The AM Forum

Question:

Suppose you were going to build a grounded grid AB2 rf linear amplifier, using a matched pair of triodes (3-500Z or whatever your favorite triode is)...

What are the engineering/practical reasons why you would use the tubes in parallel, rather than in push pull or vice-versa?

Best regards
Stu
AB2EZ

2 advantages of GG jump out immediatly Stu.

Phase and simplicity.

In PP you'll need to deal with phase of the drive signal, neutrilization ect....

GG parallel of zero bias triodes is much easier. Very simple circuits even if you have to run a tuned input for the Kenyacom radios.

once you go solid state push pull wins hands down

With grounded grid, there usually is no need for neutralisation, whether pushpull or parallel.

With pushpull, the balanced output tank circuit is more complicated than a single-ended pi-network, particularly for band changing.  However, the large (and usually expensive) rf bandswich is eliminated if you use plug-in coils, which are more practical for the pushpull circuit. Also, some kind of unbalanced-to-balanced input coupling system is required.

But particularly at higher frequencies, with pushpull it is easier to maintain the proper Q of the tank circuit, since the effect of tube capacitance is 1/4 that of the same two tubes wired in parallel.  Also, it is easier to balance the two tubes, using a small differential capacitor in parallel with the main split stator tank capacitor.  That capacitor need be only a few pf per section, with at least the same voltage rating as the main cap.  It is easy enough to construct one at home if there is nothing in the junkbox. In one of my transmitters I use two small neutralisation capacitors with the rotor shafts coupled together mechanically.

Another advantage of pushpull is that the tubes supply the full rf sinewave to the tank, instead of mere half cycles, and therefore it doesn't have to rely on the flywheel effect to supply the missing half cycle.  This can be an advantage at 160m and 80m, because you can get away with using less Q in the output tank and thus a smaller variable tank capacitor.

With the pushpull circuit, link coupling is more practical than the pi-network, although a balanced pi-network circuit can be constructed if you have a split stator plate and loading capacitor, and you split the tank coil.  I find variable link coupling more convenient than a pi-network for loading the transmitter, because with everything tuned to resonance, you don't have to retune the plate tank capacitor with every re-adjustment of the loading capacitor.  You just move the link in or out until you read proper plate current, and the plate tank circuit will still be perfectly dipped, if the transmitter is properly coupled to a purely resistive load.

A well balanced pushpull rf amplifier inherently cancels even harmonics, so there is less of a problem with the troublesome 2nd harmonic that amateurs often have with their transmitters.

I have always used pushpull in my higher powered homebrew transmitters, even though mine are class-C grid driven.  All things considered, nothing can beat the overall simplicity and efficiency of the classic pushpull triode final with plug-in coils and link coupling.

Julius
Frank
Don

Thanks for the quick responses to my question.

Have any of you heard of anyone using a ferrite of the kind used to make transformers for Class E rigs (typically a few ohms or a few tens of ohms in and out, and high current in that application) to wind a balun transformer with a high enough magnetizing inductance to be used in a tube application ... at an impedance level of ~2500 ohms in and out? I am assuming that the wire would have sufficient insulation to handle the high voltage between turns. I'm thinking of a minimum of 10 turns on the primary and a minimum of 10 turns of the secondary... with perhaps a modest step up or step down to facilitate impedance matching ... and still using a pi network on the unbalanced side of the transformer.

Stu

stew check out qro emcom/emcon something like that and uk amp they guy that makes commercial gs35 amps.  They use them on the inputs and if memory serves one of these uses them on the outputs.. The downfall is you need to wind one per band and due to the voltages @ legal limit they tent to be massive cores. 
dont quote me on any of this my mind is usually a fog running 2llc's and an inc

Who needs expensive band switches just use a roller inductor and some 10 amp micro relays about 4 bucks each for the input circuits

With a push-pull amp, it might be easy to make it into a giant balanced modulator at the flick of a switch ( and some audio injection)  ;D

.....Larry

Many commercial linear amps made in the last 10 or so years use a toroidal core based step down transformer on the output to reduce the capacitance requirements in the Pi-L output network. These are single-ended, but I don't see why a balanced setup wouldn't work similarly.

yep. I believe that claim to be true. if you got the drive, run triodes.

Oddly enough, I was pondering this very thought the other day. I have never seen a P/P grounded grid amp. I was wondering why. It would be a little more complicated than a parallel amp, but using plug-in coilz for the final tank would make it easier. I guess you would have to use either a balun of some type or a tuned P/P input transformer (plug in coil arrangement) to feed it. the low inpedance of feeding the cathode would be pretty interesting. Mayme some kind of broad band toroidal transfoma. ? ? ? ? ? ? ? ? ? ? ? ? ? ?

Another good thought to ponder.

