinite resistance.  Is it needed?  Not for breaking any loop – that’s for sure.

BUT!  The headphones are gonna be on my head!  What if one of the trannies develops a fault that shorts the primary side to the secondary side?  That’s mains voltage at my head! 😯

“So just strap the PCB gnd to the chassis”.  No!  That could create a real loop problem.

The 35A bridge rectifier provides a low impedance/high current path to the mains Earth when anything over about 1V is seen across its paralleled diodes.

Nobody tells you this stuff!  Anyway, there it is installed:

ground loop breaker installed

The fly leads from the trannies are actually quite feeble – much lower in cross-section than the PCB ground plane mask, so the solid PCB centre terminal connection of the white loop breaker wire provides an adequate current path from the secondaries.

Yeah – there are fuses in there and interconnect shields which connect back to the pre-amp (sometimes). 😛

Moving on…

Headphone protection:

Most of the power amps that I’ve built have output muting relays controlled by an ESP P33 to prevent fault DC from reaching the drivers.

Prior to installation of the protection system, this amp sent a little spike/click to the headphones when power was applied or removed and it sounded a bit crummy, so if just to eliminate that, P33 is a worthwhile addition because it provides a muting fuction as well.

By good fortune, I set out the chassis in such a way that the P33 PCB and a muting relay fitted so nicely that they look like part of the original idea. 😳 It need lots of connections to the main PCB:

P33 connections

P33 is intended to control two relays in a stereo power amp, but they both connect to the same terminal and activate together when fault DC is detected on either channel.  Here a single DPDT relay works exactly the same way.  And with around 25.5Vdc available, a standard 24V coil relay would probably be OK without a voltage-dropping resistor.  Nonetheless a 47Ω resistor is being used.  This 10A relay is ideal since it has a high 650Ω coil for very low current draw (37mA):


Here it is installed with modified wiring to the output socket:

Fiddly wiring

Bloody brilliant eh?


The P33 controller PCB:

Here is the populated P33 PCB:

P33 for headphones

Some of the resistor and capacitor values had to be changed for this low-Voltage application (R1 and R2 = 18KΩ, R6 = 39KΩ, R8=100KΩ, C1 and C2 = 47μF).

The 47Ω relay Voltage-drop resistor is under the PCB and makes useful an otherwise unused pin of the relay terminal block:

relay volt drop

In and working fine:

P33 installed

The completed amp:

Headphone amp done

All good. 😀


With a 220Ω resistor across an output, it just started to clip at 30V P-P sine (about 10.6V RMS) which is about 500mW.  The clipping was very ugly/hard with inverted triangles at the peaks and troughs of the wave.

Waiting for some cables to arrive before measuring the response with the REW software…

OK here it is:

QRV-03 distortion

That’s frequency vs. SPL with the 220Ω dummy load.  I guess much of that miniscule distortion is the PC sound card.  The red line at the top is the frequency response!  Addendum:  A test with a loop-back wire showed that nearly all that distortion (apart from those 50Hz/100Hz etc. power supply blips) was the sound card.

And here is the phase (basically 180º inversion across the board):

QRV-03 phase

Actually, the sound card might be inverting the signal.  I will check with a loop-back cable later, but it makes no difference to the sound.  Addendum:  Yes – this is all due to the sound card.

OK, what about the THD?  Checked it and it does of course vary with frequency.  At 1kHz it is 0.026%.  I know that’s not the proper way to quote it, but it’s the best I can do.

Later, I’ll try to repeat the measurements with some headphones attached instead of the purely resistive load.  Need to get some connections happening first.




One thought on “Headphone Amplifier

Leave a Reply

Your email address will not be published. Required fields are marked *