Slippery Slopes Cognitive Event Horizon Reef Madness
navigation bar

Just another OTL Headphone Amplifier

It started out as a replacement headphone amp (you know, for work...) and got lost along the way. Current version is about the 6th circuit and sounds pretty good - I tried several variations of tubes and parallel feed using cheap output trannys, but couldn't get things to make music. The current version makes music.
If you are interested in building a Headphone amp you really have to go over to the headwize site and check out all of the really groovy designs and discussions on this interesting sub-universe of amplifier design:
Still here?
OK, building a headphone amp is exactly like building a low-power integrated amplifier, except for the added complications of needing extremely low noise (good headphones will ruthlessly reveal noise, hum, grunge, whatever) and needing to support a wide range of output impedances since headphones vary all over the map. Diving a pair of 300 ohm Sennheiser HD600 'phones is a very different load from a pair of 32 ohm Grados or Sony cans.
Fortunately, I like Sennheisers and have a pair of HD580 cans, so I was able to focus my design for their higher impedance - Note: my amp is not tested with low-impedance 'phones - no warranty expressed or implied, don't blame me...
Earlier ideas (Yup, I did build and test all of these, but sorry, no schematics of these available):
  • C4S loaded 6N1P driving paralleled 6N1P parafeed via C4S into PP output transformer
  • C4S loaded 6H30pi driving paralleled 6N1P parafeed via C4S into PP output transformer
  • C4S loaded 6H30pi driving 12B4 parafeed via C4S into PP output transformer
  • C4S loaded 6N1P driving 12B4 parafeed via C4S into PP output transformer
  • C4S loaded 6H30pi cap coupled to 12B4 C4S sunk Cathode Follower OTL
  • Resistor loaded 6H30pi cap coupled to 12B4 C4S sunk Cathode Follower OTL

I used the 6N1P and 6H30pi becuase I've A) used them before, B) had a stock on-hand, C) like their sound. The 12B4 was a well-liked tube from online discussions with a low plate impedance and good power range. All of these tubes are very durable and should last a long time in the schematic shown below. I was really impressed with the sound of the 12B4, and plan to use it again if possible - BTW, it's also a very attractive tube when illuminated.
Headphone Amp
Headphone Amp: C4S loaded 6H30pi voltage amplifier into C4S sunk 12B4 cathode followers
Frame is paduak and birdseye maple with an aluminum top. Tall cans are Target kitchen department covers for output transformers bypassed in current design.
6H30pi-12B4_jul02 Schematic
The final hurdle was getting rid of the cap coupling between the 6H30pi voltage amplifier and the 12B4 cathode follower - making this amp into a really big 'ol Foreplay derivative (sorry Doc).
Just to give folks and idea that this is hobby is not all instant gratification, here's a note I submitted to Bottlehead Forum on 5jul02:
OK, so I've got this little headphone amp (you know, for work...) I've been fiddling with for about 6 months.
It uses a 6H30pi as a voltage amplifier (10kohm Rp, 4.3V bias, 15mA) cap coupled via a .47uF Angela tin foil/PP film cap to a 12B4 cathode follower (C4S loaded, 12mA), OTL output, 1kohm load across output jack, 75uF (combo of 3 film/foil cap types). PS is UF4007 FWB - CLCLC 250VDC with 1.3mV hum at last cap. Filaments are DC, Schottky rectifiers, CRC, 6.1V...
I previously was using a C4S to load the 6H30pi, but switched to Caddock 816 resistors (big heatsinks, they still get warm) because of problems with my much-transplanted set of C4S boards. When I fired up the thing I got no glow on the CF C4S boards until I touched the probe of my multimeter to the grid pin on each socket (note, CF is cap-coupled on both input and output) - then things worked wonderfully, and the music flowed!
After wondering about this odd C4S behavior I realized that the meter was providing a ground reference for current to flow, and then the C4S was kicking in; but this was not acceptable for normal powerup - and yes, it did repeat when I shut down and restarted. Another symptom was a very slow rise over time of the output voltage I attribute to leakage across the coupling cap. So I installed a voltage divider from the B+ to the 12B4 CF grid to the ground using Caddock MK132 (good quality metal films I happened to have in the parts bins in these values: B+:221k:grid:68k:GND) that holds the grid at 60VDC and the thing starts reliably 100% of the time.
The problem is this seems to have sucked the life out of the sound - It was much livelier without the voltage reference.
So, anyone have an idea for a next step? Grid choke to GND? 1meg ohm R to GND without the B+ reference? Output of the 6H30pi is around 98V which is above what tubeCAD is saying is a good input for the CF, but then tubeCAD doesn't have a constant current source option either.
Whatever, this current sound will not do.

Doc B. Kindly replied:
What's the Rp of the 6H30pi? Can't you just dispense with the CF altogether and couple the cans right to them? If not, it seems like maybe reworking the PS so you can direct couple the stages might be a neat way to go.

Aha! I repost:
Sometimes the process of posting makes things fall into place in my head (sometimes I just get more confused), but I'm now thinking that at one point before I added the resistor networks to the CF grids I had been running the thing overnight and the leakage had driven the grid up to 117VDC (you have to be quick to even see it as the process of measuring it made the voltage drop immediately). The output voltage was happy at almost 100VDC. Maybe I should just snip out the coupling cap and voltage divider, and let the C4S ride at a higher voltage and be done with it. The 12B4 has plenty of headroom. Simpler is better.
I'm afraid to take pictures of the inside of this thing, with about 6 major circuit/tube combos (several iterations attempted parafeed into some cheapy XFMRs) there are signs of wiring and soldering distress all over. At least the outside looks OK.

The next day I reply to Voltsecond:
OK, performed more surgery last night - removed the coupling cap and voltage reference dividers and just directly connected the plate of the 6H30pi driver to the grid of the 12B4 and let the C4S ride and do it's current sinking thang.
Sounds much better without the extra complications and the 12B4 is happy with the grid at 97VDC instead of 60VDC. I was shocked at the sacrifice to the music when I put in those 68k resistors - an interesting lesson. The Dowdy-lama once told me to avoid any resistor over 20k in a circuit - they are just bad news. Food for thought.
Had not worked with the 12B4 until now, an impressive little tube; as a former TV sweep tube it is durable and will take a huge voltage (550 volt rating) without complaint. They sound good and are relatively cheap.
VS, I'm using the 12B4 as a CF in OTL mode for now and it works fine - if at some point I get caught up on my other spendy projects I might take a run at a pair of 'proper' headphone output trannys - and switch the 12B4s to VA drive (or maybe try grounded plate drive?). Thanks for the suggestions.

I'm still interested in trying to build a transformer output headphone amp, but I've got some other projects to finish first, and this is good enough to enjoy at work for a while...
Finally, more links: Another Discussion site on Headphones
Bottlehead Doc B.'s site and the home of the Bottlehead Forum
VoltSecond This guy is relentless - some Headphone stuff in the Foreplay links Another good Discussion site