Power amplifier tube matching

[W4AMV]

Question on the matching of tube characteristics for some circuits, such as push-pull and balanced output. Solid state devices such as bipolar and FETS are obtained at a price premium for such applications.

The question:

I have heard or it was stated in some literature piece that mis matched tubes can be operated say in P-P and over time they will match quite well. Is this true? Is there data or evidence that this occurs? That as the tubes load share and age that thier characterisitcs tend to balance and match?? It would be nice to see for example, an IMD plot or 2nd harmonic suppression of an un matched tube system that over operating time shows improved performance.  The same for SS??

Alan

[W7TFO]

A traditional way was to have individual bias controls for each tube, so quiescent current could be adjusted equally, especially found in modulators and HI-FI amps.

No reason you couldn't use that technique at RF as well.

73DG

[Opcom]

It may be difficult at RF, but you adjust the bias and drive to each tube so that the peak plate current is the same at the verge of clipping.
The average cathode current might be an easier way to approximate that if you do not run lots of grid current.

My point is that bias setting alone is not really enough, the tube needs to be worked.

If you can only adjust the bias, and are running grid current, the average cathode current is still an OK method if the grid current is small or you know what it is via the grid current meter, because you must subtract it from the cathode current. I have done this with parallel 807's in RF (grid current), and with quads of them in audio (no grid current). My 807's come from a couple of 5 gallon buckets.

A 10 Ohm resistor in the cathode circuit and an analog mechanical type milliamp/voltmeter makes it simple, be sure to bypass it (and the cathode resistor at the socket)! I installed little pin jacks in the chassis skirt of the audio amps to be used for measuring during the adjustment.

If it is a screen grid tube, that will also show up in the cathode current but it tends to be close between tubes because large screen currents are related to high instantaneous plate current (when the plate voltage is low the screen current is more) and that did not seem to me to vary much between tubes, 807's anyway.

After this, do not expect the bias'es to match or the standing current (if any) or the "partial power" current, to match up. Only the current at peak power is important. You may have to accept the fact that one tube won't do as much as another no matter how hard you push it. You can either set them equal or let the better one do more. In parallel it does not matter but in push pull they should be equalized up and you must live with the lowest one for both currents.

If you do put a bypassed 10 Ohm resistor in each cathode lead, and a way to conveniently test the current, then you can also easily match or grade your tubes based on a known drive and bias. Just use a known good tube (the best one) in socket A and plug the mystery tube in socket B and note the plate current with all things equal, and write that on the base. or, just use one tube, no matter as long as all other conditions are equal and the tube is pushed hard as normal, not just idling.

[k4kyv]

I run a push-pull rf final with split stator tank capacitor. Sometimes I would have a slight unbalance, even with a pair of well matched tubes. The load sometimes affects balance.  I use a separate plate current meter for each tube, so even a slight unbalance in readily observable, and I tend to be a perfectionist in such matters.

I would shuffle tubes, loads and any other factor I could control to achieve the best balance, but sometimes it was impossible to get the exact same plate current for each tube. I don't like the idea of adjusting grid bias to achieve balance, since with that method, once balance is achieved, you are operating the two tubes in the final at different conduction angles and therefore at different efficiencies and modulation linearities.   So I installed a capacitive balance control.  It consists of a small differential capacitor, about 12 pf per section, spaced for at least the same break-down voltage as the main tuning capacitor.  The differential capacitor is wired in parallel with the main dual-section capacitor.

The final tank cap is 240pf per section.  With the 12 pf/section differential, I have been able to get perfect balance even on 160m. I made the differential cap from the dual-ganged neutralising capacitor assembly taken from a BC-339 military transmitter that uses a pair of 833As in the final.  I simply reversed one section 180° so that it is at minimum capacitance when the other section is at maximum.

There are several causes for unbalance in a push-pull circuit.  The tubes may be mismatched, the split stator capacitor is rarely perfectly balanced, and even if so, stray capacitances in the circuit may be different for each of the two tubes. The capacitive circuit allows fine-tuning the relative loading of each tube so that they both draw the same plate current without fiddling with the bias, grid drive or neutralisation settings.

