How much peak power can a class C 6146B put out?

[KC2ZFA]

Steve, can your pwm generator board simply plug into the 6883 rig linked to in the post above yours ?

[steve_qix]

It should work although a MOSFET is better than a bipolar (it would still work either way).

Beware: That 6883 circuit has fundamental weaknesses that will yield less than ideal results.

[K1JJ]

That 50 watt 6146 schematic has an interesting feature...

1) Notice he uses a single zener diode for fixed bias in series with the GRID leak resistor instead of placing it in the cathode. This way no power is robbed from the plate circuit.  I've not seen that done before in a class C stage to replace the protective bias fixed supply.  

Would there still be enough residual grid current to drop a voltage across the zener if RF drive is lost?


The builder obviously knew his way around transformers, as some of them look homebrew. Coming up with the perfect PDM filter inductors in compact form like that is not especially easy.


Maybe with  a few mods it would make a great little driver rig.


T

 

[N2DTS]

I think I have that same meter in my junk box.

And the chokes? look like the plastic spools radio shack sells wire on.
Just buy 4, glue them together and hook the wires together?
That is slick, no need to even unwind the wire...

Here is one such PDM transmitter using two 6883s (12V version of 6146).

http://www.lu8jb.com.ar/PWM_6883.htm

The switch tube grid is switched in turn by a HV bi-polar transistor with its base switched by the PDM signal of about 5V.

Phil - AC0OB

[steve_qix]

That 50 watt 6146 schematic has an interesting feature...

1) Notice he uses a single zener diode for fixed bias in series with the GRID leak resistor instead of placing it in the cathode. This way no power is robbed from the plate circuit.  I've not seen that done before in a class C stage to replace the protective bias fixed supply.  

Would there still be enough residual grid current to drop a voltage across the zener if RF drive is lost?

I've used the zener / capacitor approach in my own PWM transmitters (in the modulator).  In fact, the tube PWM article shows an "adjustable zener" for such a purpose.

The zener fixed bias circuit in the RF amplifier technically isn't needed because typically the modulator tube is simply turned off, and there is no voltage across the RF amplifier at that point..... so no protective bias is needed.  Usually at least from a linearity standpoint, 100% grid leak bias has typically worked better, at least in my RF amplifiers.

But, if one did choose to use it, the circuit's ability to maintain the fixed bias is dependent on the leakage of the capacitor(s), and other components that might drain energy from the fixed bias storage capacitor.

[DMOD]

...The 4d32 would be better, or if you can find them cheaply (and I think there are a lot of these out there) the 6DQ5 - which is a very good switch !

Addendum: This is a WIP.

My feelings exactly.

After my next two HB projects are completed this is what I was going to use.

I have a bunch of 6DQ5's from an HT-40 restore so I thought I would use them:

[K1JJ]

The zener fixed bias circuit in the RF amplifier technically isn't needed because typically the modulator tube is simply turned off, and there is no voltage across the RF amplifier at that point..... so no protective bias is needed.  Usually at least from a linearity standpoint, 100% grid leak bias has typically worked better, at least in my RF amplifiers.

I was thinking more protective bias is needed for a dumb mistake, no drive or cata-crapout short situation. Like when the modulator itself shorts or the MOSFET driving it shorts. I've had this happen several times and it will put full voltage on the RF final.  A shutdown circuit would be even better for the final, but more complex with floating circuits. A 33V 50 watt zener is cheap and simple insurance to cover most problems occurring outside the final.



T

[WD5JKO]

That same web site has many PDM rigs shown. This one is solid state, 220 watts for 40M AM:

http://www.lu8jb.com.ar/PWM_460_40m_(2).htm

He has many of the schematics there...The PDM is based upon a NE555 and an LM311...Quite simple!

Jim
Wd5JKo

[DMOD]

It should be pointed out that lu8jb uses a 75kHz sampling frequency whereas Steve's PDM's use about 110kHz so the PDM filters will have different component values.

75kHz was the standard sampling frequency for some of the Harris/Gates PDM rigs so it should be more than sufficient for Ham audio.

However, I prefer the higher sampling frequency of Steve's circuit since it should yield slightly smaller values of PDM filter components.

Phil - AC0OB

[steve_qix]

That's a good point about the switching frequency.  The main reason why such a low frequency was used with the tube PWM transmitters is because, for various reasons (that have already been covered here), it is very hard in tubes to maintain the PWM waveform over the entire range of duty cycles.  So, there is less mathematical integration of the waveform as a percentage of the entire cycle at lower frequencies than at higher frequencies.

