My Headphone Design

Hi Friends,

I'm a 17 years old tube enthusiast from Brazil, and this is my idea to use two XCL82 tubes to make a OTL design with negative feedback to drive headphones.

In the frist idea is drive only a Sennheiser HD580 headphone, but when I has simulated the pentode of XCL82 in TubeCad the Zout was around 86ohms ( 20mA, 200v Triode Connected, rk = 330ohm, gain is 0.71 ). A little lamp lighting in top of my head with the idea of put negative feedback in the amplifier and reduce the Zout to can drive 32ohm headphone.

I’m thinking to use the XCL82 because I can buy this in Brazil for US$1. This is a very popular tube.

But If I can’t drive some 32ohm headphones the idea is make a very stable and very good amp for mine headphone.

The schematic you can see below. Is C1 necessary? Is a good idea change the R1 to a CCS with transistors? I haven’t worked with it before.


Best Regards,
Felipe Navarro
Hi Felipe,

for negative feedback pls. connect the feedback to the grid of the input triode.

A better method for degreasing the output impedance is a "ring of three": a CC stage, a CC stage and a CF stage with overall feedback from the cathod of the CF to the cathod of the first CC stage.

Regards Andreas
Here ya go mod_evil. Feedback should be routed to the grid of the triode. Please ignore the values of components as your shema didnt have any I left them at default. Feedback becomes the ratio of R1 to R2. The input impendance also basically becomes the value of R1 so make sure it is high enough for your source to drive.


  • mod_evil.gif
    22 KB · Views: 548
I'm assuming the xcl82 is the same as the ecl82/6bm8, high mu triode/power pentode. Feedback will be required to make the gain of the triode managable, comparable to 12ax7 with slightly lower mu. It may be capable of giving decent volume out of a set of 300ohm phones but definitly will not drive 32ohm. If you really intend on using 32ohm hp's then configuring it as a standard single ended amp with 4k:32ohm output transformer is probably the best way to go. Cathode follwers are funny things. They do not have the drive current that the output impendance and idle current would suggest. It is at the mercy of the plate resistance. As such you'll find most otl designs use a very low Rp triode for the output, usually a few in parallel, but these are still pretty limited on how much current can be delivered to a low impendance load.
You may want to try the circuit shown below. I have not actually built it yet, only simulated in SPICE, but it should work. The simulation indicates that for an input of 1 volt peak, .7RMS, the headphones will see 200mV peak, or .6mW, at about 0.5% distortion. 0.6mW doesnt sound like much, but according to the specs I have, 32 ohm Sennheiser headphones for example have a sensitivity of 94dB per 1mW. So .6mW will give you 91.8mW, which is plenty loud enough for me, although some people may want more.

I could not come up with a direct coupled version, hence the inclusion of C1. Also C4 is necessary to keep the gain up.

SPICE is not reality, but this is a starting point for you. If I can find some PCL/ECL82 in my stash I will build one.


  • xcl82headphone.jpg
    40.4 KB · Views: 412


2002-08-15 10:16 pm
i built a simple parafeed SE ECL82 headphone amp with a mains toroidal as the output transformer. i never really liked the result and soon stripped it down and built a much better PP version.

I think one of the main issues that you face with the input triode is overload. Its very difficult to bias this triode up at anything greater than 1V. Since many inputs will peak higher than this the input stage is in constant fear of overloading. My version sounded harsh and edgy and I believe it was for exactly that reason.

In my newer PP version I have a microphone transformer phase splitting driving a LTP. The transformer has a step down of 9:1. There is absolutely no hint of edgyness now - coincidence, maybe but I suspect not. So my advise is use a step down input transformer, or if that is to expensive use a voltage divider to reduce the input signal before it hits the grid. The point is don't allow the grid to approach overload. I suppose that the feedback applied to the grid would help to protect from overload, but I don't like that type of feedback so I wouldn't try it. I applied global negative feedback to the cathode and though it helped, the sound was rather lifeless.

Hope that helps a little. If you want to see what I did with the PP amp drop me an Email.

Overloading the grid can be avoided by bootstrapping the grid resistor. In the example I show here you can see that I have the bottom of the grid resistor R3 connected between R1 and R2 where feedback would be applied if using cathode feedback scheme. Choose the value of R3 so that the cathode current will drop a dc value about twice the expected input ac rms voltage. I use this method for a 12av6 based hybrid. The bias voltage for the tube is only about .8 volts in my circuit yet I can input nearly 2 volts ac rms with no issue. This doesnt effect cathode feedback either, just have to choose your voltage divider based on the value of R3 for the ratio you want. Bootstrapping the grid resistor in a circuit which the feedback is applied to the grid is almost mandatory so as to make the grid resistor appear to be an extremly high value to ac, and reducing its influence on the feedback voltage divider.

Sorry the picture looks so screwy my graphics program didnt seem to like it.


  • bootstrap.gif
    6.9 KB · Views: 305
After much thought (I'll save you the boring details) I figured using a high gain triode with very heavy feedback is a bit redundant and most likely wont sound very good. So take the triode out and just use the pentode as a buffer. But having that perfectly good triode there unused bugged me. It's bed time for me so I'll give more details later. Here's what I came up with.


  • otlheadamp.gif
    27.7 KB · Views: 254