The "Dual Quads" PDM Tube Transmitter - under construction

[K1JJ]

This is a continuation of the series modulated thread, but it has morphed into a PDM modulated moderate-powered rig.  I decided that the class A series modulator was just too inefficient to bother with.

This is a rather unique rig. I wanted a tube transmitter that would function at relatively low voltage. PDM transmitter voltages can get pretty scary. Preferably clean, transformerless modulation with big peak capability and a quiet cooling system. (radiation cooled)   It must be compact, the size of a DX-60.  I want it to be easy to dance to and have a good beat.

The choice was quad 6LF6s as the switch tubes and quad 6146Bs as the finals.  With 0V to 1600 V  available on the whole thing it should be good for about 250 watts carrier and 1500w pep, maybe more. (If I can make it real efficient... 90% switch tubes, 80% finals eff.) I have a Variac on the HV supply, so can be used as an exciter too.  There is a muffin fan that can be switched on for extra cooling if needed..

It will be driven by the new DDS VFO and SS pre-amp. I am hoping to use an unturned toroidal input transformer to avoid input tuning.  The output is a conventional pi-network tank.  The switch tubes will be driven by a solid state PDM generator (WA1QIX design) and a 11N90 MOSFET.

Frank WA1GFZ suggested that I mount the 6146Bs RF tubes on a floating aluminium plate that is bypassed to ground for RF.  This will permit short connections from the cathode / grid and screens while permitting audio current from the switch tubes and PDM filter to stay above ground to modulate the finals.   With four 6146Bs in parallel, the wide plate will help stability too. Jay / W1VD has suggested some ferrite cores to wind a compact PDM filter.  My thanks for their help!

Notice that BOTH filament transformers have their center taps connected to the cathode of their respective tubes. This will keep cathode to filament voltages within specs. Notice that the secondary of the 6146B filament transformer floats at high voltage. I wound HV insulated wire on a Variac core and pruned the turns to give 6.3 V under load. Pictures later.

For four 6146Bs running at 600 V plate, -92 volts grid bias and 175V screen volts,  I calculate a grid leak resistor at 6.7K / 1.2 watts. The screen dropping resistor is 11.3K /16 watts.  Actual resistor power ratings will be rated higher in practice.

I will start posting construction pictures within a week or so.  The panels and chassis are done and need to be painted.

The RF deck is compact - about the size of a DX-60.  The power supply is slightly smaller.

I have most of the parts mounted in both the power supply and transmitter but can still change the basic design before I wire it up.

I plan to use .001 1KV disc bypass caps for the screens on both the 6LF6s and 6146Bs.

The photo of the schematic is more readable if you enlarge it and pan around. The dark lines show the floating plate.  Anyone see errors or have suggestions before I wire this sucker up?

T

[John K5PRO]

Take a look at the 1 kW Power Rock BC TX schematic on my website under Latest Articles. Its another variant on the theme, using triodes throughout, 3-500Z. The deck of the RF tubes is, like yours, also the cathode potential. Since it must be bypassed to ground for RF voltage (not PDM) the same cap is also the terminating cap for the PDM filter. The Switchtube is grid driven via a MOSFET card driven by fiber optic link.

[W3GMS]

Take a look at the 1 kW Power Rock BC TX schematic on my website under Latest Articles. Its another variant on the theme, using triodes throughout, 3-500Z. The deck of the RF tubes is, like yours, also the cathode potential. Since it must be bypassed to ground for RF voltage (not PDM) the same cap is also the terminating cap for the PDM filter. The Switchtube is grid driven via a MOSFET card driven by fiber optic link.

John,
How do we find your website?
Joe, GMS

[Rob K2CU]

Tom,

I have 10 mil kapton sheeting in a 12 in roll. I gave GFZ a piece to use with the class E rig. I could give you some for making an insulated mount for the floating RF deck. It would make a great bypass cap and mounting structure at the same time. The RF deck could sit on a cut out aluminum chassis with an inch or two of overlap with the kapton sandwiched between, and held in place with nylon screws. wadaya think?

[steve_qix]

A question first.... what bands is this rig going to cover (just curious)..

2nd -  Will the 6LF6s pull enough current with only 0 V on the grid (or are you expecting to drive the cathode somewhat negative)?  It may be that you have to provide some forward bias on the grid to get the tubes to turn on as hard as possible and achieve the efficiency you want.

3rd - being that this is a tube pulse width modulator, you will most likely need some form of analog compensation.  This can be done completely in the PWM driver of course, but with tetrodes it might be just as easy to apply some analog compensation to the screens.

The analog compensation accomplishes two purposes.  First, it linearizes the tubes, which are not great switches.  You drive the tubes harder as the on-time increases.  Second, it compensates for the problem of needing to completely discharge the PWM filter over all pulse widths.  The Problem you must solve: As the on-time gets smaller and smaller, a point will be reached where the PWM filter will not discharge completely (the PWM waveform will start to integrate).  This will result in a flattening out of the modulated waveform (usually starts at around 75% for most tube modulators) with a sudden drop to 0V when the on-pulse reaches the 0% duty cycle point.  To compensate, you drive the modulators less and less as the on-pulse gets smaller, until the modulators are almost operating in analog mode.

Another way to correct the problem is to provide an active pull up in the PW modulator itself.  This could be an additional tube with a differentiated pulse fed to the grid.  With a tube, the heather and cathode of the pull-up must float.

Another idea might be to use high voltage IGBTs in the modulator.  These are available up to 4000V !!!  You could make the whole modulator from IGBTs (make sure the circuit incorporates a suitable overload protection scheme).  You will still need the active pull up, but it is much easier to implement with an IGBT than with a tube because there is no heater or screen to worry about.  With IGBTs, you will not need analog compensation to linearize the modulator devices because they are near-perfect switches.  You still need to discharge the PWM filter for small on-time duty cycles. Analog compensation or an active pull-up may be used to solve that problem.

Note: Using an active pull up will reduce the efficiency of the modulator slightly, but will improve the performance greatly for small duty cycles.

It's an interesting technical project, albeit a lot of effort to drive a linear that's going to be - maybe on a good day - 30% efficient anyway  (isn't that the original intent of this transmitter, from the other thread?). And who who cares about another 100 watts of dissipation under those circumstances. At that point, with the amount of power lost in the linear (including the power lost in the tube filaments, etc), any additional power lost in an analog modulator for the driver is almost negligible  Just thoughts!

[K1JJ]

Appreciate the comments and suggestions!  I will incorporate some.

Steve:

160 - 40M for now. Not neutralized, but will be loaded down at the grid with a broadband input.  It uses a vacuum variable and reasonably short tank leads.  Yes, I could use solid state devices for the switch, but I wanted to keep it a tube rig, at least the switch and final. I've built a few solid state PDM rigs and big tube PDM rigs already - but this is my first low power tube version.  It's all in the spirit of a tube rig, like riding a Harley vs: a riceburner.

OK on the need for a positive grid bias for the switch tubes.  I have a variable screen voltage already on the switch tubes, from 0-250volts. Will this do it or will a resistor voltage divider off of this screen supply to the grids be OK?  IE, the positive grid voltage doesn't need to be regulated, right?

Frank mentioned that I could add analog compensation by adding a little resistance in series with the FET gate to control the turn on slope. Then a fast diode across the resistor to get a fast turn off.

He also pointed out I forgot to draw in the damper diodes.. :-) Totally forgot.    He also said: ".... 6- 600 volt PIV diodes in series. You need to look at diodes to find the fastest reverse recovery diodes you can find. 3 amp diodes will work. Heck maybe the damper diodes Steve uses in his PDM kits might work if you series up a few. I think those are fairly high voltage."     I still have some of those diodes around. Can't remember their breakdown voltage.


OK on the Kapton, Bob.  I will probably just use a single plate that is mounted on four 500 pf door knob caps. Simple  and solid.  Sub-mount the tubes a bit to hide the plate structure from the grid, etc..

The rig will certainly need some testing and tweaking to dial it in. I don't expect it to be as true as a solid state PDM rig, but if I get shark fins, I'll be happy.  HA!

Steve, I probably will not use this as a driver cuz it is probably capable of 300 watts out if pushed. The 4X1 linear is good for maybe 500 watts, so why bother?  But I'll see how it performs when the Variac is backed down to 800 volts or so.   The TOTAL filament power of the rig is only about 75 watts, so not too obscene to use as a driver when backed off..



Frank came up with some more excellent suggestions:

"1. you have no fixed bias on the final. This was very tuff on my finals in the V2 pdm rig. You could add a winding to your final filament transformer and generate a floating negative bias voltage that goes between the cathode plate and bottom of the grid leak resistors. Or get a 10 watt zener and mount it between the cathodes and cathode plate with a couple caps across it..

2) Step up transformer. Take 2 of the type 43 cores from the PDM rig and wind a couple turns for the primary. You might need as many as 4 for the primary. put it across the output of your vfo and monitor the signal with a scope. You will have enough primary turns when it doesn't load the VFO. Once you establish a good minimum turns count then you can wind the secondary. I'm not sure how much voltage you need. once you find the turns ratio you need put a load resistor across the secondary so the transformer will have a resistive load. "



T

[John K5PRO]

My website on the Continental/Collins Power Rock PDM 1 kW tube BC rig is on the left side of this forum, first item in Latest Articles. Click on the 314R1 transmitter for details. Page 2 of the schematic shows the high power components, as the rig used three 3-500Z triodes, one for switching. The PDM compensation circuit was analog, consisting of diodes and RC around Q6 on the switch tube driver card. It is discussed in detail in US Patent 4140980, which can be downloaded from the page. Patent 4187467 discussed the general layout of the modulator and RF stage for the 5 kW (same design), that allowed DC grounded output components without blocking capacitors. The trick was to have the DC modulated voltage on the cathode of the final, as a negative polarity w/respect to ground. I wrote a couple of pages about the modulator and LP Filter that can be downloaded as well.

I just spent a few hours with Jack Sellmeyer on Dec 13; he was a member of the design team of this rig at Collins Broadcast. He's active on AM in the Dallas area.

Some of this may benefit K1JJ on the new radio. Take a look at a commercial rig for ideas!

[K1JJ]

Thank for the info, John.  I looked over the patent and write-up.  They obviously put some time into the compensation circuitry.

Well, I added a few things to the schematic.  (Thanks to suggestions from Frank / WA1GFZ)

1) 6.8 ohm, 3 watt wire-wound resistor parasitic suppressors to the switch tubes.

2) A 35V, 50 watt zener diode to the cathodes of the 6146s to produce a fixed bias. This will keep the tube from melting down if drive is lost.  Changed grid leak to 5K. Now fixed and GL bias total -100 volts, which should make it hard into class C.


3) A load resistor across the secondary of the input transformer to keep a steady load on the input.

4) I plan to add a step start to the HV and screen supplies.

5) Compensation to the 11N90  MOSFET. Values to be determined.

6) Six 600 volt damper diodes in series connected from the anode of the switch tubes to the B+ to suppress switching spikes.



It’s been tough going with the metalwork.  I machined out the floating plate for the 6146s. I don’t know how I screwed up so bad, but the 6146s were too close to each other once I plugged them in. They have big bases that overlap the socket.. cheezzz...   What a space shot.  Had to make a new plate. The 6LF6 plate came out OK.  

It's turning into a bigger project than I first thought, but should be a cool little compact tube rig that can do near the legal limit when finished.  

T

[John K5PRO]

Another way to compensate (patented) is the late Hilmer Swanson's scheme that Gates Radio/Harris used in the early 1970s. In the original patent Hilmer used pentodes for example (attached), and tied the first plate to supply through an inductor shunted by a resistance. He essentially pre-distorted the rectangular waves by saturating the next tube when DF was high (positive peaks) and reducing amplitude to underdrive it for low DF (near cutoff). There is level translation with diodes to the second tube, bypassed for higher frequencies with a cap. In the actual transmitters such as the MW 5  to 50, they later used solid state audio drivers to the switch tube. A page is attached from the MW10 manual. L1 is used in similar fashion, shunted with a variable R. Q2 is the driver transistor that directly drives the grid of the switch tube, a 4CX15,000A.

There are many ways to skin the cat, but for high fidelity, you have to avoid the pinch-off distortion that happens with tube PDM. Driving your rig with a clean 400 Hz sinewave, you can demodulate and see what looks like an inverted mountain, or a upside down nipple near carrier cutoff on the negative trough at high % modulation. This sounds bad.

[W3GMS]

Take a look at the 1 kW Power Rock BC TX schematic on my website under Latest Articles.

How do I find your website?

Joe, GMS

[John K5PRO]

Its here, also listed on the AMFone Latest Articles section, "Rescues of BC transmitters":

http://www.jtml.info

[flintstone mop]

I miss the BC1-H. Plenty of cheap Chinese 833's around and so easy to plunk them in the sockets to check for RF emissions. One tube at a time in the PA. 150ma of drive told me that the tube was in good condition.
I was lucky to have the solid state audio driver for the modulator. I never liked the quad of 807's for audio. The TX audio was superb. I really regret getting rid of mine. A nice choice of being slightly over the limit to being the channel master.
The protective bias was defective in the unit I had. And that explained how the "glass" was melted in one corner from a run-away tube in the PA. Musta lost drive !!
I'm sure Tom Vue will have pictures of the latest creation and it operating in a few days. The guy is amazing how he can throw something together and have it on the air.
Fred

[John K5PRO]

The BC1H1 that I got from KD0HG had the solid state audio lineup. That rig was a 'value-engineered' version of the BC1G. Something like $1200 had to be removed from the cost of the G, and that was the basis of design for the H.
It's the last 833 rig to come out of the Quincy Tin Works. Next was the MW1.

[K1JJ]

Well, almost ready to fire it up. Still a few connections to solder or bolt together as you can see in the pics.   I tested most of the modules, so it should work except for the usual instabilities, arcing and the like.  Shakedowns are hell sometimes.

The rig is very compact -   6" high, 17" wide, 14" deep, 35 pounds. I used a nice cabinet that was really too small, but I couldn't resist. The HV power supply is also compact. It uses a strapping 240V primary, 1700V sec,  1 Amp power transformer with caps @ 3KV / 100ufd total.  (step start used)  At 1600 VDC under load, it should feed this PDM rig well. I'm hoping for an easy 300 watts of carrier with big headroom for audio.

So far the filaments light up.  

The 6146Bs use a 33V 50 watt zener for protective bias and about 70 volts of grid leak. The screen is tapped off the HV supply using a 11K 30 watt resistor.  Notice the black fil transformer made from a Variac core and HV wire for the 6146B floating fils.   I am using a 11N90 MOSFET to PDM drive the cathodes of the 6LF6 PDM tubes.  

It uses a broadband RF input network using a  50 ohm to 1K toroid. The input toroid feeds the floating HV 6146B mounting plate thru doorknob caps for isolation. I loaded the grids down with about 2.5K of resistance via the grid leak and additional  cap/resistors to ground.  It will be driven with the adjustable 20 watt DDS VFO, so plenty of power for the extra grid loading. The 6146B mounting plate floats above ground to accept the PDM audio.  The PDM filter is wound using toroids - and banks of capacitors. Pretty compact in itself.

The plate tank has a low impedance of about 1000 ohms, thus the smaller tank coil and the need for a 20-2000 pf vacuum variable for plate tuning. It is designed to cover 160 - 40M.  C1, plate tuning requires about 600-800 pf  on 160M!   I would have pushed the vac variable back with a turns counter, but no room. If it works out, someday I may do that and remount the coil.
  
I needed some extra 160M load padding due to the low plate impedance.  

In Pic #3 notice the string of diodes on the left side used to snub the PDM spikes back into the HV power supply.

* Except for the HV wire, Teflon wire was used underneath the tubes and wherever possible.

We'll have to see how it performs over the next week of trials.

T

[K1JJ]

Power supply and more shots.

The HV  and screen supplies have a step start for easy key up.  There are separate Variacs for both the HV and PDM tube screens.   The 6146B screen is supplied by a dropping resistor off the HV.


Pic#3 shows the Plexiglas cover and muffin fan to push air over the tubes and keep a good flow thru the gaps around the cabinet edge.

Notice the plate current meter in the front panel  is floating with Plexiglass supports. By putting the meter in the HV lead, this allows me to use very short source connections to ground on the PDM MOSFET for better stability.

When finished, there will be no exposed HV points.

T

[K1JJ]

Additonal views:


Pic #3:  PDM filter close up and the HV resistor string for the HV meter.
It is had to tell, but the PDM toroids are mounted on Plexiglass and spaced 1/2" off the steel panel for better coupling isolation.


Fred, your resistors arrived today - thanks!

T

[flintstone mop]

Not hardly a month goes by and there is something getting ready to be plugged into the nearest electrical socket. It's amazing what you put together in your K1JJ labs, Tom
Fred

[Tom WA3KLR]

GL OM.  I can't wait to hear u on the air Mr. Vu.

[KA2DZT]

Tom,

Why did you use sand WW resistors for the plate suppressors??  I always use a carbon resistor with the coil over it.  I thought the 50 ohm resistors were for the suppressors.

Seems you end up with a coil within a coil.

Just wondering

Fred

[K1JJ]

Thanks for the kudos, guys!

Well, I fired up the RF deck by itself tonight with 600V and it actually put out reasonable power.  (the PDM modulator tubes were bypassed)   The DDS VFO drove the broadband 6146B input FB.  The low 1K tank impedance seems right on the nose for sharp tuning and good power on 75M anyway.

Problem is I found that the 6146B tubes were not drawing enough screen current, even though I measured about 250V on the screens. (should be 175V)  I increased the dropping resistor from 12K to 25K and the voltage is now about 200 V but the total screen current is only about 20 ma. It should be about 38ma for four tubes. I played with the loading and tuning but still, the current is low.  I tried four new tubes, but same thang.  I think I had enuff drive available from the 20w DDS - the tubes seemed to go into saturation.  I have no grid meter right now, just a grid leak from grid to cathode, floating at HV.