The old Harris PWM rigs used 70kHz, and I used around 70kHz in all of my tube PWM rigs.  It is possible to change the PWM boards to put out a lower switching frequency by changing a couple of component values.

[DMOD]

The old Harris PWM rigs used 70kHz, and I used around 70kHz in all of my tube PWM rigs.  It is possible to change the PWM boards to put out a lower switching frequency by changing a couple of component values.

Yep, that was a typo. Harris' PDM generators generally used a 70kHz standard.

l8ujb uses about 75kHz which is 5X15kHz, 15 kHz being the highest audio frequency expected.

In one implementation, Harris/Gates used a 35kHz cutoff, 3 inductor PDM filter system (with appropriate caps). There was a feedback path, tapped right after the first inductor, that fed back to the PDM module for distortion correction.

My plan for the circuit was to see if it could operate at the approx. 110kHz frequency with the 6DQ5 sweep tubes, and if not, then change the component values at the RT and CT ports of the 25701 PDM chip.


Lots of fun stuff to play with. 

Phil - AC0OB

[WD5JKO]

4 - 6883 RF, 2 -6JM6 PDM => 400 watts AM!

http://www.lu8jb.com.ar/PWM_6883%20X%204_(1).htm

No schematics, but good photos of what appears to be well thought out modular construction.

Edit: added a schematic of one of his QRP PDM rigs...uses the Ti494 PWM chip....

http://www.ti.com/lit/ds/symlink/tl494.pdf

Looks like something to play with and no high voltage! See attachment.

Jim
Wd5JKO

[AB2EZ]

Steve

I added the red highlight to a quote from one of your earlier posts.

Why not do the following:

Between the output of the PDM filter and the modulated RF stage... add a Zener diode with its cathode connected to the modulator output. This will add a negative offset voltage (equal to the Zener breakdown voltage) to the modulated B+ that drives the RF stage. As a result, the PWM will not have to go all the way to zero duty cycle pulses to produce 100% negative modulation peaks.

Stu

You really should have analog compensation in your PWM driver.  This is discussed somewhat at length in the PWM (PDM) document - a link to which is in one of the previous posts.  Analog compensation drives the tube harder as the On-time of the pulse increases, and drives the tube less hard as the on-time decreases.  This both linearizes the tube (tubes are not perfect switches), and works to allow the transmitter to be modulated to 100% negative cleanly.  Analog compensation is done in all of the tube PWM broadcast transmitters that were made by Harris, etc. for this same reason.

[steve_qix]

Hi Stu,

I was wondering when we were going to hear from you !!!!  That is a good suggestion, and I had actually tried that idea way back when (wow, we're talking 1973 here!!).

Unfortunately, it didn't correct the problem because the problem is not necessarily resolution of small pulses (although that is important), but rather it is the issue of discharging the PWM filter (and associated stray capacitances) when the current falls to near 0.  There isn't enough current.

To solve the problem, I had come up with several experimental solutions:  An active pull down is one.  That definitely works, but it does affect the efficiency slightly.  Ok, I thought this next one was clever - use an inverse modulator that discharges increasing amounts of energy from the filter (as the on-pulse falls to 0), and recycles the energy back to the power supply.  Or the same sort of thing that takes a constant amount of energy from the filter and recycles it.  Never actually built this, but did come up with some designs that seemed reasonable.  You see this sort of thing (energy recycling) in many switching power supplies, and I've done it myself in transformer coupled PWM modulator/power supplies.

Ok - why, someone might ask, doesn't this same filter discharge problem occur with the class E rigs.  Same thing is happening - the voltage will eventually fall to 0 (or near 0) and there won't be any current to discharge the filter.....sounds right, yes?.... well, NO... but only because of what I call the "class E gift".  This is an odd side effect of MOSFETs, where, if you drive the MOSFET with RF, the MOSFET will produce a NEGATIVE VOLTAGE on the drain (in the absence of any other drain voltage), and at a reasonably significant current, and this discharges the PWM filter - even with 0V output from the modulator.  So that's why there is no active pull down on any of the solid state pulse width modulators that we use with the class E rigs.

