SEPTOR AT-amp (Anti-Triode) being built

[revintage]

After some planning the SEPTOR is on its way. It will be a rebuild of some old Erik Andersson(Audio Innovations, Audion etc.) 6B4G PPP monoblocks.

After the discussions in some other threads this is what came out.
Note that the cathode resistor is substituted with a CCS. The toroid is a Talema with rewound secondaries. If the sims are right, as they usually are, there will be no problem to take out over 12W.

In the pic the KT66 is arranged instead of the EL519 as it is still in my warehouse.

[ErikdeBest]

Hi Lars

Looks interesting, very interesting actually. Just some questions:
1 - what CCS will you use for I1? Will a simple LM317 do?
2 - what is the size (capacitance) of C1?
3 - have you also simulated the distortion characteristics?

Succes with the built, I saved the thread for future reference (and experiments).

Erik

[revintage]

Hi Erik,
Did not add any component values as I hope someone else will also do some testing on their own. You could probably use a 317 but remember it is ca 45V over the CCS so a series resistor might be needed. The cap works like a parafeed so you have to do some experimenting.

[kenpeter]

Have you looked into the "Wilson" 3 and 4 transistor current
mirrors? My early attempts with simplified 2 transistor mirrors
never quite seem to track each other with any linearity.

http://en.wikipedia.org/wiki/Wilson_current_source

You might also consider a parallel pair of IXCP10M45S, not
a true "mirror", but good and stable, very easy to match...

[ErikdeBest]

Hi Lars

Thanks for your comments, I think I know enough for the moment. As said, the thread is saved and I have some PP iron that needs to be put to good use.

Kenpeter

That 4 transistor Wilson current mirror looks very interesting, but I did not quite understand if you compared the performance with the 2 transistor mirror? Did you? For me simple transistors work better, as the 10M45S costs more and needs to be sourced from a USA based shop.

Erik

[kenpeter]

In ALL my simple two transistor mirrors I've ever tried (maybe 4
attempts in total) unless an external sense resistor was used,
currents in the mirror branch do not linearly follow those of the
reference.

There's supposed to be intrinsic "Shockley" 26ohm impedance
to the emitter. And all bipolar transistors are supposed to be
the same in this regard. But in my experience, nobody seems
to have told any actual devices about this rule.

Obviously, the sand wizards that be have mastered this black
art for devices that share a common die (such as in op-amps).
Though they often end up abusing a Wilson anyway.

Linearity of followage in Wilson's 3 transistor variant may be
good enough, won't usually need any extra emitter resistors.
I'd use em anyway, just incase...

[kenpeter]

OPT balance will be in error by the amount of screen current.
I don't know if the amount we are talking is significant? To a
Toroidal type, maybe any error is significant... Taking screen
from the plate (through an RC low pass) ought to fix it.

[smoking-amp]

How far the screen V drops will affect how much plate current it siphons off, of course. The usual horizontal output tubes with aligned grids spec out at 5 to 10%.

Putting an RC from plate to screen will just draw the AC screen current off to ground. (I'm assuming that's where Ken is connecting the cap.) What's needed is a HV Mosfet follower controlling the screen V with it's drain connected to the plate. (returning the current). Simplest of course would be to just use a Mosfet in the first place for the "anti-triode".

Don

[kenpeter]

Resistor from plate to screen, Cap from screen to ground.
Or optionally, from "center" of the screen resistor to GND.
Cap is what prevents becoming a simple triode strapping.
Holding screen voltage steady, just a little less than B+.
Even though the plate may regularly dip below this level.

Where else then can DC current come from, or go to?
Only from current source up to plate, where all then is
properly accounted for in the OPT. My last drawing
illustrates all three common strappings with cap to
detour the DC component through the full winding..
That drawing may have been in another thread...

When screen current bypasses the OPT, that was the
only error of DC balance I was suggesting to correct.
I don't see how Don's suggestion of FET with drain tied
to plate, holds the screen up when the plate dips? You
still need a cap in there somewhere...

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

MOSFET is a good choice if you need the Anti-Triode to
faithfully fake pure parallel SE. But sand is certainly not
the only option that argues good sense.

Pentode blended may offer a very useful transfer curve
profile for an output stage. Nice and linear across the
middle, yet dissimilarly squashed on each end... I really
want to hear how this amp might sing at full volume.

[smoking-amp]

Oh, OK. I follow you now. For some reason I was thinking AC current imbalance would be a problem too. Well, it would be if it causes distortion, ie non-linear screen current.

Hmmm, let's see, looks like the screen current increase (so pentode plate current decrease) at large signal level will sum with the triode side's increasing current draw at large signal level (non-linear effect). This will cause some signal envelope related current imbalance in the xfmr. Too bad the screen doesn't act the other way and cancel that problem instead of contributing to it.

Don