Any Tube Guys Here ?

Screwed up an edit there. The issue is it will have to be a high potted part.

I don't think wiring the screens like that is a good idea. It looks like you've got positive feedback from the tube output to the screen - I'd expect it to oscillate. I think you may need bypass capacitors across R1 and R4.

Also, there will need to be some feedback at DC from the output to the PWM stage - the output is unlikely to sit at exactly 0 volts all by itself.

You'll also have to worry about the heater to cathode breakdown voltage of the bottom and particularly the top tube - the top tube isn't going to like it when its cathode is bouncing from 0 to some kV and its heater is sitting at 6.3 volts...

The bottom tube's plate voltage will go below the screen voltage if the screen is bypassed. That's OK, that's how a pentode is supposed to work.

I've dabbled before:

Concept, but not built. Needs more work anyway:

The biggest downside to pentodes in a switching converter is, they don't conduct in reverse like MOSFETs. Think SCRs with voltage input and no latching. This makes constant-current inverters much more attractive than constant-voltage. (Assuming you can generate the constant current in the first place -- since we're already talking tubes, a massive choke, say 10H

Recieving tubes no longer hold any records on current or voltage capability: high voltage MOSFETs of substantial ratings are available as cheaply.

Note that sweep tubes are only rated for peak voltages under hard cutoff (Vg1 = -200V or whatever). I don't know what happens under forward bias at that voltage, but the inevitable ion bombardment probably isn't friendly to the cathode.

The other big downside to a pentode (or tetrode) is, cathode current is controlled firstly by grid voltage, and secondly by "virtual plate" voltage. On a triode, this is just plate plate. On a tetrode+, the screen grid (and, by even weaker effect again, the plate proper as well, hence the large but still finite plate resistance) serves this purpose. Of note... cathode current, in a tetrode+, is barely determined by plate voltage at all, whether positive or negative. So, what happens when the plate voltage drops into the saturation region? All that current that used to be flowing into the plate gets sucked right up by the screen grid. A few miliseconds later, said screen grid evaporates in a flash. Umm, yeah.

There is no simple analogy of a screen grid among semiconductors. Due to the positive backwards-transconductance (change in plate voltage --> change in screen current), static positive feedback can be implemented with a single tube, no transformer (or spooky effects like secondary emission in true tetrodes) required.

The most obvious "duh" solution, which I used in the two above examples, is simply limiting screen power with a resistor, and bypassing it with a cap to maintain reasonable AC performance. This isn't bad for a single ended example with positive saturation voltage, but probably isn't going to work as well with reverse voltage (attempting to turn the tube on, while the plate is negative, will simply discharge the screen bypass cap, then when plate voltage reverses again, screen voltage will be too low and the tube will simply 'wheeze' and not saturate properly).

One can envision various approaches, like minimal bypass, or none at all (downside: screen voltage swings around, so Miller is back with a vengance), or various sorts of passive (diode?) or active (zener / gas tube / bootstrap tube / etc.?) clamping strategies. All of which cost more tubes (or high voltage semiconductors).

Oh, and for a totem pole / bridge, you only have "N-vacuum" devices to choose from, so all voltages (grid and screen) must be bootstrapped or isolated.

Which isn't too bad, I suspect. A floating +/-150V supply, referenced to the cathode, could provide plenty of screen bias (positive, with current limiting / clamping / whatever as chosen) and grid bias (negative, with transformer coupling, gain and drive). The screen power would be substantial (watts), but nothing extreme as gate\\\\grid drives go. The heater, by the way, also needs to be cathode referenced (give or take only

Tim

One good source for tube stages is the amateur radio literature from a few decades ago, like the ARRL Handbook. There are plenty of tube output stages to up to a few kW in the HF range. Some of the most popular tube types might still be available.

For many high power tubes, the maximum screen grid voltage is much less than the maximum anode voltage, so that explains the R1/R2 voltage divider. When the tube is heavily conducting, the anode voltage can be quite low and if the screen is powered from a high voltage fixed source, the screen would become the most positive electrode, sucking all the electrons and melting the screen. By powering the screen indirectly from the anode, the screen can't become more positive than the anode.

How is that 5 kV measured, is it just B+ or the difference between B+ and B- or a peak voltage across and inductive load during current switch off ?

High voltage/high power tubes are often directly heated, so you are going to need separate heater winding anyway.

I do not understand why you want to do totem pole output with two tubes, wouldn't a normal push-pull approach be simpler ?

Of course you will need an output transformer, but since the frequency appears to be fixed (instead of spanning a decade in most amateur radio linear amplifiers) and reasonably high (150 kHz) to keep the core size at a reasonable level. With selecting a modern suitable core material, the size and losses can be further reduced.

With separate primary and secondary windings, you can chose the transformation ratio freely, so if you need 5 kV peak-to-peak, the tubes doesn't have to run at several kV, you might even be able to use transistors :-)

What tubes are used ? I have seen a similar circuit with six triodes feeding a 16 ohm loudspeaker directly.

Beautiful!

joe

If I understand correctly, the output transformer runs on audio frequencies and the 10 mH choke and the two 10 nF capacitors filter out the PWM on the primary side ?

Audio transformers going down to 20 Hz are quite big. Wouldn't it be better to run the PWM through a small 20 kHz+ transformer and then do the filtering on the secondary side ? An elliptic filter with at least

PWM signals are *wide band*, only needing a LP filter to recover the original modulation.

.... Phil

Let me try to explain it a bit more politely...

