Any Tube Guys Here ?

What do you mean? AM and SSB contain no low frequency components.

Although a plate modulated AM transmitter still requires a substantial transformer, for apparent reasons.

A better question is how an SSB transmitter can be class B without generating extreme distortion.

This is what my PP PWM example does; the imbalance *is* the signal, hence the beefy iron (which, if there's a DC imbalance, will encounter the same saturation problem).

Tim

If you don't find joy with MOSFETs, consider IGBTs. They're pretty slow

-- you may not be able to achieve more than 10kHz PWM -- but they go to much higher voltages.

And take a look at some plate V / I curves.

Screen series resistors help limit screen dissipation when plate voltage is low.

... Phil

1. What is the zener for? With proper drive,it is not needed. 2. NONE of the tubes are rated anywhere near 5KV *=BRAZZ-fap!!! Look at ratings if xmitter tubes first, and choose from them. The tube manual may even give some useful ideas. 3. That 3rd schematic is useless for your desired wants; it is a nice linear amplifier for starters, and the largest theoretical peak-to-peak output would be 300V. 4. If you use an output transformer, then a tap on it, connected to the screen via a resistor may be in order. The transformer could have (say) a 10:1 ratio to boost output to 3KV with little change in rise/fall time; design depends on pulse width.. Look in the RCA Radiotron for some basic info, maybe ch.25.1, ch.2.4, ch5.2-5.3 for starters.

1. NONE of the tubes are rated anywhere near 5KV *=BRAZZ-fap!!! Look at ratings if xmitter tubes first, and choose from them. The tube manual may even give some useful ideas. 3. That 3rd schematic is useless for your desired wants; it is a nice linear amplifier for starters, and the largest theoretical peak-to-peak output would be 300V. 4. If you use an output transformer, then a tap on it, connected to the screen via a resistor may be in order. The transformer could have (say) a 10:1 ratio to boost output to 3KV with little change in rise/fall time; design depends on pulse width.. "

As you have pretty likely figured out, the application is audio. The voltag es tell you the exact application. The actual load will be capacitive. How capacitive I really do not know yet. I got a pretty good enough idea though to proceed.

With 4,500 volt MOSFETs, to many problems are solved and with tubes, even t he worst audiophool might choke on. I think I can better be trusted with $1

1. The Zener diode is to bias the cathode, without inherent resistance. To make the tube perform as if in fixed bias mode but be self biased. The reas on for doing it this way is specifically to avoid a bad bad failure mode. I f not for thast scheme, ANY loss of drive would cause a meltdown faster tha n ever seen in any tube amp ever. I do nopt want that to be possible. There fore the Zener is of the value that will cut off plate current entirely. Th e tubes ar eonly turned on by drive.
2. If you are looking at the 6AS7 BMP or whatever, This is nothing like wha t I intended. That was included in what I mean as I AM NOT doing. If I did this with tubes, they would pr4obably be 6MJ6, but I am not sure they are r eadily available.
3. That third, like I said, was intended as a negative example. Sorry if th at was not clear.
4. The whole idea of this circuit is to avoid the output transformer. While the output filters will be necessary, the output is not transmitted by mag netic flux, and will be flat response to DC.

And now you all know what you most likely figured out. I want to build an a mp that will feed electrostatic speakers directly, without a transformer.

If the 4kV + 4 kV is not enough, THEN a transformer might be considered, bu t it wil not have to be 1:60 or 1:80, maybe 1:5 or 1:4.

Such a scheme MUST provide better fidelity. And if not, if it is at least a s good as the rest, they might buy it. That's why the 150 kHz frequency is important. It is higher than their source. Higher than almost any tape bias ,l higher than any normal digital audio scheme, and it isn't really digital .

Think about it, it is only quantized in one axis, time. The levels availabl e are still infinite. This is no more damnning to the precious audio signal than tap bias, in fact much less so. MUCH LESS. The LEVEL is never quantiz ed, only the time. And we got that cutup smaller than a gnat's ass at 150 k Hz.

