Quasi complimentary output stage?

[metalsculptor]

I am preparing to fire up an ultrasonic welder (sonopet 1000) It is a 70's vintage valve unit with 6JS6C valves for output in what looks like a class B setup with transformer drive to the grids The real weird thing is that it uses a quasi complimentary topology with 15 valves in parallel per side. PSU has +/- 700V rails and separate heater supplies tied to the cathode, one to -700V and the other to the anode of the other bank of valves who's anodes are tied to +700V, load is fed from the centre. I always liked this topology but never thought anybody would use it with valves, I thought the output stage was class C with a tank circuit initially. Rated output is 1 KW. All valves look good heaters work but it is currently smoking bias resistors, I have no load connected yet so I will arrange a load and replace all the small electrolytics before firing it up again.

Does anybody know of any valve amplifiers using this arrangement? It would seem like an excellent way to drive electrostatic speakers if the voltage was higher. Does anybody have a schematic of such a topology, mostly the bias and the drive is of interest.

 

[Miles Prower]

Look up "Futterman", "Inverse Futterman", and/or "SEPP"

If I was going to do a hollow state OTL, I'd opt for the Circlotron topology.

 

[Sch3mat1c]

If it's transformer coupled, it's just SEPP, which is a rearrangement of classic PP, probably done to save the hassle of an output transformer (at the expense of another heater winding).

FYI, tubes are inanimate, so it's hard for them to give compliments to each other. Since there are no complementary tubes (no "P type" electrons), there is also no such thing as quasi-complementary, unless you build a hybrid.

Tim

 

[metalsculptor]

Quasi complimentary is term used for transistor amplifiers where 2 N type devices are arranged in same manner as the valves described. Totem pole may be a better name for this. I suspect you are right about saving the output transformer, the load is a magnetrostrictive transducer (lossy inductor) with some capacitance to resonate with.

What is the advantage of a circlotron over the "quasi complimentary" ? Two floating PSU's seems like a lot of opportunity for capacitive coupling especially if the PSU's are over 1Kv or so each and driving an electrostatic speaker would be a pain from a bridged amplifier. In the circuit I looked at there was a 2K resistor in parallel with the load which I thought strange because there are very few valves with a plate resistance that low, the 8871 on my bookshelf would come close.

 

[Sch3mat1c]

No, that can't be right... they make totem poles like this,
[IMGDEAD]http://webpages.charter.net/dawill/tmoranwms/Circuits_2008/TTL_Output_Stage.gif[/IMGDEAD]
and they aren't complementary at all. The "quasi" part comes from substituting the P type complement with a Sziklai pair, thus saving the expense (in the olden days) of a PNP power transistor.

2k ohm load? That's not so bad, I've driven 600 ohms with a single sweep tube (at high efficiency!).

I wonder why your device has *thirty* in it -- is it a linear amplifier? If they're getting 60W from a pair, presumably limited by plate dissipation (also ~60W/pair), that's 750W output, and 750W plate dissipation! That thing must be *beastly*! I can only imagine how some here would drool, if you rewired it for audio use, or OTL for that matter. Dayum...

Tim

 

[metalsculptor]

The totem pole I am familiar with is the TTL one which is 2 npn devices in series with a bias diode in between and a resistor. If I knew how to upload images I would post an example. Anyway look up TTL in wikipedia and they have a nand gate schematic.

My amplifier is rated at 1kW output, mid point on the plate current guage is 1A and the rail to rail voltage is 1400V I assume that it is linear and probably class B1 because the drive is small. There would be no need to run class AB due to the resonant load eliminating most harmonic distortion. The only reasons I can think for using 30 line output valves is either cost or keeping the plate voltage low. It is a Japanese machine using Japanese valves and labour was quite cheap in Japan at that time.

It probably would make a good OTL amplifier but it would require a load of hundreds of ohms. The machine is about the size of a large floor standing speaker with 2 7inch fans to cool the valve array the power supply is just a 1050V CT transformer feeding a bridge rectifier like most transistor amplifiers except more voltage. Surprisingly the filter capacitors are quite generous 6 x 200uF in series, usually industrial equipment uses the bare minimum, 1 uF in a 4Kw 3Kv single phase supply is not unusual.

When I eventually fire it up I will have more idea how it works, it will need some careful CRO probing with those voltages, bit like horizontal deflection amplifiers.

How did you manage to get good efficiency driving a 600 ohm load with a line output valve?

Mark

 

[Sch3mat1c]

http://www.diyaudio.com/forums/tubes-valves/110155-blasphemy-im-sure.html

Plate efficiency was something like 80%, which is quite excellent for tubes. The output network is "class A", burning half power, so the overall efficiency is low, but that would be fixed with a proper PP circuit.

