Simple Audio Driver for 833A modulator

[Tom WA3KLR]

We better specify in terms of peak audio of a sine wave.  You appear to be using RF dbm.  I was referring to audio/telephone levels of voltage. +12dbm is 3.1V rms.

I need a little over 4 Volts peak to get the FET driver to full output.  I need to get this down. 

You say 2 Volts is what you use. Is this the rms value or peak value?

[K1JJ]

OK, Tom-

I meant about 2V p-p audio.   I can go higher or lower, but 2V p-p is what I have the various rigs set to accept.  Standardized in the shack.

But, I think most guys use line level at 0.5 volts, right?  Might want to put in a pot and more gain to accept that...

T

[Tom WA3KLR]

I don't know what people are running for audio interface level.  When I build my Class E rig and start to use an external processor I will probably use at least 1.5 V peak (3V p-p).

So I will change the design to take about 0.75V peak.  Fixed level, no pot.  For those who run a higher level interface, they can easily add some attenuation by a resistor to gnd. at the input where I have a series resistor and r.f. bypass cap.

I may be able to design out the -5V supply.

[Tom WA3KLR]

Attached is my completed 833 modulator FET cathode driver schematic .pdf file.

I did have to retain the -5 Volts supply for the op-amps.  3 power supplies total.  But all are low current.

The FET dissipation is 5 Watts worst case each.  The dissipation is 3.5 Watts each at idle.  They can be on separate heat sinks.

The total harmonic distortion at 1 kHz. with 75 volts peak-to-peak output swing
was 0.08 % THD in the simulation.

The frequency response is 65 Hz. – 7.9 kHz. for -1dB.
                                     34 Hz. – 15  kHz. for -3 dB.

C7 at the audio input determines the high frequency roll-off.  The driver circuitry following this input network has a response out to 60 kHz. This is needed to achieve low distortion via feedback in the audio range.

The delay through the phase inverter in channel 2 is estimated to be about 1 degree of phase shift at 10 kHz.  So this approach should be fine.

I've never shipped out a schematic before that I haven't prototyped myself.  But I have done quite a bit of work and simulation on this design.  Tom and Mike, it should go well for you.

[K1JJ]

Outstanding Tom!

Those distortion and response figures look FB. Regulated bias, negative feedback around the driver, what's not to like?

You obviously spent a lot of time on it and we appreciate a working circuit.

Probably Mike will build it next weekend and I will follow through within a month, once my new 833A rig is together.

Maybe a few of the guys here will have suggestions for improvements before I build mine.

Thanks again, OM.

T

[W2VW]

My suggestion is for Tom Vu to buy lunch for Tom KLR. Not that bait stuff either.

[W3SLK]

Forgive me for being thick here but what are the advantages of driving the cathodes of your modulator tubes VS the grids?

[K1JJ]

Hi Mike,

Well, the first reason was for simplicity, though as it turned out, this circuit is as complex as other circuits out there driving the grids - though uses easier to find low level components.

That aside, one advantage is there's built in negative feedback for the 833A's being a grounded grid stage. Just like a GG RF linear.

833A's are a hard tube to drive. There has been known to be transformer grid ringing with these "dumb" low mu triodes. A fet can sink plenty of current and ideally suited to service in the cathode.

Another reason is that the normal -100V [or whatever] regulated grid supply is no longer needed.  The grids get tied directly to ground and the cathode fet sets the bias via a pot for each tube.

Done correctly, an older electronically regulated tube grid supply used a lot of components. Also, when using the old IJY  FET driver board into the grids, normally needed were several HV supplies, like plus and minus 300 and +600 if I remember correctly.

Tom's requires just +5, +15, and  120VDC at 15ma.  Pretty modest considering this powers the complete driver including providing both modulator regulated biases.

Tom's design uses very little power, like 10 watts worst case - check his specs above.
And I like the .08 % THD figure for the driver...hard to beat that.

Another advantage pointed out is that internal tube arcing has been known to wipe out fets driving the grid, wheras a cathode fet is less prone to this.

I think that covers most of them - there's probably other reasons. But, in the end, driving the grids is a FB method too. I think what inspired us was this application was relatively new for most Amers with few guys using it.

Personally, I wouldn't want anything else if I were building a new rig. [and I am] Good conversation topic when it's running on the air. When perfected, we hope to showcase it and encourage others to build it who are presently running iron transformer drivers of various kinds.... or just want a good winter project to improve their rig's fidelity.

73,
T

[Tom WA3KLR]

Wondering about the power supply?  Not to worry.

The total d.c. power required is about 2.3 Watts.  Use one of those 5 – 10 VA international transformers with dual primaries and dual secondaries.  Run the 120 V line into one primary and use the other primary winding for the 120 V supply.

See attached power supply schematic for a suggestion.

[W3SLK]

I'm contemplating building a single 833 rig modulated by a pair of 810s. Could I apply that same circuit to the cathodes of the 810s and expect to modulate the 833 to at least 100%?

