Some explanations:
- the input pot is a 100K linear one. Interaction with the input resistors gives a usable (not perfect) curve for audio.
- Offset varies between 13 and 3mv in one channel, depending of the position of the pot (the other channel is in between 0.8 and 9mv).
- Input impedance is kind of low at min 13K
- The signal is taken from the pin5. The jfet ccs which biases the ne5534 also biases the led for the class AB output stage.
- The emitter resistors' value is a bit high at 33r. I picked the value according to the forward voltage of the leds I had in hand. It gives me a bias of around 12ma.
- Power supply can be about anything in between +/-9 and +/-15VDC. I used a 24vdc wallwart, regulated by a lm317 to 18v and split with a passive rail splitter.
- Some additionnal decoupling isn't shown.

Original idea from here: http://www.diyau...5703#post675703

I especially like the idea of reusing the LED you wanted anyway as a power indicator for biasing. You've gone one better than I did in PIMETA v2, as yours is used for 3 purposes, instead of just 2.

Would you consider doing a variant with a bipolar CCS instead of the JFET? Then your amp can be built from Radio Shack parts, except that you have to use the op-amp in the normal fashion, rather than driving the buffer from the op-amp's comp pin. Even the lowly TL072 -- the only sane choice, of the few op-amps Radio Shack offers -- should sound decent given a transistor buffer.

To give you an idea, here's the layout under eagle (the amp is on stripboard). It's roughly 10cm/8cm.

The opamp output (pin 6) is not connected.
The output signal is taken from one of the compensation pins of the ne5534 (pin5)

@Tangent:
A ccs with a pair of bjt would indeed easily replace the jfet one. Actually, my sims had such a ccs at first. A jfet ccs is just a bit easier to solder at half the parts (call me lazy). The jft ccs costs me 15c, the bjt one 10c (from a german supplier on ebay).

---

[Quoting 00940]

The output signal is taken from one of the compensation pins of the ne5534 (pin5)

---

---

[Quote]

A ccs with a pair of bjt would indeed easily replace the jfet one. Actually, my sims had such a ccs at first.

---

---

[Quote]

doing it this way bypasses the op-amp's output stage, effectively replacing it with this more capable one.

---

How interesting.

Yeah, I had a "???" moment too, even though I have seen this trick before.

The most common op-amps can be broken into three stages:
1) differential input
2) high voltage gain stage ("Vas")
3) current gain output stage

For easy use with feedback, we must make one stage much slower than the other two. In the LM301 you connect a pFd "compensation" capacitor across two pins. In the LM741 this cap is internal. The dual NE5532 has caps built in. The single NE5534 has small internal cap plus pins to add an external cap.

The red dots in the plan above are the compensation pins.

Note that one of them is also the point between the voltage-gain stage and the current-gain stage.

The actual 5534 plan is a lot messier than the one above. And because it is an old design, from days when integrated PNP transistors had poor performance, the 5534 uses a very tricky current-gain output stage which acts like a complementary emitter follower, however all the hard work is done with NPN transistors.

While tricky, the 5534 output stage really IS good. However since it is on a small die with a lot of other parts, including quite large input transistors, it is limited to 30mA-40mA maximum. It was intended to drive 600 ohms very well. It will drive lower loads, but 32 ohms is a real strain, and very little power (25mW) is possible in 32 ohms.

What 00940 and folks at DIYaudio have suggested is: keep the excellent input and volt-amp stage, but attach a "better" output stage. Buying discrete, we can get PNP as good as NPN, we can get devices larger than Rupert cared to squeeze into the chip, and high load currents in exteranl devices do not heat the input devices and cause bass distortion.

00940 shows a very ordinary complementary emitter follower, using high Hfe transistors. It will drive over 100mA, even 200mA.

The internal output stage is still connected, but drives nothing, and does no harm.

My only comment is that it isn't utterly short-proof. There is no separate protection for the output devices, and on paper I can find a situation which will exceed their rated dissipation. But heck, I've run a lot of crude speaker amps with no protection and much higher worst-case over-spec. Worst-case rarely happens. And headphones get shorted a lot less often than bare-end loudspeaker wires. And it IS a DIY project.... if you can build it, you can re-build it.

thanks

interesting thread and great explanations on some new concepts for me.

is there an RMAA result set for this amp? I'm curious