What's the advantage of E-linear

[Michael Koster]

Specifically, what's the advantage of E-linear over simple plate-grid feedback.
As I surmise, e-linear is essentially ultralinear plus G2-G1 feedback. In what
way would this be superior to plate-grid feedback?

Thanks,

Michael

[SpreadSpectrum]

I would think that any time you start moving the screen voltage around, you lose efficiency. By that, I mean you will have to start dissipating power in the control grid to get the same maximum power.

I would be curious to see curves drawn for an e-linear setup and a plate-grid feedback setup with similar 'mu' target. I would bet that the simple plate-grid feedback would have higher max power without grid current. Would e-linear have some advantage as far a linearity goes? Unfortunately, I am still young and haven't had time to build a curve tracer yet.

So, I don't really know but am curious as well.

[jon_010101]

To first order, when pentode drivers are used and screens are connected to the taps, e-linear functions similarly to cathode feedback, since the feedback signal is seen by the grid and screen. But it has the advantage of not requiring a tertiary winding. Of course, you can also disconnect the screens and run the output stage in pentode - which I think has some appeal.

It is thus also similar to plate-to-grid feedback when UL connections are used for the screen, and requires the same added drive capability. Of course, it is a slightly less-flexible arrangement. I think the largest benefit over plate-to-grid feedback would be simple convenience and elimination of plate-to-grid resistors. I'm sure Douglas can provide further comment. Within limits of stability, something like the Citation II feedback scheme offers additional benefits since it includes the driver stage in the loop. But again, that's a higher effort approach, requiring more tweaking.

[JoshK]

Good question(s). I've always been curious about this as well.

[Sch3mat1c]

It's just old fashioned shunt feedback. Easier to do it with one resistor from output plate to previous plate, plus you can control the amount.

Leave the screen alone, all it does is increase distortion relative to any other configuration of equal gain (including triodes and UL). The control grid is the most linear input and has earned its name with good reason.

Tim

[SpreadSpectrum]

I just finally have taken the time to study Schade's 'Beam Power Tubes'. I painstakingly plotted points for KT88 curves in Excel and then calculated curves for various plate to grid feedback ratios. I was struck by the fact that I could get much better linearity than a 300B with a higher mu. The cherry on top was the fact that I could also get much higher plate currents without control grid current along the same load line. (more power output)

Originally Posted by Sch3mat1c

The control grid is the most linear input and has earned its name with good reason.

What I don't understand is why do ultralinear circuits exist? After my curve drawing exercise, it is apparent that I can have my cake and eat it too. (Well, I do have to provide a screen supply, I guess.) Why would we give up the benefits of the accelerating grid(at a constant voltage) if we don't have to? To save a few components in a feedback network and screen supply? Is there a gotcha that I don't know about yet? As I said, I'm still young and haven't had a chance to build a plate to grid feedback project yet.

[Wavebourn]

I prefer low and stable like a rock voltage on a screen grid. Of course, in such case first grid bias must be regulated as well.

[Michael Koster]

Originally Posted by SpreadSpectrum

I just finally have taken the time to study Schade's 'Beam Power Tubes'. I painstakingly plotted points for KT88 curves in Excel and then calculated curves for various plate to grid feedback ratios. I was struck by the fact that I could get much better linearity than a 300B with a higher mu. The cherry on top was the fact that I could also get much higher plate currents without control grid current along the same load line. (more power output)



What I don't understand is why do ultralinear circuits exist? After my curve drawing exercise, it is apparent that I can have my cake and eat it too. (Well, I do have to provide a screen supply, I guess.) Why would we give up the benefits of the accelerating grid(at a constant voltage) if we don't have to? To save a few components in a feedback network and screen supply? Is there a gotcha that I don't know about yet? As I said, I'm still young and haven't had a chance to build a plate to grid feedback project yet.

Why do ultralinear circuits exist? -- Good question

A better question might be, why did the Hi-Fi world go with UL in such a big
way rather than the scheme the "inventors of the beam power tube" advocated?

Why is plate feedback not way more common?

One answer might be that Williamson and UL amplifier designs were successfully
marketed as solutions to the internal resistance problem of beam power tubes.

Behind this might be that O.H. Schade's reference design used an interstage
transformer to sum the local feedback with the grid signal. That wouldn't have
been super popular with manufacturers, leading to the marketing of alternate
designs.

I think this may have postponed the further development of O.H. Schade's
design, although there were a good number of O.H. Schade-style designs
through the years; I run across one now and then.

When did direct feedback to the driver plate show up?

[Michael Koster]

Originally Posted by Wavebourn

I prefer low and stable like a rock voltage on a screen grid. Of course, in such case first grid bias must be regulated as well.

Yes, I think that's the approach.

Add local feedback and you have a nice triode. Adjust local feedback to optimize
mu and Rp for your circuit (gm=mu/Rp). A great shortcut for ballpark estimates
is to set the "mu" of the stage by adjusting the feedback, then calculate Rp by
mu/gm where you just use the tetrode gm for your Vg2... The error is that mu
and gain are not identical for triode power output stages.

I wonder how stable the high-gm tetrodes would be with well regulated g2 and
fixed g1. It "feels" like an additional balancer, e.g. servo, would be needed.

On a related subject, who has studied O.H. Schade's discussion of beam power
tubes vs. triodes for class AB operation? Pretty compelling if you work some
case studies. The gist is that tetrodes are more suited to AB than are low
impedance triodes, and that low impedance triodes are best run class B.
(well you need to remember they's selling beam power tubes here ;-) )
But I think the basic reasoning makes sense *especially* in the context of
local feedback across a class AB P-P stage. Constant gm across 2 tubes
through the crossover region, linearized by the local feedback.

Woulld a bifilar OPT be any advantage in a tetrode AB stage with local FB,
as it would in class B with triodes?

Michael

[gingertube]

With Anode to Grid feedback the feedback resistor is effectively in parallel with the Miller Capacitance. This puts a definite upper limit on the size of resistor you can use and still maintain reasonable high frequency response.

In my Baby Huey I'm experimenting with Ultralinear PLUS screen to grid feedback. No definitive conclusions yet. Only listening so far, I need to make some measurements.
Cheers,
Ian