Feedbackless headphone amp


2013-08-14 1:19 pm
Hi, I found this cool circuit on this site:
the Headphone amplifier

I guess I briefly understand the operation of this "super linear stage"
I simulated the circuit this fellow has made. Thd probe shows about 0.05 thd at 2 and 20kHz. I think it is mostly inaudible.

Yet I´d like to know what is the main cause of the remaining distortion and how to possibly bring THD further down.

I´w read some of Bob Cordells "Designing Audio Power Amplifiers" but I don´t know how to apply the tricks to this circuit. I wonder if current sources, cascode or some kind of local negative feedback would work. I´m not very good at designing circuitry but I could try if someone would be so kind and give me some hints.


  • headphoneamp.jpg
    287.2 KB · Views: 141

Mark Johnson

Paid Member
2011-05-27 3:27 pm
Silicon Valley
For those wishing to see the schematic in greater detail


There are two common-emitter stages with 1K emitter resistors and 10K collector resistors. I observe that if you replaced the two 10K fixed resistors with a single 20K potentiometer (wiper @ ground), you would now have the ability to continuously adjust the amount of 2nd harmonic distortion AND you can dial in either negative-phase H2 or positive-phase H2.

Last edited:


Disabled Account
2017-02-07 1:54 pm
...I´m not very good at designing circuitry but I could try if someone would be so kind and give me some hints.

You have applied the simplified configuration into the original design, but I'm afraid it will not work as expected.
The 1k-10k node in the simplified config is at the pos-rail side pulled down to -1.4Vdc by these resistors, and on the neg side up to +1.4Vdc.
But that renders the input bjt's impossible to operate: Vbe is reversed (0.8V).
Simulation software tends to ingore this impossibilities in AC analysis.
Connect the upper 10k not to the neg-rail, but to the pos-rail instead, and for the lower to the neg-rail. This will open up the input bjt's sufficient.
The mirrors wll be running at some 1.2mA.

The integrator ('vas') stage runs on 1.2mA also, and is a bit meager.
Decrease the 1k Re's to 330 ohm at least (I'd say 100 ohm). Adjust the bias pot accordingly 3 (10) x lower.

Localized overall feedback (over 2 stages only) is posssible, but very tricky. This 3-stage design could have 2 loops (2nd to 1st and 3rd to 2nd), but as there are no easy 'entry points', serious dc shifts and ac pole-zero divergencies will occur which must be tackled. You better try the original design.
Last edited: