Series Modulating a Class C 4-1000A

[K1JJ]

I'm looking for ideas to series modulate a grid driven, class C,  floating above ground, 4-1000A for AM use.  The 4X1 may have 4-5KV HV on it.

I already understand the efficiency reduction using series modulators or grid/screen modulation compared to plate modulation.

But what I am not clear about is the difference between using another 4-1000A as a class A series modulator in the RF final's cathode compared to using a 1000V MOSFET for the job.   Wouldn't the final tube need less than a 300V swing in the cathode to make it turn on and off?   ie, Why use another 4X1 for the job if a lower voltage, power saving MOSFET could do it... or can it?

If I were to modulate/drive the center tap of the RF tube's filament transformer with audio, does the final have to float above ground like a standard series modulated final config or does it act more like a cathode driven linear and can be referenced to ground?  Check out the schematic below.

Or maybe I am thinking of a "series modulated" tube being placed in the plate lead, as a single ended  "plate modulated" connection. I'm not sure of the advantage of using up this extra power in the plate circuit as compared to the cathode.  In this case the final can be referenced to ground. Maybe the single ended aspect is so inefficient in the plate circuit that a cathode connection is better.

Another alternative is to modulate the grid and/or screen of the final, using a MOSFET. I want to stay away from using transformers.

I have a PW 6AQ5 X 6AQ5 series modulated rig that is very clean. I was thinking of possibly scaling it up. (the series modulated part)
Here's the basic schematic:   http://www.amwindow.org/tech/htm/series.htm 

Maybe in the end, over a full modulation cycle, all of these short-cut schemes are not much more efficient than running a linear with a clean, hi-fi exciter....  :-)


Any ideas?

T

[Ian VK3KRI]

I think your mixing series mod and cathode mod. With cathode mod (as I understand i) you're part grid modulating the tube ,because the voltage on the grid is relatively constant compated to ground. Thus 300V across the device in the cathode  may well be enough to get 100% nedative modulation because you've added 300V to the grid bias.

However true series modulation would keep the grid voltage relative to the cathode relatively constant. thus at 100% negative modulation you will have full B+ across the device in the cathode. At carrier the cathode dropping device will be dissipating the same value as the DC input - nice in winter.

                                                            Ian VK3KRI 

[AB2EZ]

Tom

I agree with Ian. It appears that you are proposing to modulate the grid - to - cathode bias by applying an audio signal is series with the cathode and ground.

If that is all you do, the result will be very distorted modulation unless you use a small modulation index.

You could try obtain some improved (perhaps significantly improved) linearity, in conjunction with a high output modulation index, by employing feedback from the output of an off air monitor (envelope detector) to the input of the grid-to-cathode bias modulator. It would make an interesting experiment... but I would try it with something a lot smaller than a 4x1 to see if you really want to scale it up. As you point out, the end result won't be any more efficient than using the 4x1 as a linear amplifier.

The objective, in using the feedback circuit will be to force the cathode-to-grid bias to move up (or down) just enough to cause the desired change in the amplitude of the envelope of the r.f. signal. Since the envelope of the rf output signal is certainly capable of changing from its midpoint (carrier level) to twice that value, if the tube is properly loaded... and there is no problem in reducing the envelope to zero... the feedback circuit will simply pre-distort the biasing voltage change to get the desired modulation. As with any feedback approach, it won't work if you load the tube so heavily that it is not capable of putting out twice the carrier-level envelope amplitude.  As with any feedback circuit, it will work well provided the phase shift around the loop is small for audio frequencies of interest... and provided the gain of the feedback loop drops below 1 at frequencies well below the 180 degree phase shift frequency.

As an aside:

1. If you are going to do grid-to-cathode bias modulation, and since you are not running the tube in a grounded grid configuration... it might be better and simpler to add the modulation to the grid bias... and leave the cathode grounded. That way, your modulator won't have to deliver any power.

2. I know that Carl, WA2UJX is in the process of building a screen modulated 4x1 rig. With proper design and with careful adjustment of the loading (using a scope with a 100%-modulating 1000Hz sine wave applied)... I believe that most people won't be able to tell the difference between a screen modulated rig and a plate modulated rig.

Best regards
Stu

[The Slab Bacon]

<snip>   
 2. I know that Carl, WA2UJX is in the process of building a screen modulated 4x1 rig. With proper design and with careful adjustment of the loading (using a scope with a 100%-modulating 1000Hz sine wave applied)... I believe that most people won't be able to tell the difference between a screen modulated rig and a plate modulated rig. 
 <snip>


Been there, done that!! Some years ago.   No one can tell the difference unless I tell them. Not even you, Stu, as you have heard it many times.


And to Mr Vu, welcome back to air last night!! It was good to hear you back on!

                                                                                    The Slab Bacon

[ab3al]

I couldnt find the proper iron to plate modulate my gs35b russian tube.. so i plan on just yelling at it really loud...

[WA1GFZ]

Tom,
Say you wanted to convert your PDM rig to series modulation. First you would have to bias the switch tube to be a stable regulator with about 2/3 of the plate voltage across the modulator and 1/3 across the final. Do the math you will need plenty of air across the the modulator due to steady state dissipation across the modulator. Now you want to modulate the thing so you could duplicate the cathode modulator KLR gave you or you could modulate the grids. Plan to waste a lot of heat. Frank's screen modulator would be a lot easier to do but again the final will be less efficient at carrier but you could run a pair of tubes in parallel.
Of course you could screen modulate a 4X5 and do it with one tube. I think the safest easiest set up would be the screen modulator like Frank uses since there are no floating stages. I bet a 11N90 on a good heat sink could screen modulate a 4X1. I suspect a DC coupled screen modulator would be very clean.

[K1JJ]

OK, I think I see what I was missing, thanks guys.

To summerize, if the final tube is floating above ground, and the grid and cathode keep their relationship constant for audio, then a series modulated class A tube in the cathode circuit could modulate the final just like standard, loose class C parameters. However, if the cathode and grid vary relative to one another under modulation, then it becomes cathode or grid modulation and has its linear and tuning limitations as described by Stu and the guys.  Very good.

Frank, your PDM analogy is appropriate - here's why I'm thinking of this system... After I got off last night I fired up the PDM 4X1 rig to get ready for today. Loaded conservatively into a 2:1 antenna load, at only 3500V, one of the PDM coils started arcing badly and caught on fire. Before I could put it out it melted a big hole in itself... :-)  This thing is a Frankenstein nightmare to handle.

Anyway, just imagine the problems when I put full voltage on it. So I'm seriously considering getting rid of the PDM coils and PDM generator and converting it into a clean, series modulated rig. Same tubes, layout, etc except I'd need the audio interface to the modulator..

So, you think those standard 11n90 MOSFETS are linear enough for clean audio and not just switch type fets?  Also, the MOSfet in there now gets pretty warm, even with a heatsink doing PDM. I wonder how much power it will take doing full analog audio service?

Again this 6AQ5 circuit works very well for me now. (transformerless and DC to light with huge audio peaks)    http://www.amwindow.org/tech/htm/series.htm 

Even though the series modulator tube is a power hog, being able to get perfect 1 hz audio and maintaining the same loose tuning class C parameters is appealing at this point. Not to mention getting rid of those troublesome and space wasting PDM coils.... :-)

BTW, right now the PDM tube does not use regulated screen or grid supplies. Do you think I'd need to regulate them in series modulated service?  Maybe it wud be easier to use a triode, like a single 3CX-2500 series tube to eliminate the screen and also handle the class A power....

T

[Opcom]

Gates had a PWM 5KW AM rig that put the modulator tube in series with the final cathode. Both were triodes. The RF drive was transformer coupled so there as no effect of changing the final's bias or drive parameters over the modulation cycle. The filament power was supplied to the final at the switching frequency by a transformer. I know nothing more about it, I was only shown the transmitter and told these things while on a tour of the site. A home version might be a complicated project from a design standoint, but not beyond the skills already shown by some on this board.

[WA1GFZ]

Well bummer on the filter. Your 11n90 ran warm because the gate drive series R was a bit high. A FET will be very linear if set up properly. And yup the 2500 will make an easier modulator being a triode. I would cathode drive the tube with an 11n90. Remember if you run 1 KW input with 2/3 the voltage across the series tube you will be dissipating 2000W in the modulator and about 250 watts in the final. Lighting all the tubes will put you close to 3KW of heat. Now compare that to a pair of 4X1s or a 4X5 screen modulated at 1KW. 
I was thinking of an h mode modulator for the solid state rig for a while and simulated a very nice DC coupled amp with feedback that was clean out to about 25 KHz. Then I looked at power dissipation and quickly changed my mind back to pdm. Floating finals is a big PITA as you have learned twice.

[K1JJ]

And yup the 2500 will make an easier modulator being a triode. I would cathode drive the tube with an 11n90. Remember if you run 1 KW input with 2/3 the voltage across the series tube you will be dissipating 2000W in the modulator and about 250 watts in the final. Lighting all the tubes will put you close to 3KW of heat.  learned twice.

Yikes, that IS a bit poor, isn't it?

Well, maybe I'd better repair the PDM coil and give that one more go first.   And also give some thought to running the finals with grid bias or screen modulation.  I've done that several times in the past with good results.  Though I've found that the carrier output is usually limited to 1/2 the dissipation of the tube, even running class C.  Thus, we are about back to running a linear... :-)

I'll have to give this some more thought before tearing down the PDM portion.

BTW, the first PDM coil arced in about the middle across its own winding. Maybe about 1" of turns burned up. Not sure why it would do this - it wasn't arcing to ground or anything nearby....  Into the dummy load it was fine, but when on the 2:1 dipole, it arced.  First time this happened after hours of previous testing.   Moving down to 3725 is what did it - as it did FB up on 3885 into the ant before.

Looks like reactive final loading has an effect and reflects back as  higher impedance/voltage into the PDM coils...


BTW, Frank/AHE ; Good to chat with you last night too. Sad to hear about the power transformer crap-out of the 4X1 rig there. I heard you sitting on the floor working on some rig later on... is that the 4X1 temp fix? Also, please fill me in on your final config for your screen modulated 4X1.... did you use a transformer coupling into the screen or direct coupled tube -  any grid audio added, what HV, grid leak, screen voltage, grid voltage, any regulation on the supplies, how you tune it up with loading, drive, audio technique - what carrier max and how do the peaks look at what %?  That's a mouthful... :-)  Thanks!

T

[AB2EZ]

Tom
et. al.

The more I think about using an audio signal to modulate the grid-to-cathode bias... the more intriguing it seems.

Let's assume that we are not talking about a self-biasing configuration.

For a fixed amount of rf drive... we know that modulating the grid bias to be more positive (or the cathode bias to be less positive, in the case of a grounded grid configuration) will increase the rf output power (i.e., the envelope of the rf output will increase, but not in proportion to the change in the bias voltage). Likewise, we know that modulating the grid bias to be less positive (or the cathode to be more negative, in the case of a grounded grid configuration) will decrease the rf output power.

The time delay is negligible for audio frequency modulation.

So this is an ideal application for using feedback from an off air monitor/envelope detector.

I may try this myself with my grounded-grid GS-35b linear amplifier... i.e., operating it with fixed rf input, and modulating the cathode bias.

In the case of my linear amplifier, the adjustable cathode biasing circuit already employs a high current transistor as a current amplifier... so that will probably make this even easier to do. I need to look at the schematic of the existing cathode biasing circuit to see what the associate time constants are.

For experimental purposes, one could apply the modulation at a low frequency (say 10 Hz), and worry about increasing the bandwidth of the feedback loop if things work out okay with low frequency modulation.

An issue that requires some care is the following. Before you close the feedback loop, you need to add a dc offset to the output of the off air monitor/envelope detector... to null the output to approximately zero (AC coupling might work, but I'm worried about phase shifts at low frequencies that could cause oscillation in the feedback control process). That way, when you close the loop you won't produce a large swing in the DC value of the bias. After you close the loop, you can tweak the dc offset adjustment on the input to the feedback loop's differential amplifier to re-adjust the unmodulated bias level.

The big issue, of course, is whether one can introduce enough loop gain to get good linearity of the modulation, with a reasonable modulation index, ... and still keep the whole thing stable.

Anyway... starting with a small modulation index, a low modulating frequency, and a modest loop gain would provide some initial indication of how "practical" this is.

Stu 

[WA1GFZ]

Stu,
I was listening to you a while ago as I worked on my modulator and my only quesstion is input impedance. I wonder how it  will change if you go to class c?
Tom,
Maybe there was something unstable in either the final or modulator when the SWR was higher. Reflected power may have made its way into the modulator. Maybe they should be inside a shielded cage. I wouldn't give up on it yet because it sounds so good. Walk away from it for a few days.  BTW I was also upwards of 2 KW dissipation making a linear modulator for the class e rig. The worst case was at bout 30% positive when the top supply just takes over in the h mode.
The PDM filter inductor may have had a series resonance or acting like a tesla coil. Remember this also happened in the 1991 rig. It might be a good idea to have gaps in the winding every couple inches.  Make sure you have a good RF ground to the Final cathode circuit bypass caps. We talked about a good power supply ground at the switch tubes but maybe the RF ground needs to go to the final bypass caps so RF doesn't go through the modulator looking for a ground.
If Rf is looking for a ground it has to go through the filter. You might run a ground strap off the final to your ground system and don't tie it into the power supply return. 

[K1JJ]

OK Frank -

Tnx for the suggestions. Will do.

Well, I decided to repair the burned PDM coil and give it another try. Maybe the coil arced to the ground plane.  I added a Plexiglas sheet under the coils just in case.

Tonight I had it on the air with the guys. They thought it sounded FB. Gary made a recording and sent it to me. I'm really pleased the way it sounded. That's what I needed to hear to keep this rig going and continue to work out the bugs. It was getting frustrating to say the least.  I'll email you a copy to hear.

The next step will be to ramp up the voltage, but I will hold off to be sure the coils don't arc.

Later -

T

[WA1GFZ]

Tom,
I would like to hear the recording. I bet you were able to boost the lows more than the 833 rig. Sorry I missed you today. I was working on the modulator for most of the afternoon.  Everybody would be doing it, if it was easy.

[John K5PRO]

The Harris/Gates MW5 thru MW50 were the rigs from the 1970s that put a PDM tube in series with the cathode of the final RF tube. The Collins Power Rock 1 and 5 kW did same thing, with the switch tube below the cathode of the RF tube. In both brands, a high power low pass filter was in there too, as Tom has. The Power Rock schematic can be seen in my pages in Latest Articles bar to the left, under the Continental 314R1 model. I plan to add some stuff about the filter and PDM topology this week.

It is common to use a MOSFET in series with the cathode of high power pulsed amplifiers, as a pulser, to put the tube into and out of conduction. As Stu pointed out, it  may lead to distortion as a form of modulator. And since all the cathode current flows through the switch, it has to have high current rating, compared to a grid modulator.

[steve_qix]

Hi Tom,

I'm sorry you had a failure    A failure in a TUBE rig??  Seems so unlikely    But, enough of that !!

First, don't bother with class A series.  You've seen the calculations.  You'll be dissipating well over twice the power in the modulator as you will be putting out !!!  N.G.

Now, on to PWM.  At this time, I don't see a clear reason for the failure.  You may have a self-resonance in your PWM coil.  How big (in mH) is it?  The input inductor should be as small (in inductance) as possible, and still hold the PWM waveform when the modulator is in the off state.  If the waveform starts to fall off near the edges (not stay flat), you need more inductance.  But, you probably have plenty at this point.

The PWM coils should *never* fail.  I ran 10,000 (10kV) in my old tube PWM rig for YEARS, and NEVER had any failures whatsoever in that portion of the circuit.  The rig was often not tuned up correctly; the antennas came down; the final failed; etc.  No failures ever in the coils, damper diode or modulator.  Yours should be able to do the same.

I would look for a self-resonance in the coil or a *parasitic* in the modulator.  A parasitic would cause the modulator to put out high energy at frequencies which are not expected.  The key here is high energy at those frequencies.  This will give rise to voltages across the coils for which they are not designed.  Might or might not be the problem.

I have found, through much experimentation, that a wider, shorter coil seems to work better than a longer coil.  This is strictly imperical, through observation.  I have conducted no scientific investigations into this.

I went into the archives and dragged out my old PWM input coil.  For reference, I took a picture and attached it.  The other items on the table are for reference (the cell phone, the casette tape and the 1/2 gallon orange juice container, to show the size of the coil.  It is quite large.


Let me know if you need any help with this !
Regards,

Steve

[WA1GFZ]

notice the spaces between turns to reduce shunt C.

[K1JJ]

OK Steve and Frank -

Thanks for the suggestions.

No doubt that after thinking it over, PDM is the way to stay vs: the series modulator.

That's right Frank, as John Kennedy once said, "We choose to build a pdm rig not because it's easy, but because it's hard."   

Yep, it could very well be a pdm parasitic. I had some in there before and did lots of extra bypassing and even put suppressors in the grid, screen and plates. This seemed to kill them, at least from what I saw on the scope.  The parasitics seemed to look like white noise - at a higher freq than the normal 135kc pdm pulse.

The recent coil arcing occurred under heavy modulation, across about 20 turns in the middle of the first input coil.. But as I mentioned, I'm not sure if maybe it just arced to the gndplane and I since have added a plexiglas sheet as insulation.   So far no problem running it at the same parameters as the crapout Saturday.

Right now the filter uses a 70mh input coil and a 35mh output.  I've heard it both ways... use a big input and now use a small input coil. I suppose I could switch the filter around and make it 35mh in and 70 mh out, as a test.

I think the Tesla coil effect is big. I wonder if there is a way to tame down the tendency to develop such large voltages - like maybe with a corona bleeder or some other dampening idea?

It's too late to rewind the coil with gaps - it was wound on double-sided tape and would be a nightmare to start over.

I'll try ramping up another 1000 volts to what happens.  I gotta do it someday, might as well now. I only have about 1200V or so on the RF 4X1 now under steady carrier condix -  3500V under full modulation.

T

[w3jn]

Did that second recording I sent open for you, Tom?

[W3RSW]

Hmmm,  the old 1kw DC plate input days being gone, probably forever, and as others have stated, efficiency shouldn't matter so much and that it should be between you and your power company....

...and as you have already mentioned linears not being too bad an idea, all other things being equal ....

Why don't you comtemplate scaling up your very clean 6aq5 rig to say, 807's and feeding the output into an 8877?  Yeah, I know very expensive pube... but if properly treated should last a long time.

65w pep input can come from a very clean 10 (lots of head room) to 16 watt (arrl headroom) carrier to yield a full 1500 pep output.   You can imagine other interesting parameters from an 807 or 6146 rig with very clean audio and only twice or three times the working voltages of your 6aq5 rig.  Hey, 4d32's are going very cheap now too. 

[WA1GFZ]

Steve and I have not agreed on filther configuration since about 1983.
Looking at an interleaved buck modulator I wish I could make a small first inductor look good in simulation because it would be a lot easier. Someone please prove me wrong because 2-70 uh are going to fill my filter cavity in the modulator chassis. I will gladly send anyone swcad files or transient and attenuation. They can easily be modified. I will have just enough room for the second section but the third will have to be in the final chassis.
I find when the first inductor smaller two things happen in simulation. Transient you get a big overshoot and attenuation you get peaking near cutoff. The cleanest configuration I've have simulated so far gives a straight line phase shift through the band pass with no peaking near cut off. In transient simulation I shoot out 50 pulses and make a nice square wave with a couple percent overshoot. Smaller first inductor I see as much as 20% overshoot. (rise and fall time slopes limited by the frequency attenuation)
Now this big inductor can store a lot of energy so who knows Mr. Tesla
I come from a switching regulator point of view. The current slope is very quick in current mode feedback if the first inductor is too small.

