Eight Part Harmony: The Tao Headphone Buffer

[Steve Eddy]

Here's a very simple (weighing in at only eight parts) but very nice sounding headphone buffer for anyone interested.

The schematic should be rather self explanatory. Q2 and Q3 need to be heatsinked and the enclosure well ventilated. Even though they're dissipating just a couple of watts, they get very hot in human terms.

R3 is specified at 8 ohms which isn't a "standard" value, but was the value of the Caddock resistors I used when I first built and measured the circuit. You can simply use 7.5 or 8.2 ohm resistors instead.

Cobbling it together for the first time with just parts on hand and 1% tolerance for R2 and R3, the offset was only about 7mV. You can take two approaches to minimizing offset by either hand selection of fixed resistors or R2 or R3 can be replaced with a 10 ohm 1/2 watt trim pot. My preference would be R3 with each end of the pot connected a per R3 in the schematic with wiper connected to the emitter of Q3.

Experiment with different resistor types for all resistors. I've become rather fond of some Stackpole carbon films.

The transformer isn't required but I much prefer it. It provides ground isolation and can readily accept balanced or unbalanced sources. And to me it just sounds much better.

The circuit isn't limited to headphone use but can also be used as a general purpose buffer (with a +/- 12 volt supply it'll swing about 10 volts peak).

You can also replace the fixed resistor R1 with a 20k ohm pot or stepped attenuator and call it Sistah Steve's Buffered Passive Preamp.

Just make sure that you use a pot or series/ladder type attenuators so that the transformer will see a constant 20k ohm load. This also precludes the use of linear pots with law faking resistors.

Althoug it's been solid as a rock under all the conditions I have used it in, I'm sure someone will find a way to rock the boat. Stability issues can be address with a "gate stopper" resistor of around 220-470 ohms in series with the gate of the input transistor of Q1. Output instability can be addressed by applying compensation capacitance between the base/collector of Q2.

Oh, and the more cautious can simply replace R3 with a 470-1,000 uF non-polar electrolytic cap.

Enjoy.

se

[Circlotron]

Originally posted by Steve Eddy
Oh, and the more cautious can simply replace R3 with a 470-1,000 uF non-polar electrolytic cap.

After several seconds the cap would charge up and the amp will stop working. Instead, did you mean put the cap in series with the load to block any flow of DC if something goes wrong?

[Steve Eddy]

Originally posted by Circlotron
After several seconds the cap would charge up and the amp will stop working. Instead, did you mean put the cap in series with the load to block any flow of DC if something goes wrong?

Ooops. Yes, I meant remove R3 and put the cap in series with the load. Oh, and sprinkle the whole thing liberally with protection diodes.

Thanks. I was in a bit of a rush when I threw in that last line as an afterthought.

se

[PRR]

Sweet and simple.

> the more cautious can simply replace R3 with a 470-1,000 uF non-polar electrolytic cap.

??? That would block all DC current; it won't be any kind of amplifier. I'm sure you meant to say "Put a 470-1,000 uF non-polar electrolytic cap in series between the output and the load". oops you caught that while I was writing...

> can also be used as a general purpose buffer

Yes, and for higher-Z loads (or for initial smoke-testing) you could use 100Ω for R2 R3 to keep current down to a couple dozen milliAmps. That would drive 600Ω to 10Vpk without clipping and make ample noise in headphones. Then if you need more power or lower output impedance or just want to run over-rich to improve linearity, you can go 33Ω then 10Ω until you get happy or your heatsinks become a problem.

As a "power amp", it is inefficient. Not just because it runs Class A, but becuse the bottom device runs a fixed current so efficiency is half of a push-pull or choke-loaded power amp. Efficiency isn't everything of course, and "wasting" power is often a quick clean path to small simple nonlinearities.

I wonder about running with no base-emitter resistor. Mostly because I am so old that I remember transistors with so much Ico that they would run-away without a base drain. That's not a real issue with today's clean silicon. There's no pull-down on base capacitance, but if running Class-Rich-A with fast transistors and audio signals that may not be an issue.

If you need a specified output impedance in mass production despite parts variations, trim R2 instead of R3 to null offset. You could instead make R2 twice the size of R3 and put a 1K pot across it, connect lower FET Gate to pot wiper, and trim offset.

[Steve Eddy]

Originally posted by PRR
Sweet and simple.

Thanks. That was the goal. I've another version that I'm working on just for Grado headphones which uses a passive load and just 6 parts, but I want to evaluate a few more output coupling caps before I post it.

??? That would block all DC current; it won't be any kind of amplifier. I'm sure you meant to say "Put a 470-1,000 uF non-polar electrolytic cap in series between the output and the load". oops you caught that while I was writing...

The ooops is all mine. T'was Circlotron who caught it.

Yes, and for higher-Z loads (or for initial smoke-testing) you could use 100Ω for R2 R3 to keep current down to a couple dozen milliAmps. That would drive 600Ω to 10Vpk without clipping and make ample noise in headphones. Then if you need more power or lower output impedance or just want to run over-rich to improve linearity, you can go 33Ω then 10Ω until you get happy or your heatsinks become a problem.

Yup. Though the 1837's maximum I_C is 1 amp so can't run too rich. The Grado amp and the power amp will use 1302s (and different FETs) but I don't see any reason why they couldn't be used in this circuit. They'll happily do a couple of watts without a heatsink.

As a "power amp", it is inefficient. Not just because it runs Class A, but becuse the bottom device runs a fixed current so efficiency is half of a push-pull or choke-loaded power amp. Efficiency isn't everything of course, and "wasting" power is often a quick clean path to small simple nonlinearities.

Yeah, it certainly won't win any EnergyStar awards.

I wonder about running with no base-emitter resistor. Mostly because I am so old that I remember transistors with so much Ico that they would run-away without a base drain. That's not a real issue with today's clean silicon. There's no pull-down on base capacitance, but if running Class-Rich-A with fast transistors and audio signals that may not be an issue.

Yeah. I've tried the circuit both with and without. It didn't sound any better with them and I couldn't come up with any compelling technical reason to keep them so... I didn't.

If you need a specified output impedance in mass production despite parts variations, trim R2 instead of R3 to null offset. You could instead make R2 twice the size of R3 and put a 1K pot across it, connect lower FET Gate to pot wiper, and trim offset.

Thanks. Yeah, that'd be particularly useful if one weren't using a matched FET pair.

se