Triode PP OP stage combining Blumlein Garter and CCS bias - anyone tried this?

[ray_moth]

It occurs to me that it might be good to combine the advantages of Garter bias (good DC balance) and CCS bias (good AC balance) for Class A triode PP. I'm not sure how this could be achieved in practice but I'd love to see ideas on this.

[kenpeter]

No promises... You tell me?

[Shoog]

Originally Posted by ray_moth

It occurs to me that it might be good to combine the advantages of Garter bias (good DC balance) and CCS bias (good AC balance) for Class A triode PP. I'm not sure how this could be achieved in practice but I'd love to see ideas on this.

Done with bypassing referenced to each other, the CCS addresses both AC and DC balance, but you are restricted to class A.

Shoog

[smoking-amp]

I think Ken's drawing will just supply constant current (and zero current at that unless they are depletion mode, 10M45 is depl.) to each tube since there is no feedback path that gets around the Hi-Z Fet drains. The Fets are doing there own thing, could care less what the tubes are doing.

I think it needs to have the Fets and Source resistors turned upside down and use P channel parts. Then the tube currents will affect the Fet biasing. Maybe have to move the grid bias pickoffs up to the resistors too, then maybe not, could work as is if too much voltage is not lost in the Fets.

[kenpeter]

Split tail pair has two DC constant current sources. And in this case,
Garter matched 10M45S sources no less! (yes they are depletion w.
approximately -5Vgs threshold)

Originally Posted by smoking-amp

since there is no feedback path that gets around the Hi-Z Fet drains.

But we also have a capacitive path across the top, for AC currents
to pass back and forth, from one cathode to the other, in Class A...
Just as they would for AC with an ordinary single CCS, long tail pair.

The usual Garter bypass caps to ground would be redundant to the
function of caps bridged across the top. Adding caps to GND would
inhibit forced AC balance that Ray specifically asked for.

The example above would pass about 100mADC quiescent per tail.
Yet due to the AC cap bridgeage, both tails could be servicing one
triode to swing 199mA while the other swings 1mA. Then vice versa...

The triodes must require a negative bias high enough to drop 10V in
the garter and at least another 10V (or is it 15V?) in the drain. So
we not talking a circuit ready for just any old random triode. It has
to be one that likes at least -25V on the grid, even at highest peak.

I can show you slightly more complicated split-tail of garter matched
Aleph current sources. If -25V bothers you, the Aleph only needs 5V
to function. I showed you the 10M45S first, because it was simplest.

[kenpeter]

My mistake: According to 10M45S spec sheet, 100mA is achieved at -2Vgk.
The -5Vgk spec I recalled from flawed memory, was for 0mA total cutoff...

The resistors in the garter would be only 20 ohms. And only 4 volts dropped
in the garter. The spec sheet is far from clear about how many volts needed
to be standing on the anode to achieve 100mA @ -2Vgk? The smallest number
given for an anode voltage is 10Vak. Yet current vs Vgk curve was probably
drawn for conditions at 450Vak. I'm thinking 15Vak+4Vgarter is still safe bet.

So thats 19VDC cathode bias, bare minimum to maintain regulation in garter.
I still need to show you the Aleph version.

[ray_moth]

Originally Posted by Shoog

Done with bypassing referenced to each other, the CCS addresses both AC and DC balance, but you are restricted to class A.
Shoog

Agreed, Class A is enforced by a common CCS. I strongly suspect the same is true with Garter bias too. That's fine with me.

[Shoog]

I have used the garter in one circuit with bypassing to each other. It achieves current matching of less than 1mA per side - remarkable really. However AC balance has to be inferior to the CCS versions I have also used.

Shoog

[Michael Koster]

My experience with 10M45s is that they won't go to 100mA without Vgk going positive. They don't follow the published curve at the higher currents, at least at up to 50Vak. By that voltage the Vak should be saturated.

I tested the newer IXTP01N100D from IXYS and it easily reached 100mA at 30Vds.

I don't know why they call the terminals "A" and "K" either, being a depletion mode N-MOSFET it has a drain, a gate, and a source...

(the garter circuit) "AC balance has to be inferior to the CCS versions"

Shoog, what is it about AC performance that has to be inferior and why? I don't get it.

[kenpeter]

Enhancement mode opens more MOSFET choices.

IRFP240 (200V) probably works fine here? No sense to
specify drain Voltage that greatly exceeds the cathode
bias and/or bridge electrolytic ratings. Drawing shows
a 600V mosfet, is probably overkill.

This the Aleph version I was goin' on about earlier...
You need about 1.34 Volts to bias the garter. Then
whatever small amount dropped in RDSon.

Though several wizards of solid scrape have warned
against operating MOSFET with fewer volts on drain
than gate. Miller capacitance Cdg goes up greatly,
so CCS may react more slowly if not enough Volts
are present. Give it about 9V to properly shape the
depletion region, and Miller should be OK.

Talking about minimum 10V cathode bias if we want
this to really work well. I could claim it works to 2V,
but might be kinda mushy for reasons as given...