Low Distortion Buffer

[nickm]

I have a distortion analyzer and I want to drive low impedance loads at low voltage with its generator.  The generator output has an output impedance of 50 ohms so I want to add a unity gain buffer stage.  The buffer needs to have distortion low enough so it doesn't add to the THD+N floor.  What I've tried so far is just a textbook amplifier which works well at no load, but as soon as it's loaded the distortion goes up.  It seems the dominant source is crossover distortion.  I don't have as much experience in linear amps so I'm hoping one of you has suggestions for improvement.

The opamp is an LME49720.
I've tried 3904/6 and IRFZ24/34 for output devices.  Both performed well under no load, but with 16 ohms the distortion was bad (3% and 0.25% at 100mV respsectively).
I've got them powered from two nine volt batteries.  2x1000uF caps + 2x0.1uF caps for bypass.
The gain should be around unity.

[dannyf]

If your application allows, a simpler emitter follower should do. Alternatively, get rid of the resistors / diodes and put a small resistor (330ohm for example) in between the transistors' b/e junctions.

Haven't read the mosfet's datasheet but it is unlikely a normal opamp can drive those things at audio frequencies.

[Marco]

More bias current? Try two diode drops and an emitter resistor to set the bias like on page 3 here :

http://www.ti.com/lit/an/snoa600b/snoa600b.pdf

[dannyf]

You may also consider a composite amplifier, like an opamp + a diamond buffer (either discrete or an IC). Many CFB amplifiers use diamond buffers as their output stages so you can use a VFB opamp + a CFB opamp for that.

I have used AD815 as power amps on 8ohm speakers, for example.

[Co6aka]

Maybe BUF634T or OPA547T would do what you want. 

[Marco]

Or LME49600 to stay in the theme of the TI audio components.

[oldway]

How low are your low impedance loads?

For higher current in the load, you don't need a buffer: you can rise the output voltage of your generator to a higher value than 100mV.
You have to measure the voltage across the load and ajust the voltage to read 100mV.

If you need a buffer, a NE5534 can drive a load lower than 50R at 100mV (without need of adding transistors) with a distortion of 0.004% in full audio range but output current is limited to 38mA peak.

[wiss]

How low THD? How much power?
A 5532 with a class A buffer will probably work ok.

[nickm]

Thanks for the replys.  I've managed to cut the distortion in half by increasing the bias current.  I'll have to try the other suggestions.  I want to drive headphones so at least 16ohms but the amount of power is less than 100mW.  The floor noise of the of the analyzer is 0.02% at 100mV so it would be nice if it was less than that.

[wiss]

Thanks for the replys.  I've managed to cut the distortion in half by increasing the bias current.  I'll have to try the other suggestions.  I want to drive headphones so at least 16ohms but the amount of power is less than 100mW. 

A NE5532 with a class A stage inside the feedback-loop, you might need to add a compensation-cap over the op.

The floor noise of the of the analyzer is 0.02% at 100mV so it would be nice if it was less than that.

You might even get away with using a tl071...

btw, build two channels with the 5532 and you have a ~10 ppm THD headphone-amp!

[tom66]

Have used LME49600 before, it's really simple. You should add an op-amp in the loop to keep the output stable with temperature and supply voltage. Distortion figure lower than the 5th zero past the decimal point.

[blackdog]

Hi,

Use one 5532a per channel in parallel, it wil blow your eardrums uit of your nose with low distortion en very cheap 
Use a good capacitor direct over the powersupply pins.

Kind regarts,
Blackdog

[dannyf]

-100mW.  The floor noise of the of the analyzer is 0.02% at 100mV-

given how much speakers and headphones distort, I would encourage you to focus somewhere else for improvements.

I would focus on high open loop linearity / low global feedback and speed.

ad815 and other cfb amps - tpa6102 for examples - are amazing headphone amps.

a discrete jlh1969 with to92 or to126 devices would be great too - though its the figures aren't that great.

[nickm]

Well I tried the LME49600 and it works great.  Its down at the floor noise all the way until clipping.  I tried 3ohm load and it was perfect into that (Although the distortion hit the cliff at a much earlier point than the higher impedance loads.  After messing with this I think the CMOY folks are a little crazy trying to drive headphones with opamps.  They must be at least 1% distortion through all amplitudes.

[Marco]

To be fair, your BJT output stage added a ton of crossover distortion.

[wiss]

The bjt-stage has to be within a feedback loop!

[dannyf]

I think the CMOY folks are a little crazy trying to drive headphones with opamps.

You are comparing a buffer with a poorly designed opamp + buffer. Not fair at all.

[Marco]

The bjt-stage has to be within a feedback loop!

It was, he just used labels instead of a wire for the feedback loop ... which does not improve intelligibility of the schematic in this case.