Headphone amplifier drop-in replacement for Phonoclone 3 and VSPS300

[rjm]

I've come up with a set of boards for a new headphone amplifier circuit that is a direct drop-in replacement for the Phonoclone 3 and VSPS300 boards, such that the power supply, chassis, and most of the connection hardware can be re-purposed. Or if you want both, it can be modular: it allows buying and building in bulk.

It's called the Sapphire.

The circuit replaces the X-reg with Zener-based series regulation, keeps a single op amp as the front end gain stage, and has a simple diamond buffer as the output stage.

The feedback loop is selectable: it can be taken from the output to keep the buffer in the loop (recommended if the buffer is biased in low-current class AB mode) or it can be routed directly from the op amp output, leaving the buffer running with no feedback (recommended if the buffer is biased heavily into class A).

Project is final beta stage. I'm still considering some last-minute adjustments:

- change values of certain passive components
- adjust size of input cap on board
- change transistor type
- add room for heatsinking the transistors

I will order a small batch of boards shortly, and they will be available at a special discount price to any "guinea pigs" who may with to participate.

/R

[jaycee]

Looks nice. I've built a similar circuit at home, just with no gain, and active current sources instead of the 1K resistors. also I just use Lm317/337 for power supplies. Sounds damn good to me!

[rjm]

@jaycee

What kind of power dissipation did you have for the transistors?

I'm still deciding what values to use myself. 10 mW class A output power into 16 ohms would seem like a reasonable target for the biasing. Much higher than that and things start to look like a mini space heater.

[Nico Ras]

Originally Posted by rjm

@jaycee

What kind of power dissipation did you have for the transistors?

I'm still deciding what values to use myself. 10 mW class A output power into 16 ohms would seem like a reasonable target for the biasing. Much higher than that and things start to look like a mini space heater.

I love playing with headphone amplifiers and like to bias them high for round about 500 mW. All you need is a heat sink from an old PC microprocessor, disregard the fan.

With a little metal box around it could look aesthetically pleasing. Why do I high bias my headphone amp? Simply because I got a set of old Rogers LS3/5 sitting on my desk and 500 mW or even 1 watt class A is rather nice to have available.

[rjm]

With an efficient driver stage and low voltage rails, 1 W class-A into 8 ohms can be obtained with a heat sink like you show there.

My circuit is not so efficient, as it's the classic LH0002 where the driver and output stages run at the same current. Also, the voltage rails I7m using are quite high. From the same heatsink (5-8W max with convection cooling) I'd be looking at about 15 mW into 16 ohms.

[rjm]

I've ordered 10 boards, so 4 pairs are available. Cost is $20 for a pair including shipping. Email or pm if interested.

I will provide support in the form of a custom BOM / instructions to each person who buys the boards, the values tailored to whatever application they propose: class A, class AB, high Z phones, low Z phones, power supply voltage, etc.

Parts cost for this project is pretty trivial, the capacitors are the bulk of the cost. $30-40 total, less power supply and chassis hardware. I will be placing an order with Mouser so if people want parts as well as the boards I can help out with this to a point.

[rjm]

brd and sch images for the production prototype boards rev 10b

[jaycee]

Here's my version (one channel obviously and without the power supply, which is nothing interesting)

It's biased at about 50mA i think. I use the heatsink from an old PC PSU for the output transistors. The MJE15032/3 transistors are probably overkill, but i didn't have any BD139/140 handy at the time. The Q103/104 transistors do get a bit warm but they're within spec.

My current breadboarded version has no gain - with the output of my PC it's not needed. I suppose you could get rid of the opamp, but the feedback gets the distortion down, and the OPA2134 opamp's FET input means low offset. Really, the coupling capacitor shown isn't needed.

[rjm]

A very sensible circuit. Zobel on the output, 10 ohms after the op amp, and RC filter on the input. Current sources replacing the emitter resistors. All very nice and proper. Thanks for sharing that.

(in Agent Smith voice) The coupling capacitors are for your own ... protection...

They aren't needed until the one day when they save your voice coils from melting.

Wrapping the feedback loop around the output buffer reduces distortion but may also reduce stability. The question is which is the more important factor in play, something I don't have an answer for right now and for that reason the Sapphire proto boards are switchable. In principle, if the output buffer is heavily biased into class A so there is no crossover distortion then feedback should not be necessary.

12V bipolar supply is a lot of voltage for a headphone amplifier, and really not needed. Well, if you are running speakers from it is, but for low impedance headphones especially +/-5V is more than sufficient.

2 of the 4 proto boards are taken.

[jaycee]

Fair note about the supplies, and I may indeed change it as I think I have a suitable 6-0-6 transformer someplace.

I tried the amp with the opamp feedback taken from the opamp output, and with the feedback taken after the buffer, and preferred the latter. I've had no stability issues except for when I was trying different opamps, some would oscillate as soon as I attached my oscilloscope probe to the output. The Zobel on the output cured that problem.

The L_VA and L_VB pins are to connect a potentiometer for volume control. The third pin of the pot would go to AGND, and you'd omit R102 - the input impedance would then be set by the pot. I did it this way so that the opamp and the attached device sees the same impedance no matter what the volume setting is.