SiC verses GaN class D on 80M

[VE3ELQ]

My latest class "D" PW Mod AM TX project presented an ideal opportunity to compare build and performance of SiC and GaN FET RF decks. The TX is a 80 Meter 350 Watt, 1.6 KW PEP, dual phase boost converter design initially built and tested using 6 C3M0280090 900V 11.5A SiC FETs with 6 NCP81074A drivers powered at 9V. A second interchangeable RF deck was also constructed using 4 TPH3208PS 650V 20A GaN FETs and 4 NCP81074B drivers powered at 6V. They are nearly identical otherwise using the same transformer design 3 pole low pass output filters with only slightly different drain cap values. The GaN deck is about 50% smaller due to the 2 less FETs and lower footprint of the TO220 package. It was also easier to build with the common source FETs bolted directly to a sheet copper ground plane/heat spreader on the top 2/3 of the heat sink. An external PS provides 150V regulated into the half bridge class "D" PW modulator.

On air performance of both decks was virtually identical at 350W. Many QSOs provided good signal and excellent audio reports. Both decks run just slightly warm after about 15 min at carrier so efficiency seems similarly very high. I still can't measure it exactly but temperature rise of the heat sink is a pretty good indicator. Then following about 6 contacts with the GaN deck in the middle of a nice chat with K1KBW on 3873 there was the dreaded flash/bang as 2 FETs blew apart. They were replaced and following a couple more QSOs 2 more let go. The exact cause of failure is hard to determine but most probably was due to over voltage on mod peaks which produces about 600V on the drains just below their rating. Interestingly my 40M GaN rig also produces 600V mod spikes with no issues for about 7 months now. But it uses the 600/800 volt GaNs which seem to be more robust. Not sure...

So in conclusion for 80 Meters performance wise 6 C3M0280090D SiC FETs are equal to 4 of the GaNs and are probably the better choice. These are the best FETs I have found and I tried many. Cost is about 1/2 including the extra drivers but construction is more difficult and size is increased. The GaNs would probably survive OK if operated at a lower voltage and higher drain current. They are the clear winners however on 40M due to their higher switching speed.

The small deck is now being re-built for 4 C2M0280120D 1200V 10A SiC FETs which I have used before and found to be excellent performers, about equal to the C3M0280090D but with higher gate capacitance. It should be able to run at even higher voltage to achieve 300 to 350W. That means another power supply,   oh well its still fun stuff.

Some pics of the GaN deck construction attached.

73s Nigel

[ka1tdq]

I marvel at your construction techniques.  I like how the transformer can unscrew from the heat sink so that you can get to the FETs.  It makes it really convenient to replace them.

When I built my 8-FET rig, as a passing thought I made the heat sink assembly removable.  It saved my butt since I've had to do a FET swap out twice during my learning phase.  

Anyway, great info here!  If I have difficulty with my 40 meter rig working that I'm building now, I can try GaN's.

Jon

[vk3alk]

Hi Nigel and Jon....

Congratulations thats heaps of work ... and a nice power level as well.
Just wondering on your output transformer.....have you tried twisting the wires together ?...
On the photo uploaded if you look at my transformer and see the twists there..was able to achieve an extra 2% in efficiency ... might not seem much but it all adds up to nice cool FETs... !!!

The FETs ... is the drain lead the one on the right.... a little different .....I prefer it really...

If possible could you post the waveform after the transformer before the filter.....

Have also uploaded my waveform pre filter ...... TX is a H Bridge....actually are building 2 units and applying to a combiner .......

Your regulations must be different then ours in Australia.....harmonics need to be below 45db here so thats why I use a 7 pole filter to be on the safe side..


Wayne

[n1ps]

Nice work!  TNX for posting.  On the GAN failures, I think the 600V rating may be the issue.  Do you put out the same RF power on 40?  I would think it would be a little less?  I wonder if it may help to use some 540V transorbs across the drains?  Steve does that with his designs.  A good indication is if they get warm on voice peaks.   Can a 900V GAN fet be had at abt 10A?

I like the common source mounting too.

Anyway some useless thoughts from here.

Love to read your posts! 

Peter

[VE3ELQ]

The 6 FET deck has been re-installed in the TX and is working FB.

The small deck has now been retrofitted with four 1200V 10A SiC FETs and is performing very nicely, no explosions.
With the same transformer ratio it produced exactly the same power as the GaNs did, as it should, but ran just slightly warmer.
A 1/2 turn was removed from each secondary to unload the deck a little which tamed it down to 310W from 350 with a corresponding reduction in drain current since its now only good for 20A vice 40. It modulates well to about 1.5KW PEP.

Pete thanks for the comments, not a big fan of transorbs, these FETs are good to 1200V so lots of voltage headroom, should be OK.

As requested a scope trace attached with it dialed up to 400W carrier showing the transformer output (yellow) and across the load after low pass filtering (blue). Harmonics, especially the 3rd, are within regs even before the external L tuner which adds another 2 poles of filtering. Its a keeper so will build yet another TX for it as small as I can.  

73s  Nigel

[vk3alk]

Thats a nice waveform after the transformer....

Great..



Wayne

[W1DAN]

Nigel and John:

Excellent development! Good job. I have been enjoying reading about your pioneering developments.

73,
Dan
W1DAN

[VE3ELQ]

The past GaN FET failures were puzzling. Peak drain pulse voltage never exceeded 600V which is 50V below a conservative spec so over voltage was unlikely the cause. Some thought and a bunch of scope measurements uncovered the cause as drain pulse overlap combined with large drain cap values.

In a boost converter configuration, as this design is, an inductor is magnetically charged  from B+ through a fast FET switch to ground.  When the FET switch opens the collapsing field produces a high voltage pulse which is shaped with a capacitor to produce an approximate ½ sign waveform. Very low ESR silver mica caps are used  capable of delivering very high current. If the drain pulses exceed ½ wavelength then the capacitor will still retain some charge and the inductor will also still be pumping current when the FET turns on for the next cycle.  This combination results is a very high current pulse through the FET which according to a bunch of math can exceed 100 amps at full modulation. Not a good thing for any FET but especially bad with GaNs with their very fast turn on and low RDS where capacitor discharge will be very fast at very high current. If  modulation is driven into positive peak clipping then full B+, which in this case is 150V, can last for several milliseconds or thousands of RF cycles where the FET gets “banged” with this big current spike every RF cycle. Well they don't like that much and we now know the result, they themselves go BANG. 

The solution was two fold. Tune (adjust) the caps to match the transformer primary inductance to keep the drain pulses just below ½ cycle.  And then use the smallest possible drain capacitance value by increasing the primary inductance to achieve the desired waveform.  So a new transformer was made with increased primary turns on a T200-8 core giving about 4 times the inductance. Drain caps were reduced from 660pf  to 270pf tuned for about ½ wavelength, and the low pass output filter was changed to a 4 pole to handle the slightly higher harmonic content.  The test deck set for a 350W carrier point was then torture tested with sustained severe over modulation giving a PEP of 1.7KW badly flat topping.  Not only did it pass the test but the efficiency actually increased as it ran amazingly cool.  A new 4 GaN FET TX has now been built with modulator and RF deck conduction cooling into the back plate with no traditional finned heat sinks. Its currently undergoing “on air” testing so if you hear me on 80M a signal and audio report would be most appreciated. If it doesn't blow up will post some pics of the new TX in a few days in a new thread.

73s  Nigel

[WD5JKO]

  Nigel,

   Thank You for pioneering the way using newer devices. I patiently await your followup.

I also look forward to a 40m version sometime in the future!

Jim
Wd5JKO