Please help me improve my headphone amp

[kenpeter]

Diego wants 1.29V under his V1a (post#1), Darlington for your V3 would be about that...
NPN might offer better choices than PNP, MJE350 is so just barely getting it in mine...

If you could hold V1a at sufficiently constant current by Triodlington current dumping
method alone, the entire top triode can go away and become a simple resistor. It not
like V1a would see plate resistor and load currents anymore, only plate voltage and Mu.

[Wavebourn]

I will breadboard it later and report. Should be a nice amp. I even see a pair of GU-50 output...

[diego6339]

Sorry for the late response. To be honest you totally lost me as I´m just a noob at this.

The voltage stated in the first schematic are taken from the real amp. I built it following a design without changing anything. So, the bias of the 6sn7 does not need to be 1.29, it could be 1.7 or 2. The plate voltage needs to be so low because it sets the bias of the 6as7 which is now around -23, I can increase that voltage up to 60 and the 6as7 goes to around -19. That´s why I thought the ac coupling could be useful to increase the plate voltage even more. However, I don´t know if the coupling cap will damage the sound more than the gain obtained by running the 6sn7 at higher current.

In any case, I already built the amp so I´m restricted to the same B+ and the 3 tubes and a bit of space restrictions.

[kenpeter]

If using all same tubes and no sand, go with schematic of post #12.

Yet we are still in process exhausting other options. In Post #1, Diego
threw sand at the problem of CCS for V1a. I propose the same, but in
a totally different way. Taking Wavebourn's idea one possible extreme.

Here: Q1 & Q2 hold V1a cathode at constant current by dumping any
current variance. The Triode experiences only voltage variations, and
operates on nearly flat Mu loadline.

Oddly enough: This is also sort of shunt regulator, slaved to Mu. And
improves power supply rejection observed between plate and ground.

V1a can't tell any different that the current regulation is parallel shunt
rather than series... And we make V1a appear as-if a LOT bigger Triode.

[regal]

Your output coupling cap looks too low, calculate the -3db with your headphones

[kenpeter]

300ohms = 1.3Hz
32ohms = 9.8Hz
8ohms = 38.4Hz

And keep in mind the output impedance is not 1.2K
Invert factor we used to boost parallel current.
It is plate resistance (7K?) divided by about 35.
So whats that now, maybe 200 ohms???
With 1.2K in parallel, maybe 170ohms...

Not even close to as low as I thought it'd be.
I'm almost sure the collector follower impedance is
lower than that. Where did my math go wrong?

Unless boost to quiescent current is different then
boost to Gm? I've went and lost myself again...

[diego6339]

As I had it before with 300 ohm headphones is 11.29 but I think I can squeeze in a 100uf MKP for 5.31hz.

[kenpeter]

No cap is perfect, all have other strange things going on besides capacitance.
Sometimes those other effects are significant. Mixing cap types and/or values
helps to flatten out those quirks. Almost always closer to ideal than a single
large cap.

Since you obviously stuck on MKP (and nothing wrong with that) I'd suggest a
pair or more of smaller ones. And none the same value.

--------------

I totally computed output impedance wrongly for the circuit of Post#24.
Quiescent current magnification and Gm magnification are NOT the same.

Plate voltage is set by total quiescent current, resistor drop from B+,
and location on the triode curve that places 1.3V under the cathode.

Triode current is set by 1.3V shunted across 750 ohms. The rest must
come from sand assist. Sand will pull the plate voltage down untill the
rule of 1/(Mu-1) from plate to cathode is satisfied. This voltage rule is
modulated by Mu*whatever is happening at the grid of course...

Triode plate impedance is divided by HFE of the darlington pair.
HFE is highly variable with temperature and manufacturing tolerance.
25x30=750 only the garaunteed minimum for this pair, could be higher.
Fair to say 6300 going by the "typical" 25degree published curves.
What exact impedance does this get us? Dunno... 1 ohm therabouts?
Lets just say it will be plenty low enough to drive headphones.

[kenpeter]

Lumping headphone in parallel with 1.2K plate resistor (if we regard B+
another virtual ground) is about 240ohms load. But triode gets at least
750x Gm assist from the sand. So sees loadline slope as-if it were 180K.
And this is worst case, bottom of the barrel, almost reject transistor.

Consider Gm boost with typical HFE is more likely 6300x. Loadline slope
could be 1.5M or higher! Begins to look like a CCS. The triode does not
know or care that CCS effect comes from shunt rather than series.

[Wavebourn]

Actually, your Darlington stage has 1k2 output resistance; voltage amplification factor with no load is 1.6. I played with such (well, similar) beasts, they oscillate like crazy, that's why I went with pentodes in my Barracuda project.