If they did the same thing in your welder, they could do 1kW with about four of those sweep tubes, and a number of extra damper diodes (necessary because tubes really only conduct unidirectionally, ever -- all transistors work backwards to some extent, but tubes won't!).

I wonder if they went with a bigass linear amp for that reason, to handle reactive current (generated by harmonics or imperfect tuning). Or maybe simply no one was thinking of class D back then!

Tim

 

[metalsculptor]

http://www.diyaudio.com/forums/tubes-valves/110155-blasphemy-im-sure.html

Plate efficiency was something like 80%, which is quite excellent for tubes. The output network is "class A", burning half power, so the overall efficiency is low, but that would be fixed with a proper PP circuit.

LOL well I suppose a valve is like a mosfet with a lot higher Rds ON so why not PWM. I loved the valve catch diode, I probably would have used a mosfet cascode arrangement to drive it PP with floating bias supplies and standard mosfet high side drivers and 5% THD is not much more than a typical single ended pentode large signal THD.

If they did the same thing in your welder, they could do 1kW with about four of those sweep tubes, and a number of extra damper diodes (necessary because tubes really only conduct unidirectionally, ever -- all transistors work backwards to some extent, but tubes won't!).

The thought crossed my mind to add a couple of ultrafast diode strings across the valve arrays. If valve life wasn't an issue I think a lot more than 1Kw is possible from 4 6JS6's They can hold off 7.5Kv so using bit more of the plate voltage potential is an option with trap diodes the plate voltage can be precisely limited unlike in a flyback circuit.

I suspect that they had not thought of class D, because most valve oscillators produced sine waves and overdriving a linear amplifer with a sine wave only give you class C. Digital circuits were only just becoming widespread and the SCR was the mainstay of power electronics. You are probably right about the valves being over rated to handle the reactive current normally shunted by an antiparallel diode.

I have seen valve multivibrator circuits in industrial electronics and there is always the hydrogen thyratron (lower frequencies) if square waves were needed.

 

[Sch3mat1c]

Oh heck, industrial electronics were almost entirely xenon (e.g. C6J's in VFDs) or mercury (for the big assed stuff, from trains on up to megavolt HVDC power transmitters, which have only fairly recently been upgraded with SCRs). I don't know of hydrogen thyratrons being used for anything industrial, but they were huge in pulsed radar and stuff like that.

Hewlett Packard even has some old tech briefs describing some of their fast-edge pulse generators, which used a hydrogen thyratron. 1ns edge into 50 ohms isn't bad for 50s tech.

Unfortunately, xenon and mercury vapor are too slow for active commutation. In the miliseconds they take to turn off, you really need an AC source to let the "smoke" clear before they're ready for another shot. You could use hydrogen thyratrons for a particularly beefy subwoofer amp, but the darn things are still pretty expensive even today.

What's remarkable about sweep tubes is how much worse they are compared to MOSFETs, which is not much worse at all. Even if there were such a thing as a 5kVds MOSFET, it would be like a hundred ohms. Typical 6KD6 "Rpk(on)" is more like 50 ohms. Capacitance is lower, too, though it's worth noting the Miller capacitance of sweep tubes isn't as low as signal pentodes, plus you need much more drive voltage.

Tim

 

[metalsculptor]

I have never seen hydrogen thyratrons used in industrial circuits either, I was suggesting them as valve based square wave source. Radar and beam diverters are uses for larger versions. I have some C3J thyratrons a few 2d21's and a half a dozen class D ignitrons + assorted mercury arc rectifiers. The places I still see the occasional thyratron is in elevator motor controls and old spot welders where they fire the ignitrons.

Driving a sweep tube (US terminology) should not be difficult with a mosfet even using a grounded grid, switching an amp or so at 200V with sub us rise times should be well within the capabilities of a modest 400V mosfet.
I see your point about on resistance, the data sheet for the 6js6 suggests 540mA plate current at 70v with 0V on grid one so it should be lower with some positive grid drive.
The 6KD6 curves look very healthy I wonder how they would look with 20 or 30 V positive grid 1 drive

You are tempting me to grab some 6DQ5's to see how much power they will switch, 6DQ5's are a sweep tube I am familiar with from my childhood, not as common as the 6CM5 in Australia but far better power capabilities. We did not get colour television in Australia till 1974 so large sweep tubes were not required by the time transistors had replaced them. The 6DQ5 is cheap enough to blow up, if the valve flashes over the power supply will happily supply however much power it takes to clear the fault. The 6KD6 like the 6JS6 is expensive, probably because it got used in Japanese transmitters

I made a mistake in a previous post the valve on my bookshelf is 8171 not an 8871, 12Kw plate dissipation in a package smaller than any air cooled transistor and heatsink of equivalent power