[Tom WA3KLR]

The scheme is to keep the modulator grid grounded and only drive the tube in AB1 as the cathode only gets within a volts of ground at the peak.  The grid could be biased positive in this scheme to allow some positive grid swing.  So there is much less power developed than the maximum the tube can do in the data sheets.  You want to have the plate voltage at the maximum ratings.

We’ll have to see what Mike 2ZE gets with his modulator but it’s probably at most 15 – 30 % of the maximum output in the data sheet.  I am just guessing at this point.

For a pair of 810’s, I think it is looking rather iffy for developing enough power for a legal limit final. 

[W2VW]

OK you just lost me. 2.3 Watts power consumption? I'd expect the 833 stage itself to have less than 10 db of gain. It doesn't add up. The Fets need to make a lot more soup or I am missing someting??? Lost in Joisy.

[W2VW]

Oh AB1. That's it. But I still dunno if you can develop the grid cathode Voltage at that power.

[Tom WA3KLR]

The FETs are acting as variable cathode resistors. The current comes from the tube and it's plate supply.  The grid-to-cathode voltage swing is from -110 Volts to -1 Volts.  The FET basically doesn't take any power to control it.

[W2VW]

Gotcha thanks.

[K1JJ]

The power supply is pretty slick too, using that extra 120V winding. The transformer can even be ordered from Mousier for $9.

I see a typo on the driver notes under "DC Gain Trim". It says to set the bias pots to 1000V. I think you mean 1V, right?

Can't wait to build and try this or whatever version we have a month from now. You're one heck of an engineer, Tomas!


T

[Tom WA3KLR]

Mike,

I’ve been studying the plate current/platevoltage/grid bias charts for the 810 and 833.  I think I can say what the power being developed is for grid swing to 0 Volts. 

For 810’s at 2500 Volts, it looks like you can get about 330 Watts out.

For 833’s at 3000 Volts, it looks like you can get about 770 Watts out.

The rule of thumb, then looks like for AB1 only, 40 – 45 % of the full output in the Class B modulator tube table.

As I mentioned before, with this FET driver you could drive the tubes into AB2.  You put positive bias on the grids and raise the cathode pull-up supply voltage.  The FET driver puts out a greater cathode voltage swing and handles a higher peak cathode current.  The 2 FETs will dissipate more power.


Tom,

The decimal point is in there after the 1.  But it is very small OM.

[W3SLK]

Thanks Tom. I never even considered driving the cathodes. I'll keep this circuit in mind when the time comes. I think we'll wind up discussing this on 160 one of these nights.

[Steve - WB3HUZ]

JJ sez:

"833A's are a hard tube to drive. There has been known to be transformer grid ringing with these "dumb" low mu triodes. A fet can sink plenty of current and ideally suited to service in the cathode."

Why would a transformer be in play in the driver if you are running the modulators AB1?


"Another reason is that the normal -100V [or whatever] regulated grid supply is no longer needed.  The grids get tied directly to ground and the cathode fet sets the bias via a pot for each tube. "

Now you have a 120 volt supply. What's the difference? No doubt this is a cool circuit. But if you are running the tubes AB1, I don't see much of an advantage. A simple RC coupled driver will do FB. (Better yet, use tube you can get some power out of in AB1, like 4-400s). If this circuit drives the 833 in B/AB2, that would be a very hot ticket!

[WA1GFZ]

Hi Guys,
Looking good but a bit more complicated than I thought. Tom KLR I think C4 and C6 need a series resistor to set the input Z of the op amp U3 pins 2 and 6 negative inputs. Does it also need a feed back resistor to the negative input??? The loop feed back looks much better but what sets the stage gain. You may be able to connect negative input to output and run as gain of 1 follower and still have negative feedback.   Also does the Gate time constant allow high frequency to pass. I suspect yes but there looks like a bit of phase shift. My head in in switch mode so may be ok.
Now here is something to consider. You could have pulled the feedback off the 1 ohm resistors and eliminate some parts. I think we are getting close. The high Z pullup
I still think can be eliminated with a 50 Kohm across the FETs. How many GG amps have a puullup on the cathode. How many cathode modulators have a pull up on the RF final cathode? It is a feel good though to have it like belts and suspenders.

[WA1GFZ]

Tom KLR,
Why not eliminate the phase inverter stage and add a simple phase splitter with 1 transistor and 4 resistors. This would eliminate the phase shift on one side.
 Any unbalanced phase shift will increase distortion. Also you eliminate parts.
 I see you are trying for a gain of 100 but still think you need 1 k resistors in series with C4 and C6 to balance the input Z of the op amps.

[K1JJ]

Why would a transformer be in play in the driver if you are running the modulators AB1?

Now you have a 120 volt supply. What's the difference?

Good questions, transformer breath...