Even if the PWM signal is switched on and off with a high frequency, the (moving) average of the signal has a low frequency, i.e. the audio signal. Small HF transformers get immediately into saturation if you'd feed them this kind of signals.

A bit more in detail: If you switch a signal with a high and fixed frequency from 0 to 1 and vary the PW such that over every period of the HF signal the period's average follows a LF sinewave, then the flux in the transformer's core will follow this LF sinewave, not the HF switched signal, as the flux is equal to the time-integrated signal, which is LF. That's why the transformer needs to be big, fat and heavy nevertheless.

joe

That's a nicely formulated explanation, thanks.

How, with a PNP tube?

The first time I heard of PWM I immediately tested it out with a CMOS quad comparator IC, 4 BC-something switching transistors (2 PNP, 2 NPN) and very few capacitors and resistors. I used two comparators for a PWM oscillator/comparator, the other two as drivers and made a double- balanced push-pull output. Made an incredible noise for such small transistors and I had a marvellous time with the Beatles when it worked...

Just look at the picture to which the above link points. Your only excuse can be that you're trying to watch it on a smartphone ;p

joe

"joe hey"

Connect the screen and anode of any pentode and you have a triode ( with all the disadvantages ) but both can then swing over large voltages.

It is *normal* to allow the screen voltage to exceed that of the anode, this what makes a power pentode tube efficient as it allows the anode to pass high current at low voltage.

Screen current ( hence dissipation) is safely minimised by two factors, having a low surface area to collect electrons and crucially being wound so as to sit in the electrical "shadow" created by of the control grid.

.... Phil

Ok, but why the '** Shame'? So he will never again try to explain anything?

("RCA Receiving Tube Manual, 1940"; p8.)

Interesting, I didn't know that. Thanks for _your_ explanation.

joe

Yeah, for reasons already explained...

You can transform PWM if the flux returns to zero every cycle. For example, a pulse (0-49% duty) on, then off, then a pulse in the opposite direction (0-49% duty), then off. With the two pulse widths equal and opposite, flux is zero over a cycle. But the average DC component is zero for the exact same reason.

Say you detect this signal with a rectifier and LC filter (in common use, this is called a PP forward converter). You get DC of varying offset, zero to max. You must *dissipate* the DC in a resistor, and cap-couple the output to the AC load, dropping your theoretical efficiency to a paltry 25% class A.

In fact, this is exactly what I'd done in the "compact tube challenge", except it was done at the plate directly, without transformation (at

The "Borg" example, as a PP forward converter with primary side filter, does not require a secondary rectifier and coupling stuff, and can use differential (a pulse 0-x% down, wait, a pulse x-0% up, wait, ...; where x = 0-100%) rather than common mode PWM. Which is familiar from most class D amps.

The amount of overlap controls what "sub-class" it operates in: if MOSFETs were used, it could always run at a total of 100% duty (i.e., one switch or the other is always on), which would be Class D-A, but with full efficiency this time (because it's push-pull). This would be counterproductive for the screens, so a control sets the 'PWM bias' from Class D-A to D-C. D-B would be ideal (the opposing switch turns on just as reverse current runs out and voltage swings positive), but it is impossible to predict reactive current in an application like this (the reactive current, absorbed by the damper diodes, is the sum of the reactive current in the filter choke and the load's reactive current, which may be quite reactive at certain frequencies, given the arbitrary load a speaker presents).

Tim

That may be the best part right there. Thanks. I might go with the MOSFET instead. The only reason I was using tubes was for the voltage. The MOSFET solves a whole lot of probleems at once. As far as efficiency as well as simplicity of design.

Other than efficiency, a good reason I went class D was so that nonlinearities in the tube would not affect the output. I had no idea there were any transistors that could handle that kind of voltage.

So, all you really need is a dumb class A to B linear amplifier?

Barking up the wrong tree -- you didn't mention how much current you need but a single pair of transmitter tubes will do the job without linearity problems. If nothing else, wrap the thing in an op-amp for ideal (time-asyptotic) performance.

Tim

I think there are some 4 kilovolt mosfets around (or maybe they are IGBTs.) There's tons of 1200 volt stuff, silicon bipolar and silicon mosfet and SiC.

One fun amplifier is a 1B3 high-voltage rectifier tube, good for 20 KV or something. The output is the plate current and the signal input is the filamant voltage. Bandwidth is mediocre.

Very good. What happens with increasing cathode current and falling anode voltage, such that the anode load can no longer support all that cathode current?

That excess current has to go somewhere, why wouldn't it fall back to the screen grid?

It is usual to limit screen current with a series screen resistor for precisely this issue.

Agreed, it isn't common practice to slave the screen grid to the anode voltage, as this undoes most of the advantages of the extra electrode.

I certainly agree, depending on conditions, that the screen grid will often be above anode potential.

With push-pull I understand a center tapped primary, in which the B+ is connected to the center tap and the ends to the tube anodes (transistor collectors in which both are either NPN, usually silicon or both PNP, usually germanium).

What do you understand by Push-Pull ? A complementary emitter follower, in which a NPN sitting on PNP ? A totem pole would then be NPN sitting on NPN ?

Then how can RF linear amplifiers handle AM/SSB without huge transformer cores ?

How about a push-pull stage with slightly inbalance, which would saturate the core immediately. Iron powder cores with "built in" air gaps handles this quite well.

At 44 bucks a piece it will be really heartbreaking every time something goes wrong and you blow one up. A 6CB5A tube costs $4.75 and will take a lot more abuse.

In a class D arramgement like his, if the circuit is operating properly, the tubes shouldn't ever see the peak supply voltage except transiently. Isn't that the whole point?