I've got a Valiant around here that I am very tempted to modify with a PWM.  The original modulator isn't there anyway.  The hard work will be to float the negative lead of the RF amplifier, and I haven't looked into all that's involved there, but it can't be that hard.  Put a bridge around the power transformer (with a cap input power supply) which should yield around 2000V (or a little less).  Use a couple of 6DQ5s in the modulator and the publish the design.  And, fit all of it inside of the original Valiant case, of course!!

It'd be an interesting project for sure.

[AB2EZ]

Steve, et al.

Yes... very interesting!


A great example of: "the Devil is in the details"

I'll move back into "listening mode" on this discussion.

Stu

[K1JJ]

Update:

Got the IGBTs for the modulator today and mounted them on the heatsink.  Hope to try some tests by tonight. Received the triangle audio generator too, so loaded for bear.

Steve, I'm gonna try it first without the pull down circuit to see how it looks and go from there. Say a prayer that two $45 devices don't go up in smoke...

A PDM Valiant would be a hoot.  What ever happened to your 500 watt PDM Ranger?


BTW, it's interesting that the LU8 built a quad of 6146s X  two 6DQ5 PDM rig. I didn't see a schematic with it. Maybe he's the "QIX or Tron South of the Border."  


T

[K1JJ]

MUCH BETTER, but a new problem.

I fired up the IGBTs and got the rig working.   At about 500 volts on the rig (200V on final and 300 on modulator) the 100 HZ triangle tone looked good.  The high end above 1700 HZ was much cleaner, but still not perfect above 5 KHZ.   I think the problem with response may be the screen bypasses, so will change them to 250pf each.


* But the big problem now seems to be that the IGBTs are going into a slow thermal runaway when the voltage is above 600 volts.  When the modulator is not in the circuit (cathode of the 6146Bs grounded)  the RF output is stable, even with 800V across the finals.   But when the modulator is working, when above 600V, the plate current  starts at about 400 ma and then after 45 seconds it creeps higher and higher until it pins the plate at 1A.  The IGBT collector waveform starts to get rounded at the top and shrinks down as the power increases in runaway.   Everything is stable when at 500 volts for five minutes at a time.  But go above 600V and it starts to creep.

I felt the big heat sink and it appears to get warm all over its surface quickly, so there is good thermal transfer.  The IGBT case does not seem hotter than the heatsink, which is about 320 sq ".   I tried it with a fan blowing on the heatsink, but it did not have a big effect on the minimum level when the runaway started. That seems strange, but may be related to the IGBT junction.   I am using a copper spreader too.

The pulse on the IGBT input looks very square with no spikes and stays stable thru out the runaway event.  The collector output has no spikes, but is slightly more rounded at the top  - but doesn't seem much more rounded  than other PDM pulses I've seen.  I see no evidence of parasitics.

I have a 30 ohm resistor is series with each gate. I shorted them out and it made no difference to the runaway problem. I also have a 320 ohm resistor from gate to emitter for each IGBT for stability.

I'm using two stacked pink Sil-Pads to mount each device.

Maybe I need resistors in the emitters for thermal stability?   The 6LF6, (140 watt total) tubes handled 2KV no problem, but the IGBTs are stuck at 600V and are rated at about 320 watts total.  It can't be dissipating THAT much power even though the pulse isn't  perfect at only 600V. . What gives?

T

[DMOD]

Do you have a current schematic of the circuit you can post?

Phil - AC0OB

[K1JJ]

Nothing drawn up yet, but I will later, Phil.

I just realized that my input pulse AFTER the 30 ohm gate resistor looks rounded. It is square coming outa the gen.  I have it acting too analog right now.  I need to change some values. 

Also, I am using two stacked SIL-Pads.  Frank tells me the thermal resistance is too high with two and probably causing the junctions to get too hot.  

So I might just mount them on the edge of the copper with paste and do it right. They have built in insulated pads for 4KV, so might as well.  The leads are too close to the heatsink, thus my reason for the SilPads.

I'll get this yet.  The audio is starting to come alive and look decent for a change.

T

[steve_qix]

Steve, et al.

Yes... very interesting!


A great example of: "the Devil is in the details"

I'll move back into "listening mode" on this discussion.

Stu

Well, don't listen too long.  You are very often the "voice of reason" in technical discussions!

[WD5JKO]

So I might just mount them on the edge of the copper with paste and do it right. They have built in insulated pads for 4KV, so might as well.  The leads are too close to the heatsink, thus my reason for the SilPads.

  Tom,

   This issue sounds thermal for sure..