Looking at the 833A specs, for AB2 operation they show 20 watts of drive, 70V of bias and 400V p-p to the grids. That means grid current in AB2. This is about 1650W of audio out. I realize that Tom's circuit currently does not put the tubes into grid current, but I was referring to most applications where transformers are required to supply grid current. Maybe adding a positive bias to the grids is all that's needed for the grid current requirement in this circuit. But then that's another regulated supply... sigh.  I agree with Tom that 800 watts of audio from a pair at 3KV is sure babying them. But, grid current with cathode drive would be the goal and show the advantage over transformer drive. [phase distortion, etc]


As for the 120V pull-up supply compared to the usual -70V bias supply....  Well, looking at Tom's circuit, the tube's bias regulation is done entirely with a chip low level in the cathode.  The additional 120V pull-up supply is unregulated, thus simple.  Frank thinks it can be eliminated. The jury's still out. Generally, guys use electronic regulator tube schemes in the grid to pull it off with greater complexity. But it's probably a wash there cuz -70V could be reasonably regulated high level with a zenor.

Even if cathode drive turns out to be a little more complex than grid drive, I am still leaning towards it for my project here.

We'll have to see if Frank's well thought out comments lead to an even simpler circuit.

T

[WA1GFZ]

The pull up would hold the regulator stable when in receive. This could be a good idea so the stage isn't slammed full on when the rig first goes into TX. The 120 volts does not to be regulated because Tom's circuit is a shunt voltage regulator with the change to DC feedback. It just needs low ripple.  Today I think the pull up is good insurance for stability on start up.
I would change the front end to two op amps inverting/noninverting or a simple phase splitter to match the phase delay.

[Tom WA3KLR]

Frank,

There are no missing resistors in this version.  With the FET on the output of the driver op amps, the FET gives a voltage inversion.  Since the feedback is from the output of the FET back to the input of the op amp, the input signs of the op amp change assignments.  You can think of the op amp and FET together as one power op amp.  The gain is defined by( R10/R17 ) + 1, since I am driving the non-inverting input.

The simulation shows distortion of 0.06 % at 1 kHz, per channel.  So I can’t have made any major mistakes.  One bad thing in general about the circuit is that I am modulating the non-inverting input which you should generally avoid.  It requires that both differential inputs are being swung and thus calls heavily on the common mode rejection of the circuit.  To avoid this I would have to add 2 more op-amps to the “simple FET driver”.

The distortion at 10 kHz. with a large voltage swing is up to 0.36 %.  The phase delay between the 2 channels at 10 kHz. appears to be about 210 nanoseconds in the simulation.  This is about 0.75 degrees.  I previously estimated that there would be about 1 degree of phase shift.  In view of the fact that the individual driver distortion is 0.36 %, I don’t think that the differential delay in channel 2 is a concern.  Off the top of my head I can’t think of a quick and confident way to combine the 2 push-pull signals to get a valid resulting additional distortion.  Maybe an ideal transformer?  But it’s not worth the time.

The circuit may work without the +120V pull-up supply, but I did think about the start-up of the tubes without it and like you, it appeared to me that the tubes may draw a spike of current at B+ turn-on until they settle in.

Late last night it did occur to me that if I modified the design and grab the feedback from the 1 Ohm FET source resistor, then we are including the 833 in the feedback loop.  This could be the best way to go.  The audio input waveform is compared to the current of the 833.  The present circuit delivers a low-distortion voltage waveform to the 833 input and hopes for a perfectly linear transfer function.  With the 833 only operating AB1, this may not be far off.

Tom and Mike,

Ideally you should measure the distortion in your stock modulators first, at 2 or 3 power levels that would be repeated with the FET drivers.  You would need a low-distortion sine wave generator, audio distortion analyzer, large value dummy load resistors.  I don’t know what r.f. susceptibility the FET driver may have.  So it would be best to run the modulator without an r.f deck running, if possible.

This proposed 833 current feedback version would be a minor modification of my last published circuit.  It would be good to have distortion data of all 3 configurations.  In the end, the modulation transformer may be the component limiting the distortion level.  But these experiments would be a very interesting opportunity to learn what can be achieved.

[nu2b]

Hey Guys,
Maybe I'm still missing something, but if you look at the load line requirements it seems that in order to get about 0.6 amp peak plate current at say 250 volt minium plate voltage swing you need at least +75 volts on the grid and have capability to draw 0.1 amp of grid current. This means  that you need to bias the grids up to +75 volts and have the initial cathode bias at the -70 volt volts for about 30-40 ma of idle current. this means the cathode needs to sit at  75 -(-70) = +145 volts volts. Then when the FET sucks the cathode to ground the 0.1 amp of grid current can be drawn to support the peak plate current at Vpmin.
The other issue is the pullups. To maintain the sine wave into the feedback loop you need at least 2X145=290volts(minimum) . If you try for less, the loop will try to do the impossible and internally the loop will go nuts and internal limiting and phase shifts will occur.
This is how I came to the original estimate last week of using +100 and +300 volt supplies.
Regards, BobbyT