Anti-Triode SEPP, how to do best?

[Michael Koster]

Originally posted by revintage


Someone else can do that! Just joking, will do a sim for you when I get home from work.

MJK, do you have any ideas of what value the two "source" resistors in the triode-antitriode configuration ideally should be? Reread the thread and found you presented the constantvoltage version already at page one.

My main concern is if the internal caps of the DN2540 really shows up on the sims? I use the Supertex Spice model.

But at the moment my own interest is to find an "el cheapo" sand-replacement for an anodechoke.

So a totem pole circuit is indicated.

Here are all the variations I can think of right now, ignoring
for the moment the common anode circuits. For each of
constant current, constant voltage there are 3 basic modes:

1. "Mu-Follower" with very low impedance and current source
capability. I like this mode for driving a power grid tube in
class A2 or AB2. The triode operates on a flat load line.

2. "Gyrator" simulates an inductor. Can source current into the
load approaching Iq. The triode will "see" 100% of the load.

3. Anti-Triode is meant to drive a load such as a loudspeaker
or 600 ohm line input, where symmetric drive, low impedance,
2X idle current drive capability, and power supply efficiency are
desired. The triode and MOSFET equally share the load 50/50.

The resistor values for the anti-triode might be optimally chosen
to be different by a factor of 1/gm of the MOSFET, but in practice
I don't think it matters much.

I would simulate the anti-triode circuit as a parafeed output stage
in comparison to choke-parafeed and CCS-parafeed. The Anti-Triode
will have 1/2 the Zout though, so I would adjust the load accordingly.

Perhaps it should be compared with PSE, as it produces equivalent
outpower and has equivalent Rp

The AT circuit seems to work well with damping factors ~4 and above
IRL.

In summary, for grid drive I would try the mu-follower, for choke
substitution the gyrator, and for a line driver or output stage I
would use Anti-Triode mode.

Cheers,

Michael

PS I excluded the common-anode variants (hybrid-super-cathode-followers)
but there are three variations having different drive characteristics
as well.

[revintage]

Hi Michel,
So now you got six alternatives !
There are a few of them to eliminate. Lets stick to what Salas asked for. Think it was about preamp output. Wich ones would you consider?

Kenpeter,
Remade the the tests at 1kohm load. Not so much more impressing with the triode anti-triode, even if it was slightly better.

[Michael Koster]

Originally posted by Salas
Hi guys

I have made a couple of variable anode CCS sources. One cascode enhancement Mosfet LED biased of mine, and one Jung's LM317 10M45S cascoded. I am running experiments on my 6V6 trioded line pre. My B+ is 330V regulated and there is a capacitor bypassed 680R cathode resistor installed. I measured much better THD at the same bias point where the previous 5K resistive load was sitting (22mA). Sounded dynamic and clean, but in the long run I miss the previous tamer but more natural tone. It got more SS taste, and it got a bit of more hiss on my 95db speakers. When I took it down to 15mA, it backed up on hiss and SS tone, but still the same impressions. I wanna retain the dynamic and THD gains, but bring back some triode tone. I guess that valves can be pushed a bit with anode CCS and hiss if not lower current than previously with resistor is chosen? OK, the Mu is pushed now, but I have got more hiss compared to gain.
It reminded me of when running a noval hotter until it starts to simply hiss. Anyway, that is why I wanna listen to the Anti-triode before I go back to resistor loading. How it can be practically done best? I mean, just put two resistors between CCS and anode, and tap off from the center for output? OK then, is there a formula to calculate, given the B+, freely chosen CCS current, and given Rk as I described above?

EDIT: Here is the chart of 6V6 trioded. Suggestions for anti-triode?

The hiss is an indication of RF oscillation. It is a frequent problem
with CCS or MOSFET use, but should be able to be eliminated
through short wires, gate and grid stoppers, etc. Normal random
noise levels should be practically inaudible in a line stage, assuming
reasonable gain/level structure.

I'm not sure, however, if anti-triode will give you back the triode
tone you seek in a line amp. The resistor is probably best for that.
I think of a steep load line with some asymmetry in the swing.