One thing I did once was to put the electrical isolation between the copper spreader and the heatsink, while mounting the power device directly to the copper spreader. This way I get the low thermal resistance between the transistor and the copper spreader, and the next thermal barrier between the copper, and the heatsink was also very good since the area insulated had a lot of square inches. I used a overhead transparency sheet for an insulator, cut with a scissors, and holes made with a hole punch. A Kapton sheet would have been better, but I used what I had around. Used insulated shoulder mounts from old TO-3 transistor mounting kits along with counter bored holes in the copper spreader.

This was in a DC power supply where I only needed to hold off 50 volts. I was not worried about capacitance to ground..YMMV.

For thermal grease, consider the microprocessor stuff, such as Artic Silver Five:
http://www.newegg.com/Product/Product.aspx?Item=N82E16835100007
keep in mind that it's electrically conductive whereas the white "Owl Sh_t" is not.

Jim
Wd5JKO

[steve_qix]

Tom, it does sound like a thermal runaway.  They are bipolar transistors (with an insulated gate), so thermal runaway is definitely a real thing.

The bigger question is: why are they getting SO hot?  Normally, PWMs don't get overly hot.  I modulated 6 6DQ5s in parallel with 2 LITTLE TO-247 package IGBTs, and the heat sinks barely broke temperature.  I used a pure square wave drive, and drove them to +12V. This is important.  They have to turn off all the way, too.

Something's amiss.  Might be a turn-off or rise/fall time problem - could be lots of things, but measurements with the scope should reveal all.

[K1JJ]

Found the runaway problem...

Turns out it was heatsinking.  I had the two IGBTs each sitting on two stacked Sil-Pads.  The junctions were probably on the verge of roasting.  

As a test I tried mounting just ONE IGBT directly on the 1" X 5" copper spreader with white paste, no Sil-Pads and no external heatsink.  Fired up the rig. It ran warm at 500V and warmer at 700V for a minute with no sign of thermal runaway. The direct paste connection did the job.


Reduced the input resistor values and believe the input waveform cleaned up somewhat.  The collector waveform is very good - just a slight sign of slope.

Plan to mount two IXDD414s  right next to each of the two IGBTs to give them a good kick in the pants at 12V.

I see the audio is clean up to about 5 KHZ.  The screen caps need some decreasing and I shud be there with 10KHZ audio limits - the design of the filter.   At this point I think the PDM filter is fine.

T

[K1JJ]

Update:

The original heatsink location was cramped and  had poor air circulation.  I cut a 5" hole underneath  the heatsink and added a muffin fan to blow air thru the fins UP.  WHAT A DIFFERENCE!   With just one IGBT installed, the heatsink gets luke warm running a 175 watt carrier at 750w pep.  

One of the IGBTs shorted after the runaway abuse before, so I ordered another.  But with the new airflow, ONE actually does the job.

I added a 2uf / 5KV cap right next to the damper diodes to dump the spikes.  The short leads are important unless the HV supply filter caps are located very close by.

I need to run some higher power tests yet, but waiting for the new IGBT to arrive.

It appears the rig needs to be loaded very heavily to see perfect audio highs. Loaded lightly and the highs slant like phase distortion.  I think the 1000 ohm PDM filter needs to go higher, like 1500 ohms or so. I shouldn't have to piss beat the 6146s so hard to satisfy the PDM filter.

As a low power driver with the Variac at 600V, it will do 50W / 300w pep  to drive my 4X1 linear amp.  As a stand alone rig at 1400V, I think 200 w carrier / 1000w pep is a reasonable goal without beating it.

Thanks to Frank / GFZ for his many helpful suggestions!

T

Below:

The new air system, heatsink holding the two driver  IX414s and two IGBTs on copper spreaders.

[K1JJ]

BTW, I added flexible copper braid to the tube plate cap connections.  The solid wire was too stiff and pulled a cap from the tube - the glue failed.   Now there is flexibility to thermal expansion and mechanical strain.


Notice the rig is evolving....  I added the legs stand to give the fan below plenty of room for air intake.  I found that  it needed a few inches of clearance or else the air eddy currents ruined the flow.

There was now room to mount the new 2uf / 5KV snubber cap below. (and who knows what else)


It's sounds absolutely great in the off-air monitor and has the normal PDM "shark fins."  It sweeps from 1 HZ to about 10KHZ with flawless audio.  The triangle tests look good at 10, 100, 1000 HZ, etc.    If I can just iron out some minor issues, it will be a FB rig.

T