6L6GC Power Tube Matching

[R.G.]

Let me interject my standard sermon on matching. When embarking on a journey, it's always good to know where you hope to get; otherwise, like the Cheshire Cat's advice to Alice, if you don't know where you're going, any road will do.

What, exactly, is being matched, and why?

Matching idle current is indeed an exercise in matching zero-signal current in the tubes. Depending on the amp's circuitry, especially how much feedback and at what frequencies the feedback is effective, this can produce less hum at idle. It also matches the idle current in the OT, which is designed (we hope) to have exactly equal idle currents in the two half-windings be such that they exactly cancel any zero-signal magnetic offsets in the core, and not lead to unbalanced clipping. But it's subtle.

Other things that could be matched are the DC cutoff point for the tube, which is transited twice each half cycle by one of the tubes. Getting rid of crossover is one of those things that bias is intended to do, so presumably having matched cutoff points would be a good thing. I have always wondered why no one looked at that.
Another thing that could be matched for is incremental AC gain, and the second order effect of matching the incremental AC gain's variation over signal levels. Having one output tube have a bigger AC gain means the half cycles are different sizes, presumably not good for fidelity, either. Of course, in some circuits and at some frequencies, the resulting even-order distortion might be thought to sound good. Hard to say. Quirks in the matching of change in AC gain with signal level will add little quirks to one or the other half cycles, again, subtle.
Another thing to be matched for might be the current at Vgk=0. That varies with construction too.

My thoughts on matching are that if you want your tubes matched, fix the circuit so it can force them to be matched. Instead of sorting through boxes of tubes for two (or four, or six) close-enough matched ones, or worse, paying someone to do it for you, stick in controls that can match whatever tubes you get. Put in independent bias controls per tube. It's easy, and a whole lot cheaper than buying matched sets. If you can discern the differences in AC gain per tube, or test for it, then diddle the high one down with a "master volume" per output tube. Again, easy, and a good test for whether your ears really do hear differences in output tube gain. Both of these may well tell you that you require better matching than you can buy, or alternately, that you don't care as much as the internet thinks you should.

[roberto]

Hi R.G.,

about the AC gain balance, I would like to ask you if the idea I've got can be more transparent than the multiple volume pots.
Can different voltages on power tube heaters cause different emissions and therefore different ac gains?

Thank you
Roberto

[R.G.]

Sure. If the shaping of the cathodes and plates are the same, there is a zone of cathode temperatures that will act more or less the same. This happens when the cathodes are pretty much identical shapes and can supply enough electrons to keep a cloud of electrons around them all the time. If the plates are nearly identical, much the same currents flow at identical voltages.

What is critical is the construction of the grid and its spacing from the cathode. Even tiny differences in spacing matter because it is the physical size and spacing of the grid wires, and its spacing from the cathode that determine gain in the active region. So tiny differences in grid wire spacing matter a lot to gain. With devices from the same manufacturing machine and the same manufacturing run on the same day, tubes are likely to be very much alike. Yesterday's run, a little different, and last year's old tubes, more different. Another maker's tubes of the same number, who knows? Even more, the consistency of the spacing of grid wires matters, because the flow of electrons is closed off or opened by the negative voltage on the grid. The electrons see the negative charge on the grid as the grid wires growing larger, a region around each wire that the electron will be repelled from. If the grid wires are not constant spacing, the sections of grid where they are closer together will shut off completely before the wider apart regions and the grid will effectively grow smaller or bigger with grid voltage.

Vacuum tubes are mechanical devices!

The transconductance (current change per grid voltage change) does change with heater voltage, but not hugely. It's probably more true in power tubes than signal tubes. But the biggie is keeping the space charge from being depleted.

Tinkering the heater voltage to adjust AC gains is a subtlety that's going to be hard to do. My own approach would be to simply raise or lower the signal to the grid to make the effective gain higher or lower.

[pdf64]

The electron cloud around the cathode is intended to operate in massive surplus, and it must not become depleted; if it does then the cathode will be damaged. I think Tomer describes the mechanism.
http://tubebooks.org/Books/Atwood/Tomer ... 0Tubes.pdf

[R.G.]

Yep.

[pdf64]

...Oh that makes sense, maybe that was a point the signal just started getting much quieter before cutoff then? One I specifically recall was one I was troubleshooting with your help sluckey, and it was really thin sounding and horrid, and you told me to increase the bias current. I don't recall the exact details, but I think the bias circuit was not set with the right resistor for that setup and I was getting too little current. Then I swapped a resistor, and was able to adjust it to a level that was producing real sound.

Or am I just remembering even that wrong? does it just go from 'normal clean tone' to 'nothing' in one fell swoop? how do you know the point where it is 'too cold' but 'not yet near cutoff' or is there no such thing? If so then there should never be a debate about this, you'd just set the bias slightly to the point where 'current is flowing' and forget it, no?...

The following is my understanding of things, hope it helps
Have a look at the plate characteristics chart at the bottom of p7 https://tubedata.altanatubes.com.br/she ... /6L6GC.pdf
At low plate current, the plate curves squish up together and also aren't as smooth. In an amp, when it's operated down there, that results in less linearity / more distortion (low order, eg 2nd, 3rd harmonics), sine waves don't look quite right on the scope. Bad news for hifi, but this is a guitar amp and it gets overdriven anyway, so a bit more distortion at lower signal levels shouldn't really be an issue, as long as we keep the tubes in the region where things happen smoothly.
What we definitely don't want in a class AB amp is both tubes to get pushed into cut off at the same moment, under any signal type / level below clipping, as that cause sharp changes in the waveform, which generates high order harmonics, eg 5th and above, which can sound unpleasant perhaps because some of them are actually dissonant to the fundamental tone. (This also may be referred to as zero crossing crossover distortion).
Hence bias should be set so there's at least enough plate current at idle to ensure that under load, HT sag / bias shift (ie due to grid rectification) don't push the amp into class B.
When pushed much beyond clipping, grid rectification will force class AB amps into 'class B' / 'class C', ie both tubes will be in cut off momentarily (conduction angle 180 or less); but that's a normal part of power amp overdrive, and we seem to be used to the higher order harmonics that adds, perhaps because they're in addition to a heap of lower order harmonics from the top and bottom of the wave being flattened off.