Using either a mu-follower or anti-triode will flatten the load line
even more than the gyrator, resulting in less even harmonic distortion
and less triode tone. There might even be more higher order
distortion evident with the flatter load line (especially assuming
6S45pi).

If you do want to try a "real" anti-triode SEPP circuit, I think
the anti-triode operation pretty much depends on nearly equal
resistance either side of the tap. Either the constant voltage or
constant current anti-triode circuit would work. For CC, set the total
resistance as you would a CCS current set resistor. For CV, I would
start with something that gives about 2V total drop at idle and
maybe tune by ear. As ken mentions, CV does not require a d-mode
device.

Cheers,

Michael

[Salas]

Thanks for your comments.

[Michael Koster]

I just thought of a reason the constant voltage circuits might sound
different from the constant current ones.

It has to do with the average current shift that occurs in a nonlinear
(e.g.triode) power amplifier at large signal conditions.

A constant current load responds to this by reducing the anode
voltage at large signal conditions such that the average current is the
same as the idle current. This is in effect a compressor, as reducing
the anode voltage will lower the gain under most conditions.

The constant voltage circuits (Wavebourne, gyrator, CV above...),
on the other hand, may have a more linear dynamic response
due to the average current being allowed to increase as signal swing
increases, at a constant average anode voltage.

I don't know how large the effect may be, but it may be a good reason
to try both.

Michael

[Wavebourn]

So, how many kits to prepare?
I'm going to order PCBs for 2 types of gyrators: like mine from a mic preamp (cascoded), but more flexible, for adjustable voltage drop and different power dissipations (like to load 6V6), and one with 1-FET like Michael suggested.
The price per PCB will highly depend on the number of them I order.

[revintage]

As ken mentions, CV does not require a d-mode

Once again, what´s wrong with depletion mode?

Think I´ll make an IRL test with the one-transistor circuit I showed earlier. Will try two versions: 2SK2700 and DN2540.

Will also come back with a simmed comparision between IRF610 and DN2540.

Did some PSRR sims yesterday and it seems like the N-channel versions are better.

The constant voltage circuits (Wavebourne, gyrator, CV above...),on the other hand, may have a more linear dynamic response
due to the average current being allowed to increase as signal swing
increases, at a constant average anode voltage.

In other words, the CV lets the triode breathe easier.

[Michael Koster]

Originally posted by revintage


Once again, what´s wrong with depletion mode?

Think I´ll make an IRL test with the one-transistor circuit I showed earlier. Will try two versions: 2SK2700 and DN2540.

Will also come back with a simmed comparision between IRF610 and DN2540.

Did some PSRR sims yesterday and it seems like the N-channel versions are better.



In other words, the CV lets the triode breathe easier.

Hi Lars,

I don't think there's anything wrong with d-mode devices; in fact,
I rather like the way the DN2540 transfer curve plays against triodes
from 0-100mA. The gfs increases as the current increases,
perhaps compensating for the gm decrease of the tube at low current
in anti-triode mode.

I would try a DN2540 against a 1N60 to see if either one sounds
better. I also like the idea of the DN2540 as the constant voltage
device in a cascode controlling an IGBT etc. I have a bag of them in
TO-92. I also have a bag of 2N7000s...

Interesting analogy; the triode breathes easier... I also think of CV
mode not cutting the triode any slack on the peaks; makes it work
a little harder.

Michael

[revintage]

This is an idea of a DC-coupled SE with gyrator-loaded driver.

Instead of DN2540, 2SK216 could be used. IRF610 is not recommended due to its high capacitance.

[Michael Koster]

Originally posted by revintage
This is an idea of a DC-coupled SE with gyrator-loaded driver.

Instead of DN2540, 2SK216 could be used. IRF610 is not recommended due to its high capacitance.

Excellent!

If you use Mu-follower mode, you can drive some grid current,
as in this amp I'm currently breadboarding (simplified schematic...)

Cheers,

Michael