Spud-Assist: Totem-Pole Current-Mirror PP Hybrid

[Michael Koster]

Or; 1.5 watts from a 417A spud amp.

Here is a concept I've been playing with, which is a current-mirror
circuit using totem-pole (series push-pull) topology. The idea is to
drive an active load so as to produce a current swing into the
external load exactly opposite the amplifier tube signal current
swing, thus dividing the load between the amplifier tube and the
active load.

In the test amplifier schematic below, a 417A is used as the amplifier
and a DN2540 is the active load. The 68R resistors on either side of
the external load connection function as the AC load current divider
between the bottom device (tube) and top device (MOSFET). The top
68R also sets the idle current as per CCS. In practice, there is a little
loss due to the finite Gfs of the FET, and the R values can be tuned
for a load split > or < than 50/50 between tube and MOSFET.

When the tube is driven to greater conductance, current is sunk from
the load and the voltage drop across the lower 68R increases,
causing the MOSFET conductance to decrease by a proportionate
amount. Likewise when the tube is driven to less conduction, the
voltage drop across the lower 68R decreases, causing a
corresponding increase in MOSFET conduction, sourcing current into
the load.

In this way, the tube sees only a fraction of the load (here it's about
1/2) as the tube and MOSFET are basically in AC parallel driving the
output transformer. The tube sees about 4K anode load in this circuit
(actually closer to 3.5K due to the above mentioned loss).

There is an advantage over SRPP in that the signal current swing is
balanced for any external load. This makes it a good driver for
reactive loads, like the headphone/miniwatt amp here. Current
source capability could make it a good driver for an A2 power stage.
It's also a way to get better anode efficiency from CCS loaded
parafeed circuits.

When I get more time, I'll make some measurements of distortion
etc. but it's mostly f2 until you get over 1 watt, then some f3 and
above show up.

I'm actually able to get 10V p-p into 8 ohms (1.5 watts) by driving the
417A grid positive. The current mirror works fine into A2 as it senses
anode current and just follows the tube.

Michael

[Wavebourn]

It is very similar to what I played with few years ago, but a tranny was Flathead with 17 + 17V outputs, and a CCS on top was modulated feed-forward. The tube was 36L6 triode strapped.

[kenpeter]

I see exactly what you are driving at. Bottom triode sees a
constant current source, and xfrmr primary in series with 68R.
The plate swing is close but not quite exactly Mu due to the
non-infinate loading.

Top Depletion MOSFET follows the plate, and also drives the
primary thru 68R, the low 68R is seen as very high impedance
because the plate beneath it is at the same AC potential.
Very similar load as what the triode below is seeing.

The sum of currents (measured by the pair of 68R's) is always
held constant (whatever voltage drop turns on the MOSFET).
Thus the MOSFET on the top of the totem is operating in true
Anti-Triode emulation mode.

The output signal is quasi-SE, as the antitriode pulls with the
curvature bent backward of a "normal" complimentary device.

[kenpeter]

Should the OPT have pathed directly back to the cathode?
Instead of including the bias LED into the output's loop?

[Michael Koster]

Hi Ken,

> Should the OPT have pathed directly back to the cathode?
> Instead of including the bias LED into the output's loop?

I thought about that, but then the load current sourced from the
MOSFET would be returned through the LED. "Currently" the LED
sees the tube's share of the load. It's going to be in one loop or the
other and it's always in the input loop, so I thought it best to keep
in in the tube path and out of the MOSFET path.

> The plate swing is close but not quite exactly Mu due to the
> non-infinate loading.

Lightly loaded, as in a drive stage, this is true.

The load line seen by the triode will be defined by approximately 2X
the load impedance for equal value resistors. There is some leeway
to split the load unequally which is what I plan to experiment with
next. I want to try a 100R anode resistor and split the load 60/40.
Should be a little better damping factor, gain, and output swing.

The 417A is rated for 4.5 watts anode power and at that the anode
takes on a dull red glow... So I can abuse sand and tubes at the
same time!

Cheers,

Michael

[smoking-amp]

looks familiar
http://www.diyaudio.com/forums/showt...005#post638005

Since Ken later referenced the O'Schade article on square law tubes due to g1 island effect:
http://www.diyaudio.com/forums/showt...35#post1516035

I would guess now that the finite Fet loop gain is not going to cause much of a problem, since both devices are approx. square law. So maybe a little 2nd harmonic cancellation, but mostly the triode signature is going to be left intact.

Michael:
"When I get more time, I'll make some measurements of distortion
etc. but it's mostly f2 until you get over 1 watt, then some f3 and
above show up."

Should be interesting results, "sounds" like its on the right track.

I'm thinking now whether there could be any utility in using a pentode on the bottom with screen grid feedback to get triode sound. So far its just degenerating into a pentode wired triode or a source follower off the plate driving the screen grid (with a little DC drop), a little more thinking required. Idea would be to eliminate the Fet drive from the output altogether if possible.

Don

[Michael Koster]

Hi Don,

All variations on the same theme of complementary current drive.
I have built two amps based on the idea, trying to further simplify.

I'm cooking an H-bridge parafeed now that looks like your #3, but
with both triodes on the bottom. Looking for >30 watts from a pair
of 300B run conservatively.

I'm trying to see how simple I can make the circuits, got rid of
the op-amps and finding that the slight nonlinearity of the
MOSFET is not an issue as you say.

Here's a twist on #1 that probably can't get much simpler.
I'm going to repurpose my old anti-triode amp to test it.

Cheers!

Michael

[Sheldon]

That certainly is simple, which makes it attractive - and PP power from a single ended input. But how much problem will thermal issues be, due to the different affect on the transfer characteristics of the two devices? A class A amp should reach something close to a steady state, but that will vary with ambient conditions too. That should be manageable if the transformer can tolerate some imbalance. What about signal induced transients?

Sheldon

[Michael Koster]

Great points, Sheldon!

I expect some kind of diode string might help keep the relative
g-s voltage constant over temp. The tube Iq will drift also, so
I thought maybe a CCS on the B+, but CCS in series with the
inductance of the OPT makes it a noise catcher at best...
needs some fiddling with bypass caps, etc.

The other issue you mention is the real reason I'm going forward
with the series totem pole toplology for the moment.

The parallel circuit with OPT is stable in a steady-state sense but my
experiments with high loop gain (op-amps) in the current feedback
path show some instability under music conditions. I put that
investigation on the back burner but I believe it's the problem of
the MOSFET current coupling to the input loop through the common
current sensing resistor. I think there is a positive feedback loop
under some conditions I have yet to understand. Maybe it's just
due to the op amp gain and the simpler circuit will not have the
problem.

Also think I may go with an enhancement mode MOSFET and an
adjustable gate voltage, and in a practical amp a cascode may
work better.

Basically, I decided to think about things for a while and go
stupid simple to start with, and add only what's found to be
needed through testing and listening.

Michael

[Sheldon]

Originally posted by Michael Koster
Basically, I decided to think about things for a while and go
stupid simple to start with, and add only what's found to be
needed through testing and listening.

I like that, if for no other reason than I have better intuitive understanding of simple stuff. Might as well try the simplest possible, just might be OK. A cascode would take the heat off (so to speak) of the controlling FET.

Or, instead of a diode, could you use the same FET with gate tied to drain, for compensation?

Sheldon