For a regular p-p amp, eg Fender 6L6GC / Marshall EL34, about 30mA idle cathode current is usually plenty to push any chance of class B operation into the long grass. However, it might be desirable to idle at higher current than that because -
1/ as plate current increases so does stage gain, and more gain always sounds 'better'.
2/ the tone can seem to become fuller as the stage is moved into the more linear area of operation; it may be that a lot of 3rd harmonic can make the tone too thin and 'reedy', so as idle current increases, the tone can seem to become fatter.

It's worth pointing out that at high signal levels, a class AB amp will always exhibit crossover distortion, due to the significant change in operating conditions when the signal pushes ONE tube into cutoff (may be referred to as changing from the class A area of operation to class B area of operation). The effective OT primary impedance drops to 1/4 of its previous level because half the primary windings are taken out of circuit, as there's no current flowing through them.
That will always cause a kink in the amp's transfer curves, causing distortion of the waveform.
It can't be removed by increasing the idle current, though it does become less apparent as idle current is increased, because -
3/ the stage becomes more linear.
4/ the crossover point is moved toward the peak of the waveform and becomes more difficult to perceive on a scope trace
5/ if there's a negative feedback around the power amp, more gain within the loop increases the ability of negative to correct the error.

The key thing is to set bias so that there's always some plate current flowing in one tube or the other, ie there's a class A 'area of operation'. If not, the amp is in class B, and the crossover distortion point is all the way down at 0V, and it becomes zero crossing crossover distortion; the amp will sound kinda broken, even a lot of negative feedback can't correct things.

I've read and played with cold bias vs hot bias and know 2 things I found:

1. colder bias means lower output wattage but longer tube life - and I personally didn't like the sound of a cooler biased amp
2. hotter bias means higher output wattage but shorter tube life - and I prefer it... (but around that 70% number for class AB to avoid tube abuse)

Depends on how it's measured, but bias doesn't affect the amp's power output (ie output voltage across nominal load at the onset of clipping).

If biased with a low idle plate current (close to class B), stage gain will be low, hence the amp will require a bigger input signal for it to achieve max power output.
Lower gain will reduce the effectiveness of any negative feedback to correct distortion.
Distortion will reduce the accuracy of non true rms meter used to measure the voltage at the amp's output (they're calibrated to expect a sine wave).
The nature of the distortion tends to reduce the heating power of the signal (the area under the curve is reduced), so even a true rms meter will measure the output as being lower than a true sine wave of the same Vpeak.

If biased with a high idle plate current (close to class A), all the points above will become reversed, and the amp's power output (ie output voltage across nominal load) will measure higher; but the Vpeak will be the same. May even be a little lower, as the heaver continuous loading on the HT will probably cause it to sag a little.

[pdf64]

I don’t see how it's ‘needed’ to idle the tubes at 60%? Plate dissipation is a limit, not a target, doesn’t affect transfer characteristics, tubes don’t work differently according to how hot their plates are.

But they do! The hotter the bias the larger the signal size where the power amp remains in class A.

This may seem pedantic, but the hotter bias causes higher plate current resulting in an increase in the class A area of operation. Yes the plate dissipation will inevitably increase, but (like the increased class A area of operation) that's just an artefact of the higher plate current.
In itself, the increased plate dissipation doesn't increase the class A area of operation; rather both are the product of increased plate current.
Similarly, biasing to idle the plate at its limit doesn't necessarily equate to class A operation.

[pompeiisneaks]

Oh wow, that was an outstanding explanation, thanks!

I especially seem to understand:

1/ as plate current increases so does stage gain, and more gain always sounds 'better'.
2/ the tone can seem to become fuller as the stage is moved into the more linear area of operation; it may be that a lot of 3rd harmonic can make the tone too thin and 'reedy', so as idle current increases, the tone can seem to become fatter.

a lot more, that seems to fit in a bit with what I was trying to say I think.

also I really like how you explained the confusion around how measurements mislead people into thinking the output is higher/lower than it really is due to improper measurements, thanks!

I think that's still going to take some time for my brain do digest, though, but I love this kind of insight!!!

~Phil

[pdf64]

Also 60% is not the limit, 70 is on a proper spec OT and output tube along with the plate V+ and even more so the screen voltage !

I recently found the one-electron site with its repository of RCA application notes.
This one, which introduces the design max rating system, seems relevant here http://www.one-electron.com/Archives/RC ... 0Tubes.pdf
p2 shows that tubes with a design centre rating system include an allowance (ie de-rating of the tube's limit) in order to accommodate for component and supply variations, whereas design max tubes don't.
Therefore those who set bias on design max tubes take on responsibility for including component and supply variations, which are already accounted for with design centre tubes.
From that, it seems apparent that if design centre tubes should be biased so as to idle at up to 70% of their plate limit, there's good reason to idle design max tubes at a somewhat lower % of their limit, in order to accommodate component variation (eg OT winding and speaker impedance tolerances) and supply voltage variation.