The XEN XELF Headphone Buffer

I can measure, will do for BJTs, but would prefer to mess with the MOSFETs as little as possible... too expensive to risk damage from ESD or another mistake.

I was thinking of trimpot aswell, with 1 power input for both channels traces would need to be cut if you were simply measuring current draw from the supply, is there a better way than that to measure bias current individually?
 
It is a bit rough looking, the caps (Nichicon ES non-polar) were backordered so the gold ones are some spares sitting in for now.
 

Attachments

  • IMG_0137.JPG
    IMG_0137.JPG
    559.2 KB · Views: 237
About 200mA bias with ±8V supply.
Loads so far were some 300ohm sennheisers and another 32ohm headphone.

The low voltage supply is for using super caps (3x 100F 2.7V in series per rail), I already feel like using higher rail voltages will be more worthwhile but it'll still be interesting to see if there is any effect from the super caps.
They were free samples, otherwise wouldn't have bothered.
 
For initial testing and listening a D10S DAC was used, switching after a few days to TDA1387 DAC with passive resistor I/V that suits the XELF a lot more .
This combo sounds just fantastic, though more interesting was how plainly obvious the flaws of the D10S become with the extreme resolution of the XELF.

I am using the super regulators for 8V supplies but plan to switch to UltraBibs, there will be even less technical reason for using the super caps with bibs, I can't really be bothered testing them, it would only be a test since there are other, better places I want to use them.

Naturally you would raise up the voltage from 8V but there is always complications, I need to change to a higher voltage but smaller and lower quality trafo from the current one, because of this it might be difficult to make a fair judgement on only how rail voltage is effecting the amp.
It can probably be pushed to 9-10V with current trafo and my input levels are usually very low so i'll prob just leave it.


According to the LTSpice sim increasing rail voltage causes THD to increase at higher frequencies.
If this is a real and understood indication of something happening at higher voltages (e.g higher capacitance) having any small additional reason to stick with a lower rail voltage would be nice.
 
Last edited:
As long as you are happy, it is the most important thing. :)

You can easily try out higher voltage if you have lab supplies.
If not, using 16x AA Alkaline will give you a few hours testing.
Enough to give an impression as to what suits you better.

And the circuit should have good PSRR.
I don't use any fancy regulators myself.

Lateral FETs have very linear capacitances.
The BJTs do have Early effects and hence voltage dependency.


Patrick
 
Took a while but I've been experimenting with XELF as ZFB speaker buffer.
Initially it sounded poor (rough treble and lean bass) and better with the op amp (ne5532) to 'smooth' things out and improve bass, this was with 15V rails and 700mA bias, increasing rail voltage or bias did very little/nothing to help with these issues. Luckily it was a simple problem of not having enough supply capacitance.

Replacing some smaller parallel lytics with 2 x 33mF per rail helped immensely, I think Nelson said somewhere that you can never have too much supply capacitance so there may be further improvements to be had, .

In this state it sounds really good, better without the op amp, I think it's best speaker amp I've heard.

This is used as a pure buffer driven directly by Topping D10S, with multiple op amps in the DAC output stage I think the real purpose of this ZFB buffer is already defeated.

The real fun will be to combine this amp with something like TWTDAC-LT or a Soekris DAC, direct output with nothing but a simple buffer in the signal path.
High sensitivity/efficiency speakers will be another important piece to this system, allowing higher volumes from only the ~2Vrms DAC output while also reducing power demands of the buffer.
 
Last edited:
Well, for use with speakers, it is just a source follower using the lateral MOSFETs.
You will find some improvements using the Exicon double-die devices, and with a higher bias (say 1A).
No need to mention, power supply to suit.

If you want to drive this without any opamps in the signal path, you should go for a R2R DAC like the Miro AD1862.
And then use a high-gain / high-voltage discrete IV converter for the output.
You are then not restricted to 2Vrms out.

At least that is what I would do.


Patrick