I think it is doable. Now I wan tot know how much voltage really is needed. You know, that info is not as readily available as I thought it would be. I got figures up to 20,000 volts, but I find that a bit hard to believe.

It would arc over. Damn, if you touched the front of the speaker you might get ZAPPED !

What exactly I am going to do with this is a different story. I am also not in a hurry.

If I do need more voltage, I am wiling to try stacking those MOSFETs, reall y. That would be a bunch of fun of course, to make sure they wswitch at the same time. Stacking like that (as done in the Ampzilla, which did not impr ess me) is alot easier. It's just a resistive divider.

Here, I have to MAKE SURE they switch close enough not to take the Vds over the max. Any little error could cause that, and you canot fight it with ca pacitors or resistors. The drive must be exactly in phase all the time, whi le running from almost 100 % to 0 % duty cycle.

Not the easiest thing in the world...

Thanks all. Though this is not quite front burner, I would really like to d o it. Direct coupled. I have no love fior tubes. The only thing about them is running them in linear mode for their "audio exceptionalism" or whatever . I had a pair of Mac copies built by my Uncle that kicked ass. The monoblo cks had each a pair of 6550s for outputs in a Wiliamson (a REAL one with ta ps)and put out about 100 watts RMS, and they were going low enough to repro duce turntable rumble.

The best tube amps I have seen, emulated solid state amps. What people see in tubes is some specific forms of imperfections. They love it.

So...

I still like the idea of using tubes here. Tubes are rugged, and the 6CB5A, ,for example, is only $5. Every time you blow one of those $40 high voltage MOSFETS working on your prototype, you will weep.

At this point I have to try to be practical. It would be novel as hell really, to use tubes, but there are problems. One of them is global warming.

Another is the filament supply. It would have to be isolated and without alot of capacitance as those cathodes will be a swinging.

Now if we wanted to go with class AB or even A, and the nonlinearities are worked out by feedback or something, maybe they would pay for that. It would solve alot of the capacitance problems with the filament of course.

Perhaps a later offering...

This is JURB. I made a mistake and posted from someone else's computer.

Sorry for any confusion.

turned on by drive.

You're thinking small..

A pair of 3CX10000 or maybe a 3CX5000, anything larger may require some extra service to operate them..

For really serious business, you get a three phase service and operate some ceramic 275Kwatt tubes like we do, those can handle 15Kvolts at the plate with no problem! of course you do need a 250 amp filament transformer to heat those babies up!

For home projects, I do have some 3-500Z and a few 4-500's if you don't like triodes.

Jamie

The OP asked about 5 kV PWM, so I assumed that he wanted to generate a DC voltage somewhere between 0 and 5 kV.

At some stage, the discussion turned into audio class-D amplification issues and unfortunately I missed, when the discussion turned this way. Sorry.

That was my point. For audio class D reproduction, generate an amplitude modulated signal, do any impedance/voltages with small RF transformers, on the secondary side, do a full wave rectification on the secondary side and you have the audio frequency signal with a DC bias.

If that DC offset is a problem (such as driving an loudspeaker), the simplest case would be to add a big 2200 uF electrolytic capacitor as commonly done in the past with unipolar power supplies.

Alternatively, use a bridge configuration with opposite phase full bridge configuration as in many car amplifiers.

Perhaps a synchronized rectification would be enough to get a bipolar signal.

That was decades ago. Some Brown Bowery 500 kW HF transmitters used to generate the anode voltage with audio modulation with semiconductors, without huge audio transformers already in the 1980's.

For narrow band operation with a resonant circuit in the anode circuit, what is the problem ? For truly linear wide band designs, e.g. the early Motorola HF linear semiconductor amplifier designs were push-pull, so what the problem with class B ?

Class C might have been an issue.