                                                    the Slab Bacon

Or simply a balanced tank circuit link coupled to the exciter, feeding the cathodes through lo-Z taps on the coil or a close-coupled split coupling coil, adjusted to give you the proper excitation with the proper power input to the exciter stage.  A balun connected ass-backwards also might do the trick.  Just make sure that whatever you use can take the final amplifier DC cathode current, or else use a choke and coupling capacitor.  Of course, with directly-heated cathodes this would require two separate filament chokes, or chokes split into two symmetrical sections.  The filament choke and coupling coil functions could conceivably be accomplished with one unit.

Running push-pull reduces the effects of the interelectrode capacitances across the tuned circuits. These intelectrode capacitances are effectively in series instead of in parallel resulting in only half the capacitance of a single ended device and only one quarter of that of using two devices in Parallel. This helps to preserve or extend the useful frequency response of the amplifier. This effect can be seen in amateur transmitters that use conventional circuits. They tend to roll off in gain and output power on 15 and 10 Meters. If the driver and finals were run push-pull (hard to do by the way) the output power of the radio would be just as high on 10M as on 80M as long as the tank components were up to the task.

Here is a wideband transmitter: http://www.transmitter.be/gat-hf1.html
Notice how it uses push pull.

Running grounded grid is another way to extend frequency response. I think that using both together would be a very effective way to make a high power UHF amp.

At HF, my guess is that with a P-P G-G, the input impedance would be 4 times higher than with two tubes in parallel and that the drive requirements would be 4 times lower.

Mike WU2D

That was done in the past. I had a Bendix amp, probably 40-50's vintage that used two triodes in P/P. With tubes like the 4CX250 (good up to 500 MHz) and the many of the recently available Russian tubes like the GS-35 are good up to 1 GHz, the need for going P/P for UHF isn't nearly is great as it used to be.

And push-pull linear amps were also done in the past, most often to boost the output of a Heising modulated stage in transmitters of the late 20's and early 30's, before push-pull plate modulation schemes became common place.

You can see a cool buzzardly one and get the dope on it too at the link below.

http://www.arrl.org/tis/info/pdf/29hb098.pdf

Steve (I enjoyed the article that you provided a pointer to)
Mike
Larry
Bacon
Mack
Derb
Don
Bill

Thanks for the additional comments and pointers.

I performed an experiment this morning to test the feasibility of making a balun transformer using ferrite cores. I used what I had on hand, and made a few educated guesses... so this is far from optimal.

I took a pair of SB-1020-43 cores (made by CWS) and wound 10 turns of insulated #22 hookup wire on each core (separately)... using about 85% of the core's inner circumference for each of these coils. Then I linked them together by looping two turns of 5kV-rated insulated wire through the remaining space in both cores. So, in effect, I constructed a 10:10 rf transformer by using a 10:2 transformer in series with a 2:10 transformer. I used two separate transformers, back to back, to avoid crowding 20 turns on to a single core.

I inserted this 10:10 transformer between the plate blocking capacitor (the transformer's primary runs between the plate blocking capacitor's output and ground), and the tuning capacitor of the pi network of my linear amplifier (the transformer's secondary runs between the tuning capacitor input and ground)

My linear amplifier uses a GS-35b, and the tube wants to look into an rf impedance of 2500 ohms. That is why I used 10 turns for the primary and the secondary of my transformer. A few very quick and rough calculations suggested that 10 turns on this material would give me about 10,000 ohms of inductive reactance (rough, ballpark figure) for the the impedance of the magnetizing inductance of the transformer.

Results:

I used 6 watts of input to my linear amplifier, with the 10-10 transformer in place between the output of the GS35b and the output pi-network, and I got 60 watts output. Each of the transformer cores got warm (but not too hot to touch) after around 5 minutes of key down operation.

Conclusions, the transformer I made is too lossy to do the job... but, using larger ferrite cores and/or more ferrite cores and/or a different ferrite material (other than type 43) and/or more turns on the primary and the secondary should work. The goal, of course, is to have no more than 5% of the rf power lost in the transformer. My "quick and dirty" design looks like it has a loss that is roughly 50%.

 

Stu,
My mentor Harry, W3FDY (SK) build a really great grounded grid push pull amplifier after the original AM class C grid driven final was lost in his shack building fire.  It used a pair of 3-500Z's and used the same B&W HDVL plug in push pull tank coils that were used in the old class C 250th push pull final.  I remember he modifed the B&W grid coil to match the impedance of the new input circuit.  I used the transmitter many times and it worked very well all the way up to 10 meters.  It was extreamly stable and he conceived the idea while in the hospital recovering from a stroke.  It took a fair amount of drive as I recall.  He put 5KV on the plates of the tubes and when he whistled the 3-500Z's looked like orange juice cans!
Regards,
joe W3GMS