6N3/6N6 headphone amp using PCB.

[kodabmx]

I forgot there's an easy fix for that. Since the feedback resistor is 390k ohms (pretty large) you don't need a huge capacitance to pass all audio frequencies of interest from the output to the input tube's grid, while blocking DC. I think 0.47uF will do, but 1uF might be better if you're worried. A 1uF 400V film cap isn't too big of an ask.

I redrew it and I'll give it a try soon.

 

[kodabmx]

As you can see its not linear for driving 32R phones.

View attachment 944921

I would be adding a say 10:1 transformer on the output to overcome this.

Two questions - was the distortion present at lower levels? Could you change 100R to 68R and see if that helps? How about adding in Rongon's loop as it should lower the Zout?

Thanks!
One of these days, I'll get spice going but since I can't get it running with wine right now, I need to install Windows in a Virtualbox. Then install it and learn how TF to use it

 

[baudouin0]

will look a tonight if ok

 

[kodabmx]

Certainly. Thank you!

 

[baudouin0]

i have used lt spice on wine no problem

 

[kodabmx]

I get a segmentation fault. There's just not a working package at the moment. The ltspice package actually installs it's own Wine sandbox. The old version worked sort of

 

[Ryssen]

Is the 100k resistor on the input needed?I mean for the feedback.?

 

[kodabmx]

Yes. 100k and 390k define the feedback ratio.

 

[baudouin0]

The 68R helps a bit. The feedback reduces the distortion up to where it runs out of current. I may have installed net framework 4 on wine.

 

[baudouin0]

I suppose since the output swing is quite small replacing to bottom 6N6 with a resistor and moving this to the top would give more current. Nothing beats a transformer.

 

[kodabmx]

Thanks for that.
How about with 6N2P instead of 6N3P on the 68R with feedback?

 

[baudouin0]

Certainly that gives more gain for the FB to work on. About 11.5dB of FB.

 

[baudouin0]

Seems you can increase the output drive by making the bottom tube active.

 

[kodabmx]

Is that a WCF? or am I looking at it wrong? Nice result though

 

[rongon]

Yup, that's a WCF.

R2 is the (small) plate load on U2.
The inverted signal from the plate of U2 is fed to the grid of U1 through C4.
R12 is the grid leak for U1.
The output is taken from the cathode of U2/plate of U1 (same thing).

New Adventures in the White Cathode Follower

According to theory, the WCF should be able to swing almost 2X its quiescent current, so if you bias the 6N6P to Ip = 20mA, it should be able to swing close to 40mA into its load.

 

[rongon]

Of course now we have to look at the distortion spectra to see if it's predominantly 3rd harmonic instead of predominantly 2nd harmonic, because now the output stage is push-pull.

 

[baudouin0]

Should have been an inventor - never mind. More drive much lower 2nd harmonic, more gain for the NF. Downside is that it can couple HT hum into the output. Does look much better into low impedance.

 

[baudouin0]

That's good reading. By tweaking the current sense resistor you can null out the HT hum as the first stage generates some. Just a win win win.

 

[rongon]

...but only into one specific load impedance. Change the plate load resistor (current sense resistor) value and the null goes away, or... Change the load impedance and the null goes away.

The usual compromise solution is to pick the value of current sense resistor so that the WCF works 'good enough' into the expected range of load impedances; in this case from 32 ohms up to 300 ohms. The target might be 125 or 150 ohms.

That Broskie dude has some wild ideas, eh? That is a good article there. Auto-bias using BJTs....

 

[baudouin0]

With and without WCF. Gain goes up clipping level goes up, distortion goes down. Possible to drive 32R headphones with this. The HT hum actually get a bit better. Some good stuff in there.