So you intend to run some QUAD electrostats with a tube stage ? IIRC, they did not need that high voltages.

These are quite low power devices (10-100 W), so I really do not understand why you want to run them with class-D, any linear high voltage amplifier would be usable.

Anyway, the QUAD electrostats are not that great below 300 Hz, so you would still need a separate subwoofer, so there is not so much to worry about the bass response of the amplifier.

If you just intend to drive electrostats, I really do not think that you need that expensive FETs.

What happens once the capacitor charges up? :-)

I said class B (unqualified), not just PP. I have the ARRL handbooks to prove it -- they describe, e.g., a single 6146 class B final.

I guess IMD pushes the spurs out-of-band, or something, where they hit the filter and go nowhere. But then, what of the harmonics, shouldn't those mix into various IMD products back in the passband?

I'm honestly not sure. SSB isn't something I've played with (aside from crude reciever experiments).

Indeed, I don't think I've ever seen class C for AM (direct) or SSB, and for good reason.

I suppose AM might not be so bad, if you started with a signal that's not quite 100% modulation, and you set the cutoff bias equal to the negative modulation peak. And assuming the class C amplifier is linear for the rest of the range (i.e., it goes into cutoff at the low end, but remains linear otherwise, not saturating as well). But then, you'd just get better performance modulating it correctly in the first place.

Tim

It is not necessary to preserve the carrier wave form in SSB linear amplifiers, faithful envelope and carrier phase are sufficient, if there is a resonant filter at the output.

Have a look at the 6146 final. There is usually a pi filter with a Q of 10 to 15 at the output. It is necessary also to swamp the output capacitance of the tube(s). The ham transceiver of 1960's and 1970's usually had a pair of 6146's in parallel.

There will Vcc/2 across the capacitor.

Unipolar power supplies were used quite common in the 1960's and 70's and still in battery powered appliances.

Of course, you my hear a loud bang from the speaker during charging up the capacitors. Especially bass reflex speakers do not like this, the cone might fly out of the magnet, so some preventing tricks are often used, such as injecting some current into the differential input stage through a RC circuit so that the amplifier output voltage (before the capacitor) is first driven to the low voltage (Gnd) and then slowly charges up to Vcc/2. Alternatively, a relay is used to connect the speaker, after the voltages have stabilized.

These are for narrow band operation, in which the tank acts as a flywheel, the tube is just supplying a variable amount of energy during 180 degrees, the tank takes care of the rest.

There are of course harmonic distortion which can quite well be filtered out with a pi filter at the transmitter output.

Of the intermodulation products, the worst is the 3rd order, since it falls just within or adjacent to (splatter) the desired audio band and is impossible to remove at those power levels. This becomes a problem only when the stage is overdriven.

Some digital modulation methods with possible multilevel symbols, such as 16QAM, is very sensitive to the signal distortion, so you may have to run the transmitter well below nominal power.

Where, off the rectifier? Where does Vcc get into it -- the transformer ratio could be anything?

I was referring to getting audio out of the RF rectifier, not coupling it up there (or linear amps with single ended supplies...where did that come from?).

Tim

If you look at the feedback circuit in such an amp - single ended, you will see that the feedback is or should be designed to brign up the "center" vo ltage slowly. If a guitar amp for example is turned on, the feedback is set to not bother with low frequencies so much. So if it is single ended there will be a pop heard because that voltage, because of the time constant of the feedback network, rises quickly. Of course everything must be made to t ake that surge but it usually pretty much is.

I am amp designed to reproduce down to 20 Hz, the rise of the center voltag e will be much slower. However, whatever the low end of the output cap, sho uld match the low end of the feedback. That way the response falls off pret ty much the same as the power bandwidth.

