As I've got some old parts laying around, I decided to build a headphones amp for the office.

Here's the idea. Not the highest quality but it only uses what I've got in hand. Signal is taken from the pin5, loaded with a 5ma CCS. Output stage is very conventionnal class AB, with a bias of around 12ma.



I'll probably try to find some time tomorrow to solder a channel on perfboard. I'll welcome any idea on how to improve this thing (improve... not turn it into a beta22 )

Since you've shown the schematic here, I guess you want some feedback. I think this is going to be a fine and neutral sounding amp, with the slightly edgy sound of BC327/337 balancing the slightly soft sound of NE5534.

Personally I find a diamond buffer to sound better than a single emitter follower. The drawback is more parts and there'll be two ccs's to adjust unless you build it with current mirrors à la PPA v2.

You've choosen R4, R5 and R6 to balance the highish bias currents very nicely, but using high values like this will give you a quite noisy amp, and it's going to be audible with IEM's or low Z high efficient phones. I've used NE5534 without input capacitor and the DC offset is something like 5 - 15 mA, varying with the volume with the typical scratching sound of DC going through the pot, but I think it's OK.

With a 1uF input capacitor the cutoff frequency is a little high.

Finally, does the output emitter resistors really have to be 22R?

Hello, thanks for the comments.

Yes, the resistors values are a bit high. However, I plan to use the amp with highish impedance cans so that should be ok. I don't have much choice because the source will be a simple pcm2702 and I would like the input impedance to stay reasonnably high.

The 1µF at the input might be bumped to values up to 10µ.

The 22r resistors combined with a red led gives a reasonnable bias. One could go down to 10r, but the dissipation in the output trannies rises to about 200mW. Which is ok but should not be exceeded considering I'm dealing with TO92's (max rating 625mw).

For biasing you can use a vbe multiplier like eg in JISBOS (you don't have to use a pot, just two resistors), or you can use a single resistor (if it's not going to be battery powered), or you can use two 1N4148 diodes + a small resistor to set the bias (15R or something like that).

With 22R emitter resistors, the current will be somewhat limited.

are the base resistors really necessary?

These are usually used for MOSFETs with high input capacitance.

I'm listening to the little beast right now. It's a quite decent little amp, with no obvious faults. Clearly better than my previous cheapie (see http://www.head-fi.org/forums/f6/my-...-2-2-a-316277/ ).

The design changed a little bit. I made things slightly more simple. The CCS is now a jfet BF245C instead of the BJT CCS. I removed the base resistors. As the LED I had on hand had too much of a voltage drop, I had to increase the emitter resistors to 33R. It still gives me a bias of around 15ma and keeps the bc337-327 cool. I use no caps at the input. Using a linear pot combined with the inputs resistors gives a decent enough attenuation curve. Offset is in between 3 and 13mv on one side and 1 and 9mv on the other side.

The power supply is also dead simple. I use a 24VDC wallwart, regulated down to 18VDC by a LM317. The virtual ground is provided by a passive rail splitter (1K resistors + 1000uF caps). Even at high volume, the voltage split is fairly equal.

I attached the layout on perfboard and the schematic. Red is copper, blue jumpers.

Does it power your AKG K340 well ?