F5 Headamp ?

There are already umpteen versions of the F5, starting from the Selectronics Profet, then F5, F5X, F5T, F5 with 2SK2013/2SJ313, High-Voltage F5X for electrostatic headphones, ……. So why add another one ?

There have been various posts vaguely mentioning a downscaled version of the F5 as headphone amp. But I could not remember / find a proper schematics for precisely that. Since there have been recently some discussions about a discrete Pass type headphone amp, I thought I should add yet another one.

This is purely a lunch-time mental jogging exercise. That means, translated in English, that I have not built, nor do I intend to in the near future. But I think it is so easy and simple, that not much can really go wrong. So I leave here the Spice files as well as a Vero-board layout, so as to encourage those fearless DIY’ers to make a real attempt.

Performance in simulation is excellent. Even at 14V pk-pk into 50 ohm load, distortion is less than -80dB. Frequency response is equally impressive. Bandwidth over 1 MHz, and just a touch of overshoot with a 100n pure capacitive load (which is of course unreal).

And other than the input JFETs, the rest are easy to source components. You do need matched devices if you want low distortion.

Straightly speaking the feedback network is more Profet than F5.
The original F5 feedback network will however be required when using low Gm JFETs (e.g. 2SK246/2SJ103).



  • F5 Headamp IRF THD.asc
    4.7 KB · Views: 973
  • F5 Headamp IRF Freq.asc
    4.6 KB · Views: 433
  • F5 Headamp Vero.pdf
    369.3 KB · Views: 2,657
Last edited:
Let's mix it up somewhat and save those Toshiba JFETs, without comprimising SQ.
P1/P2 set the Iq of the output stage to about 50-100mA as well as DC offset at the output (start adjusting with maximum resistance of the pots).


  • ha-F5-BJTin.gif
    7.9 KB · Views: 16,010
  • Like
Reactions: 1 user
Well if we go BJT then there are hundreds of circuits, e.g. the JLH headphone amp.

But Nelson is known for FETs, so I considered it appropriate to pushing a FET circuit.
And you can of course use LSK170/LSJ74.

The 2nxxxx parts will cause more distortion.
So one needs to jack up the OLG somewhere else ......

I have been asked by PM :

> Can i use this circuit as a Line preamp to drive my Aleph j power amplifier ?

Yes, but you probably do not need so much current driving capability for a line level preamp.

> How can i null the offset voltage and bias current for the out put devices ?

By adjusting the values of R3, R4.

> what kind of head phones can i use with this circuit ?

In principle anything from 25 ohm to 300 ohm, except for a couple with really miserably efficiencies.
Would have no problems with a HD800, HE500 or K701, to name a couple.

In response to post #3, here are Spice files for the Fairchild JFETs.
As expected higher distortion, despite the lower Idss and the correspondingly increased values of R3 & R4.

One should also note that the Fairchild TO92 devices are also now obsolete.
One can continue to use the SOT23 devices (MMBF5457/5460).
Dissipation should be OK.

And still a good, low cost project.




  • F5 Headamp IRF FAIR THD.asc
    4.6 KB · Views: 387
  • F5 Headamp IRF FAIR Freq.asc
    4.5 KB · Views: 265
Weather was miserable on Sunday.
So instead of going out for a drive on my Boxster, I built a pair of quick vero protos with what I had in my drawers.
All as posted except for the gate stoppers where I now used 1k.
(JFET drain resistors need trimming of course to set bias & DC offset.)
This ensured stability with no load, and -3dB bandwidth is still 2MHz (smooth 1st order).
Power supply was just my lab supplies, +/-15V 330mA for 2 channels.
You can use LiPo batteries if you do not listen for too long.

How does it sound ?
Fast, clean, have plenty in reserve, background dead quiet.
I'll let someone else listen and give you his subjective review. :)
For what it takes to build, this will beat a lot of other headamps around, discrete or not.
Oh, DC drift within +/-20mV because of very good heat sinking (~30°C).
So no servo needed.

Photos to follow once the weather is good enough.





  • 150330 F5-HA Freq 68R Load.jpg
    150330 F5-HA Freq 68R Load.jpg
    91 KB · Views: 645
  • 150330 F5-HA 100kHz Sq 68R Load.jpg
    150330 F5-HA 100kHz Sq 68R Load.jpg
    81.6 KB · Views: 6,719
  • 150330 F5-HA Freq No Load.jpg
    150330 F5-HA Freq No Load.jpg
    88.2 KB · Views: 6,879
  • 150330 F5-HA 100kHz  Sq No Load.jpg
    150330 F5-HA 100kHz Sq No Load.jpg
    82 KB · Views: 7,229
As promised.

As I said, only a quick & dirty build to prove circuitry.
Took 5 hours from scratch to first sound.





  • 150331 F5-HA Top.JPG
    150331 F5-HA Top.JPG
    253.6 KB · Views: 1,547
  • 150331 F5-HA Btm.JPG
    150331 F5-HA Btm.JPG
    239.7 KB · Views: 2,176
  • 150331 F5-HA Module 1.JPG
    150331 F5-HA Module 1.JPG
    287.8 KB · Views: 2,546
  • 150331 F5-HA Module 2.JPG
    150331 F5-HA Module 2.JPG
    243.8 KB · Views: 2,236
Last edited:
Each channel of the F5-HA requires at least 100mA, preferably 120~150mA.
The minimum voltage you would need is about +/-12V (the Vgs of the MOSFETs eats up ~4.3V).
Anything less you are not operating the circuit properly.

With a 50 ohm load, it can output 10V 200mA and still has low distortion (120mA bias min.)
That would be 1W rms.
If you want to use a 30 ohm load, then I suggest increasing bias to 150~180mA.
If you have a 300 ohm load, you can increase voltage to not more than 20V.
But the hungry phones for which this is designed are all 50 ohm or below anyhow.

The F5-HA already has JFET input and hence high input impedance (essentially same as a complementary JFET source follower).
There is no need for any additional buffer (of which the so called B1 is on of many).

In my listening test, I use a 10k stereo pot between DAC out and F5-HA.
No problems whatsoever.