With the 250V on the screens I can get about 250w out. But with the lower screen voltage and current, it is limited to 150w out. Strange.   The recommended voltage is 175V and 38 ma for 4 tubes, plate modulated class C service.


Fred, the 50 ohm resistors I used for the RF deck are metal films that I found in my stash. They do look like wire-wounds, I agree.   The rig seems very stable on 75M so far, so if I have problems on 40M, I will switch over to the carbons you sent me.

TMW I plan to try 160M and 40M to see if it tunes up there. If so, I'll give the PDM tubes a try and see if the overall modulation system works.

This is turning into a pretty cool rig - really pleased with it so far. I really needed a quiet, efficient, medium power rig for those lazy late afternoons and early evenings when the band is local and loud....  Shuda built one a long time ago.  

T

[John K5PRO]

Looks interesting! I had the same thought that Fred did, those parasitic choke resistors look like sand/wirewounds. If they have inductance, they won't dampen the parasites well.

Two other comments: The HV current meter will become a dust collector when floated at positive potential. Some cheap BC rigs did that, and they ended up being a headache as the meter was hardly readable after so many years unless cleaned regularly. Make sure you cannot adjust the zero screw on it while running.

Do you have a good RF bypass cap at the cathode end of the 6146's to chassis? As it is also the output terminal for the switch tubes, this capacitor may be part of the PDM filter. Without an RF ground here, however, the screen, grid and plate may be working with a longer inductive path to chassis, with reactive voltage drops.

[steve_qix]

Hmmm.. You might need something to measure the grid current.  Either some type of voltage indicator across the grid leak (I'm assuming it's not hot with RF, but maybe not), or a traditional (floating) current meter.

For whatever reason, I thought you were using 6DQ5s in the RF output.  The 6DQ5 is a nice RF tube because you can pull a lot of current with relatively low voltages (for tubes).

Back in the day, I ran 6  6DQ5s in parallel.  The rig ran at the 600 watt level no problem with 600V on the plates.

Anyway, something to consider if the 6146s don't work out.

Regards,  Steve

[K1JJ]

John,

Yep, I just ran an MFJ 259 test comparing the 50 ohm metal films I used against standard carbon 50 ohm resistors.  They are identical with an swr of 1.2:1 at 40Mhz.  Both have little to no inductance below 30 Mhz as was advised here.

That big red capacitor next to the 6146B floating plate mount is 4600 pf and bypasses the cathode. I also have the 6146B mounting plate sitting on five door knobs to ground. These caps are all part of the PDM filter's last cap.


Steve,

The grid leaks go directly from grid to cathode, thus, there is RF there and difficult to tap for metering.  I usually meter grid and screens, but thought I'd get away with short connections this time. But I'll figger something out for metering.

Gee, I didn't realize the 6DQ5s were that robust for current. 600 watts is a double what I expect with the four 6146s.    The four 6LF6 pdm tubes will be more than enuff, and wud probably modulate six 6DQ5s.   

Frank says he has a breadslicer that will fit in place of my vac variable.  I usually like to use a counter dial with vac variable, but not enuff room in there.

T

[KA2DZT]

FWIW,  I remember with my single 6146 rig the screen voltage ran high, I think about 200 volts when it was suppose to be around 150 volts IIRC.  I never did anything about it, just let it run at 200 volts.  I think drive levels will affect the screen current.

Fred

[K1JJ]

I found the problem with low screen current was simply not having enough RF drive. It needs about 15 - 20 watts of drive because I loaded the grids down heavily for stability, being that it is four tubes in parallel.  So, the MRF-150 IPA amp that drives the 4-1000A rig will be included in the chain, no big deal.

With the proper drive I now get the normal 175V on the screens at 38 ma.  The grid leak shows about -95V  at  16 ma.  (four tubes)

It loaded up nicely on 160M and also on 40M with no problems.  It will put out 300 watts at 700V with used tubes, so looks OK.  I have a  NOS set of 6146Bs that I'm saving until the bugs are out.

Putting the 33V 50 watt zener in the cathode for protective bias works FB, but it robs 20-30 watts of output power. Frank / GFZ suggested I make up a floating -35V negative supply to feed the grids for protective bias and get a free ride.  Might do that.

I found a way to monitor grid leak bias voltage, thus current - and to monitor screen current. They both require floating meters in Plexiglass.  It was like flying blind without them.

So on to the PDM modulation testing.  That's when thangs should get interesting since we will be going from 600V carrier to 1600V peak modulation.

T