Gio, a MAKE subscriber from Winnipeg, Canada, sent us a link to this well-documented low voltage headphone amp project. The tube-based amp uses a single 12AU7 tube for voltage gain and a IRF510 MOSFET to supply current to drive the ‘phones. The small hybrid amplifier operates off a 12V SLA (sealed lead-acid) battery, so there are no high voltage concerns. An LM317 regulator was used as a constant current source to bias the MOSFET into class-A operation. The amp can be built for about $40-50, less if recycled parts are used.

(Psyched to see the builder using the Grado SR60 headphones, my all-around, all-time faves.)

NP-100v12: 12AU7 (ECC82) / IRF510 Headphone Amp

Gareth Branwyn

Gareth Branwyn

Gareth Branwyn is a freelance writer and the former Editorial Director of Maker Media. He is the author or editor of over a dozen books on technology, DIY, and geek culture. He is currently a contributor to Boing Boing, Wink Books, and Wink Fun. And he has a new best-of writing collection and “lazy person’s memoir,” called Borg Like Me.

  • Ross Hershberger

    I work with tubes all the time. I really admire the neat execution of this project. I’d question the choice of running the 12AU7 with a 12V B+ and 5K load. That’s really too low for good fidelity. A clean operating point for this tube for headphone voltage amp use is about 200V B+, 33K plate load and 3ma (100V plate voltage). This tube is not designed to run at low voltages and will have high distortion at low plate volts. The 6DJ8 is better at low voltages and might be acceptable on a 12V supply, but for best performance I’d want at least 24VDC on the plate (48V B+)

  • Ross Hershberger

    The IRF510 MOSFET is an ideal device to drive low impedance headphones as a source follower. Some applications use the 150ma tube heater as a resistive load, but that wouldn’t work in this case.

    The Grado SR-60 has been my favorite budget headphone for 20 years. When the earpads fall off, replace them with the more comfortable yellow Sennheiser HD-414 earpads ($5 from Sennheiser).
    This week I got a pair of Stax electrostatics, the SR-202 phones and the SRM-212 amp. If you don’t mind 580 volts clamped around your head these are pretty sweet.

  • Ross Hershberger

    The article makes reference to grid leak bias. I think this is mistaken. Grid leak biased tubes have their cathodes grounded and a large value grid resistor. Electrons hitting the grid cause it to charge up negative, biasing the triode. This circuit has a 100K grid resistor, which is at least an order of magnitude too small to allow charge buildup. The bias in this circuit is provided by the cathode resistor. The cathode current flowing through the resistor forces the cathode positive W/R/T the ground referenced grid, giving negative bias between the grid and cathode.

  • Gareth Branwyn

    Interesting stuff here, Ross. I’ll ping Roggom and see what he says to all this.

    That’s great advice on replacing the Grado pads. Mine are pretty worn down (I’ve probably had mine for a good 15 years). We used to have a set of Sennheisers and I remember those pads. I wondered what they’d look like on the Grados and then I found this:


    Not bad.

  • Ross Hershberger

    I got my mom a pair of SR-80s (bigger than SR-60, with more bass). When she complained about the comfort I ordered a few pairs of the HD-414 pads. She’s delighted. The foam is much softer.

    W/R/T this circuit, I think I’ll lash it up on the bench and measure the distortion. It’s the only way to know for sure. Should have results within a day. The parts are all within arm’s reach.
    If possible I’ll test it with 6DJ8/6922 triodes in place of 12AU7s to see if that makes a difference.
    I just finished testing thirty seven Type 45 vintage triodes from my stash. I haven’t used a 45 in years and have to get these out of here…

    • Gareth Branwyn

      That would ROCK, Ross! I’d post an update for sure!

  • Ross Hershberger

    Actually, in the case of this circuit you could use the tube heater as a resistive load for the MOSFET. Ground the heater CT (9) and use one half of the heater (pin 5 or 4) as the MOSFET load for each channel of the amp. You get 150ma through the MOSFET, which will work fine. The output stage current heats the tube. That eliminates the heater draw from the PSU and eliminates 2 LM317 regulators. The downside is that the LM317 configured as a current source has a very high AC impedance, optimizing the gain of the follower stage. The filament half has a resistance (hot) of only 40 Ohms, which isn’t a lot of impedance for the MOSFET to swing current through. I dunno what the MOSFET’s Z is at 150ma, but it’s the inverse of the transconductance, basically.

  • Ross Hershberger

    I mocked the triode portion of this circuit on the bench and ran some tests. Here are the results:

    Test level = 300mv AC out
    Signal input = 70mv AC
    Gain = 4.28
    THD @ 1Khz = ~1%, primarily second order.

    I’m surprised that this works at all, but it does. 1% of 2nd order THD will ‘warm up’ the sound a bit, giving it an authentic tube sound.

    The THD might have been lower at lower output voltages, but the sensitivity of my THD analyzer is low so I picked that level to get full scale on my lowest sensitivity setting.

    The gain is rather low. Run at a more conventional operating point the amplification factor would be more like 10 – 15 rather than 4.28. That might be plenty of gain for a headphone amp.

    Technically the THD is high and would not be acceptable for a high fidelity circuit, where you’re looking for distortion a couple of orders of magnitude lower, but in a headphone amp the relatively benign 2nd order distortion might be pleasant or even desirable.

    This circuit could be built with one of the commonly available surplus 48VDC power supplies. If a current source was used as a triode plate load in place of the resistor, the triode would see most of this voltage at its plate. The resulting distortion, once the operating point had been adjusted with a different cathode resistor, would be about 10 db lower.
    With a 48V supply the MOSFET section would have to be reworked to get the operating point right as well.

  • Ross Hershberger

    I ran this circuit with a 12AT7 in place of the 12AU7 because I know the AT works well in some low volts applications. The distortion went down a lot – to a measured 0.4%, and that measurement included a lot of noise and hash because of my ad hoc lashup. The real distortion was probably 0.2% in a ‘clean’ circuit. That compares favorably wit hteh 12AU7’s 1% THD. The AT7 biases up at exactly half of the B+ voltage, +6 volts. The big difference was the gain. Rather than the 12AU7’s 4.28 gain the AT7 gave a gain of 20. It took only .01V of input to get 0.2V of output. This is one of the reasons for all of the noise on the output. The AT7 does have a higher design amplification factor (mu) so this is to be expected.
    I’m not sure the extra amplification factor is important for a headphone amp, but there it is.
    I’m going to try this again with a different 12AU7 and see if I get different results before I tear the circuit down.

    • Todd

      I’ve been wanting to build a tube headphone amp for a while now. When I saw this circuit I figured it would be a good place to start. Would you suggest using the 12AT7 instead? From your results it looks like the way to go. Is there anything that needs to be done differently, or can you simply substitute the 12AT7?


      • Ross Hershberger

        The 6DJ8/ECC88 works well at some low voltages. 12V is pretty seriously low but it might still be OK. I’ve been looking at a lot of tube/MOSFET headphone amps lately. Many use voltages around 50 – 70V
        The 12AT7 has pretty high distortion. I’m using it in a new tube power amp circuit I’m developing for publication specifically because its distortion somewhat cancels out the distortion of the output tube.

  • http://mofact.org/wp-content/temp/ air max pas cher

    thanks for almost all information about unseen aol checker’ nevertheless does anyone understand how to determine which usually askjeeve chat room somebody is actually on the web?
    air max pas cher http://mofact.org/wp-content/temp/