Simple Class A Headphone Amp Using THAT1646
 

Here's a headphone amp I've been playing around with using a THAT1646 line driver in an unusual configuration to provide bias. It's fully DC-coupled and has low output offset. It's a rather long thread - I've linked in to where the schematic (as a headphone amp) develops. RMAA and FFT tests follow in later posts. People who've built it like the sound. I've posted it over at HeadWize but we can't update there right now and I thought you guys might like to have a peek at it.

Pro Audio Design Forum • View topic - Using the THAT1646 As A Transistor Pre-Driver/Headphone Amp

Wayne

Wayne, that looks like a cool little project.
I like the fact that the output runs open loop.

Experience suggests that 30 - 60 mW is more than enough for most people with modern high-efficiency headphones, so your design seems spot-on.

Can a DRV134 be used instead?

Thanks Jen-B. The open loop output throws a lot of people with FB internal to the 1646. I tried a closed loop approach with overall FB during design (to reduce idle current to A-B) and just couldn't get as good of THD performance.

It's unity gain so for some 'phones like the old high impedance AKG-240s it might need some upfront gain but I haven't seen the need for it with modern 'phones. The drive power is indeed adequate for the MDR-7506's I use. The SR is about 8V/us which is plenty for the 1-2V P-P drive.

It's a power hog though - my application was for mastering and non-portable. I've had a few people ask about batteries and I kinda shook my head.

What really surprised me about was the DC performance. I haven't had to match OP devices and the worst offset I've seen is 6-7 mV. So it doesn't require FB or a servo which keeps it simple.

theAnonymous1. I really don't know about using the DRV134. I've never tried it. The THAT1646 uses OutSmarts and I tricked it into driving the outputs in "common mode" by using the sense inputs (pins 2 and 7) for signal input and the "normally-used" differential input, pin 4, for complimentary DC Vbe bias.

The DRV134 and SSM2142 use a cross-coupled approach (versus OutSmarts) to line driving. I don't know if driving the sense lines with input produces a "common mode" output. I suspect that they will, but I never tried them. It's easy enough to find out though. I have a DRV I can drop onto the PC board.

I've done comparisons between the DRV134 and THAT1646 as line drivers driving several hundred feet of cable and the FFT performance of both are nearly identical and only second and third visible. They're both very good parts.

I would really appreciate that.

I'm interested in building this, but I hate buying new parts. I have plenty of DRV134.

I tested a DRV134.

It looks like the DRV134 will allow "common mode drive" using its Sense pins as inputs the same way a THAT1646 will.

What is different however, in this unique configuration, is that the differential gain of the DRV134 appears to be twice what the THAT1646 is. This creates bias issues. It does however pass signal and appears to work with an idle current that is far too high.

The Vbe multiplier which drives the THAT 1646 differential input is "mirrored" on the outputs, pins 1 and 8. Other than a few mV of internal 1646 offset the outputs are copies of the Vbe multiplier. One is +Vbe the other an inverted -Vbe. Both Vbe copies are pinned relative to ground in the absence of signal. Since the Vbe multiplier is a third transistor sitting on the heatsink (also a BD139) the thermal tracking is quite good and the offsets low and relatively drift-free.

With the DRV134 having a diff gain that appears to be twice what the 1646's is, the idle current is significantly higher. The Vbe multiplier's output needs to be scaled by approximately 1/2 for it to have adjustment range for idle currents in the 60-80 mA range.

I need to look at why the DRV's (and SSM2142's) differential gain goes from two to four when it's used this way. It's likely a trait of the CCOS topology. A good reference for CCOS outputs is here:

http://www.proaudiodesignforum.com/I...ES.pdf#page=31 (Figure 24.)

My hunch is that the Vbe multiplier needs to be divided by two before being fed into differential Vin. SInce the tempco of the Vbe multiplier would be scaled and then multiplied again it should track OK. That would require a hack. So I guess the answer to whether a DRV134 can be used is maybe.

Anyone willing to experiment on a similar circuit based on DRV/SSM? I got a pair of those but hardly any time to throw a test circuit together, even more so because I think my SSMs are SMDs. Maybe some day.

I've toyed with this circuit in a simulator using DRV134 model and here is what I came up with:

• Leave everything unchanged. This gives Iq in the 200 mA region. This means 3,5W on every transistor with +-15V supplies. No big deal, I'll just use a bigger heatsink, 14W isn't something to be scared of.
• Use darlingtons istead. Solves current problems but I'm not aware of other possible issues.
• Use a Schottky diode as a reference, according to my simulation, this results in 90 mA'ish currents. Not sure about thermal tracking.

I finally had a chance to try this circuit using a DRV134 and it only requires the addition of two 470R resistors to scale the Vbe multiplier output by 1/2.

The DRV134 in the "common mode drive" configuration has a differential gain of 4, the THAT1646 has a differential gain of 2.
This causes the DRV134 to have twice the delta Vbe at each output compared to the THAT1646.

http://www.ka-electronics.com/Headph...6_Vbe_Mult.JPG
THAT1646 Class-A headphone amp.

To use the above circuit with a DRV134 (or SSM2142), add a 470R between the collector of the BD139 Vbe multiplier and the 1/25 uF at pin 4. Also add a 470R from pin 4 to ground.

An interesting little circuit. Have you ever tested capacitive driving stability with no "build-out resistor"? (Output series resistors are thoroughly "out", you know. Even a number of "big" cans appreciate <10 ohms, and several multi-driver BA IEMs even want an order of magnitude less.) Since the whole affair is running open loop, that probably won't be too much of an issue, but anyway.

If you've got some sensitive BA IEMs floating around (like Shures, UEs or Westones), you can listen for audible noise, too. If these show a pitch black background with the input shorted, so will everything else.