Headphone Amps
I'm a sucker for headphones. Speakers can be OK as long as they're good and located in an acoustically sound room. Most of the times they're not, including my own speakers. So most of the time I listen to my trusty Sansa Clip MP3 player. The earbuds they came with are surprisingly good but sometimes I need a little more oomph.
Enter OP275.
This is a dual audio opamp that will happily drive 600 ohm loads. My aging Beyerdynamic DT990 is just that. So I concocted a little board with a 5 to +/-15 volt converter, some resistors, two supply filter capacitors. I plonked it in a tin mint boxie (Curiously Strong Mints), added a LiIon 2 Ah laptop battery cell, a switch, a LED, 1/4" headphone socket and a short cable with 3.5mm plug for my MP3 player. The opamp is a straightforward non-inverting configuration with 10k and 1.0k precision feedback resistors, resulting in a gain of 11 times. The input is loaded with a 270 ohm resistor so the Sansa's capacitorless output (AS3525 chip) is happy. Done.
The results are excellent. The Beyer has a nice drive and is very loud before clipping sets in. Until that happens there is no distortion whatsoever. It's clean with a very high damping factor due to the direct connection of the phones to the outputs. The power inverter gets slightly warm during use but the lithium cell has enough power to last for hours, way beyond the Beyer getting uncomfortable. Charging is done with my bench power supply set to 4.2 volts and limited to half an amp. The Sansa Clip (the Sansa in the pictures is an old one) produces surprisingly good sound, much better than the crap Apple iSnot Nanu (second generation) that I purchased some time ago (no, don't get me started on the fruit-themed maker of entertainment gear!). The laptop LiIon cell (2 Ah) will give more than 20 hours of playing time. Current drawn from it is about 80 mA, depending on the type of DC-DC converter. A stabilized type will draw more current.
I also tried it with my newer Beyerdynamic DT990 Pro. These phones have a lower impedance at 255 ohms but the opamp doesn't mind. There is no difference between the two headphones on this amp. Of course, the old Beyer shows its age...
Enter ECC82.
While I deride people who feel that "tubes sound better than transistors" I do have a weak spot for them. So when I studied the SRPP (Series Regulated Push Pull) circuit on Mr. Valve Wizard's page and came across the remark in his paper that the high impedance of this circuit would not be a problem to drive headphones I decided to put this to the test. A quick proof of concept showed the viability of the design and I had a simple chassis plonked together from copper clad PC board in no time.
Turns out that it could have been a bit bigger but this was what I had handy. The big choke was liberated from an ancient measuring device with lots of late forties tubes (EF22, EBC41). No clue what its purpose was...
The results were a bit disappointing. Power was lacking and distortion was high: around 2% at only 1 mW. While this is rather loud it's put to shame by the OP275. And it's not quiet: there is an audible hum in both channels from the heater. Even though they're twisted it gets in the preamp. Maybe I should use a modded switchmode power supply to create the 6.3 volts. Or maybe the fact it's not floating but raised to prevent cathode-filament breakdown causes it. The anode voltage choke does a great job of cleaning the supply voltage without any solid state device. Ripple is very low. The voltages are consistent with the design. First triode has cathode voltage of 2.8 V, anode is 131 V. Second triode (bottom): 2.3 and 126 V, top: 128 and 256 V. Nicely halfway the supply voltage (261 V). Both grids are zero (milli)volts so there's no grid current or leaky capacitors.
I tried different tubes: an ECC81 (or rather a GE 12AT7) for the input provided more gain due to its higher transconductance (S). The digital computer tube (!) 5965 provided a higher output power and still more gain. Nonetheless, at more leisurely listening levels the sound is rather good. I had the feeling that it's a bit woolly. This may be due to the high impedance, almost current-sourcey nature of the drive. The output capacitors were replaced by computer grade electrolytics which solved the slight roll-off at low frequencies. The larger capacitance also mostly solved the higher voltage swing at low frequencies at the SRPP output. I will do some more testing and experimenting but first I need to fix my Lindos audio analyzer (fucked up display).
But... the view... Even with the quick 'n' dirty chassis kludge the tube glow is just wonderful:
Addendum 6 June: it turns out that the cathode resistor between the two triodes is critical. In my kludge it was 330 ohms but with ECC82s it should be more like 680 ohms. I put a variable trimmer resistor in that spot and I could tweak the distortion so the output swing was more than twice as big. It would now pump a whole 7 mW into the test load of 168 ohms. Replacing the ECC82 with a 5965 I could even coax 8 mW out of it. This at a distortion level of 2%. With this tube the resistor should be about 160 ohms. A distortion analyzer is very helpful but adjusting the trimmer to a symmetrical sine wave on a scope also yields good results. Clipping occurs at about 8 Vpp. The positive half clips harder than the negative so distortion at high, uncomfortable levels is very unpleasant. But it will cause ear damage in any case...
Final experiments (10 June): the hum is 100 Hz and this means that the supply voltage of the first stage (now ECC81) is still not smoothed enough. It has 20 mVpp ripple. This was fixed by adding another RC segment in the supply line: a 8.2 kohm/2 W resistor and a 150 µF/400 V capacitor. This effectively killed the hum. I also decided to put a dehumming circuit across the filaments: a 100 ohm/3 W trimmer with its wiper grounded. This did not make one iota of difference. Floating the filaments was a disaster however as this caused a nasty rattle.
When I built an A/B jig (1/4" sockets and a switch to immediately switch over between amps) I noticed a distinctly more pronounced bass in the tube amp. It turned out that the Beyer did not like the high impedance drive after all. Using a dummy load showed a flat frequency response. The bass lift was about 2 dB at 125 Hz. The OP275 is flat within 0.1 dB across the board with any load.
In conclusion: this is a typical case of "Nice try, but no cigar". The result from the SRPP are better than expected but not great. The "woolly" sound I noticed at first definitely is caused by the Beyer not liking a Hi-Z drive. I'll stick with the wonderfully portable OP275 mint boxie. In audio, I feel that total submission of the transducer is of utmost importance. The OP275 duly delivers. The experience was great though.
Thanks to Frank Philipse for the datasheets and Mr. Valve Wizard for the inspiration!