If the output cap is within the AC feedback loop, care must be taken when i t comes to phase shift or you might just have built a large oscillator at t ens of Hertz. I remember one I tried to modify that used maybe a 1,500 uF w ith like a 4,700 uF. It resulted in low frquency oscillation, though in my view it shouldn't have. I did not have a print so I backed off and I think I wound up with a 2,200 in there. Not a bad value really. At 20 Hz, the Xc

n that low because very few can reproduce it, so, that isn't so bad.

In my "little" project here, I am not that concerned with the low end. It h as been fiigured out what I am doing now, and we all know how ESLs are not for low frequencies anyway. If Martin Logan for example wants to use it tha t way fine.

I read up on their website all the s*it they had to do to get decent low fr equency respomns out of their full range units. In fact most of their line uses standard dynamic woofers. I would like to hear those full range panel jobs, but I doubt there is a dealer nearby. If there is air travel involved , it is just not that important.

I know many have done alot of things, but what else am I going to do ? Gimm e a decent job and I'll do that. Until then I will just do freelance work a nd work on my pipedreams. In fact, the speaker project is on the burner beh ind a nice beench power supply project. I got that covered in the design as pect of it I think. Right now i am thinking features. I think the device wi ll include like a pulse generator, several frequencies and all that. Also a specially builty generator to test capacitors with. I'll get into that som e other time, but that part of it will use a scope. I am NOT including a sc ope. It'll a;; be in the FM, which you can RTFM.

Anyway, enough for now. For the valveheads around, sorry, the first offerin g is going to use the MOSFETs. there shold be no difference in the sound in class D. The outputs have almost noting to do with it.

However, if someone were to actually want a class AB type tube amp for this purpose, I know I can do it. Part of it wil probably be solid state, AT LE AST the voltage regulators for the bias and all that. Nut if a market is no t demonstrated to me there is no way I am investing in making that happen.

it up there..."

I hjave tried befroe to configure what I THINK you are all talking about - a transformer coulled switching amp. I have found it nearly impossible. Sur e, it can be done but in my view, to no avail. There seems to be no way to make it efficient and clean.

Maybe I am wrong, but without the print for the idea, right now I have no i deas. If you took two transformer secondaries for example and tried to rect ify the audio off them, the positiva and negative will always work against eachj other, but additionally, the positive side will NOT let the output go negative and vice versa.

The deal with PWM going through a teanssformer, I never really explored. I saw it here, but right now I see it as most likely horrendously inefficient .

But I could be wrong. I'll give it some thought.

Synchronous rectification is one of the better ways, but not very amenable to tube-only application. See Dave Berning's "not quite OTL" tube amps -- diode/MOSFET switches at 200kHz effectively make DC transformers; compare to a chopper amplifier with no gain.

Another way could use a constant voltage switcher to deliver the output voltage, and a constant current "sucker" to provide the bias of the DC load resistor, without actually wasting power. This has a certain logic about it, as you can only supply one aspect at a time (current or voltage), since rectifiers won't let it back out the other way (duh). So you need the conjugate property (CC complements CV, or vice versa) to make a full amp.

Hmm, I should scratch up something based on that.

Tim

-- Seven Transistor Labs Electrical Engineering Consultation Website:

up there..."

I hjave tried befroe to configure what I THINK you are all talking about - a transformer coulled switching amp. I have found it nearly impossible. Sure, it can be done but in my view, to no avail. There seems to be no way to make it efficient and clean.

Maybe I am wrong, but without the print for the idea, right now I have no ideas. If you took two transformer secondaries for example and tried to rectify the audio off them, the positiva and negative will always work against eachj other, but additionally, the positive side will NOT let the output go negative and vice versa.

The deal with PWM going through a teanssformer, I never really explored. I saw it here, but right now I see it as most likely horrendously inefficient.

But I could be wrong. I'll give it some thought.

I have. And also seen appreciable screen current even where the screen grid is in the shadow of the control grid.

?? So where does all this cathode current go?

I can assure you that limiting screen current will also limit screen dissipation.

Yes, usually for high power tetrodes, where the screen grid is typically attached to a tap on the output transformer.