[MikeKE0ZUinkcmo]

I have a couple of NCL-2000 linears and matched tubes are difficult to find and when you find a pair, they are a lot more money than I want to spend.   I duplicated the original grid bias circuit and run separate bias to each tube, equalizing the Idle current for the two tubes.  Been this way for about a year now and all seems well.

[KM1H]

With several NCL-2000's here and getting them in regularly for service I modified the stock bias circuit decades ago to use dual pots.

With the original bias circuit a deviation maximum of 1.5V has no effect on IMD on SSB. Above that it decreases rapidly.

With dual controls 3V is where the IMD starts declining and 4V is down to -25dB 3rd. With the G3SEK screen and grid regulator circuit there is an additional 3-5dB improvement to the basic amp but the only amp I have with that included is the 6M conversion which has a pair of new Burle matched 8122's from a Burle employee who I did some work for on another project. With that setup the IMD is -33dB at almost 1500W PEP, those tube are especially "hot", but I run at 1200W.

Unmatched tube pairs without available adjustments will eventually even out since one will always be running much harder. It will also continue to get weaker faster.

Ive seen very little IMD change between matched and unmatched 572B and 3-500 pairs or quads in the case of 572B's in GG circuits. Ive even mixed brands and country of origin.

Carl

[Opcom]

... I installed a capacitive balance control.  It consists of a small differential capacitor, about 12 pf per section, spaced for at least the same break-down voltage as the main tuning capacitor.  The differential capacitor is wired in parallel with the main dual-section capacitor.

The final tank cap is 240pf per section.  With the 12 pf/section differential, I have been able to get perfect balance even on 160m. I made the differential cap from the dual-ganged neutralising capacitor assembly taken from a BC-339 military transmitter that uses a pair of 833As in the final.  I simply reversed one section 180° so that it is at minimum capacitance when the other section is at maximum.

That is a very interesting way to do it!  Is there any concern about one tube being 'less resonated' or, tuned to a different frequency, than the other? Or is the overall tuning of the tank with the big capacitor what matters?

[k4kyv]

Is there any concern about one tube being 'less resonated' or, tuned to a different frequency, than the other? Or is the overall tuning of the tank with the big capacitor what matters?

The resonant frequency of the tank circuit is determined by the total capacitance across the whole coil.  In the case of the balanced circuit with split stator capacitor, the total capacitance is that of the two sections in series, plus tube and stray capacitances.  If there is a slight difference in effective capacitances of the two sections, this will cause an unbalance in the push-pull final.  The two capacitor sections in series sections act as a capacitive voltage divider which defines how the two tubes share the load, and should be exactly the same (assuming the tubes are balanced).  The differential capacitor corrects any unbalance.  Since the capacitance of the differential cap is only about 5% of the maximum capacitance of the main cap, its setting will have a negligible effect on the resonant frequency.  With mine, even on 40m with the main cap almost fully unmeshed, adjusting the balance control has negligible effect on the plate current resonance dip.

Very few split stator caps are built precisely enough to have exactly the same capacitance per section throughout their range.  I have a  little solid-state capacitance meter that  reads down to 1 pf or so, and most split stators measure a few pf difference between sections, at any setting. Stray circuit capacitances may upset the balance, even if the split stator is perfect.  Also, this control allows one to compensate for a small unbalance in the tubes.

Regarding one tube being 'less resonated' or, tuned to a different frequency, than the other, that is what happens when the mid-tap of the coil in the balanced tank is strapped directly to the rotor/frame of the split stator cap, or when both are grounded.  This is why one or the other should be left floating. Usually, it is easier to maintain balance by RF-grounding the split stator cap and leave the mid-tap of the coil floating at RF, feeding the +HV or DC grounding to the coil through an RF choke.

[Opcom]

Don is a genius.