[steve_qix]

Hi Tom,

Hmmmmmm. Maybe 70mH is too big and maybe it isn't.  Depends on the impedance of the load, and the overall design of the filter.  Where did you get that value?

The filter coil in the picture (I posted earlier) is 30mH.  This was part of a 6 element filter which included a series resonant circuit at the output of the filter, to ground.

30mH was probably the smallest I could get away with for the impedance in use, which was 8000 ohms (load presented by the RF amplifier).  If you have a lower impedance, you could use a smaller value inductor than this.

However, I am not saying this is your problem.

With 30mH, I got a near-perfect square wave at the output of the modulator.  The response of the filter was flat up to 15kHz or so (audio).  The input filter was set at 7.5kHz, to avoid putting audio frequencies into the filter which would become part of overshoot, ripple and other anomalies which are normally part of all filters.  This is very important.  The input filter (anti-aliasing) MUST have a lower cutoff (preferably quite a bit lower) than the output filter's cutoff.

Do you have a good overload shutdown?  It's probably important to have.  Then, at the first arc-over, the rig would just shut down and hopefully, no harm done!

Regards,

Steve

[K1JJ]

Did that second recording I sent open for you, Tom?

Yes, it opened FB, thanks John!   I know you made an accurate recording cuz Gary sounded like his normal mellow self.


Steve:   Frank and I did a lot of computer filter modeling as I worked out the bugs and more or less settled on this design. It showed good specs and a decent compromise.  When the rig runs, the pulse is square at the switch tube plate and the 135 kc sideband attenuation is down below 120db now, due to the improved rig/filter layout and series resonant trap.  The rig even hits negative 100% modulation cleanly, so I'm happy with the compromise.

So, I'll bet the filter design is OK for now. I need to exhaust the other bug possibilities first.

Today I'll ramp it up another 1KV and see what happens next... fun fun  -   


Oh, BTW Steve,  how would you suggest I make a shutdown circuit for the rig?  I have a breaker in the mains, but that is as sensitive as a rock.  Maybe a latching circuit that senses the current flow through the RF tube's fil CT.... or am I looking to measure the swich tube's current? At least the switch tube is at a ground reference point - while the final if floating and hot.  Any circuits would be appreciated.

T

[WA1GFZ]

Steve KA3KLR gave Tom a filter with 2 37 UH inductors. It looked good but switch noise sidebands were only down about 60 dB. After a bunch of simulation we arrived at a bigger input inductor. I think we use 500 ohm input and 2000 ohm output. I didn't look very closely at the transient response but when Tom added the second inductor the sidebands went away. I think he swapped them once but don't remember the results.
GFZ analog compensation. Slope the turn on of the FET which Tom seems to like. Reverse diode yanks it off fast. I was even thinking of some R in the Source lead. Yea, the fet heats up a bit but sure easier than all that extra hardware. Tom's drive to the switch tubes is the same as I used in my V2 but with an 11N90.  KISS
I was only guessing in the input Z of the filter. Not sure what a saturated tube does so just took the screen voltage as the saturation voltage times the peak plate current. All I know now is I can't hear the sidebands at my place and before the second inductor was added I could easily measure them at around   -60 dBc using my SDR spectrum display. Tom bought a high voltage probe so he can see what the switch looks like.

[WA1GFZ]

Gee maybe Steve's overload board could be modified with more turns through the hall sensor. Teflon wire would work in series with the FET drain. There will only be a few hundred volts there when the tube is off. The relay could be placed in series with the PTT line. It might take a gain change but it would be a cool circuit and it has inputs for SWR shutdown. I don't thing the efficiency meter circuit is worth doing at these voltages but who knows.