The AM Forum

In my search for better switching FETs for RF amps I found the specs for the Cree Silicon Carbide series FETs the C2M0280120D in particular, very promising. It exibits a low gate charge and a Ciss of 259 pf all in a 1200V 10 Amp FET. I already have a box full of MOS FETs that I ordered in twos and fours and most turned out to be poor performers at 4 mhz and above.  This one looks like it might work well in a high voltage half bridge.
Anyone tried this series of FETs before I waste any more money.

http://www.cree.com/Power/Landing-pages/MOSFET-products

73s,   VE3ELQ

   I don't know much about the silicon carbide FET's, but look at this one available for $37.50:

http://www.rfmw.com/ProductDetail/DE475102N21A-IXYS-RF/232147/

I attach the data sheet. I can say that two of these at 13 dot 56 Mhz Class (E) make 3 KW CCS with resonated RF drive to the gates. Runs with a B+ of about 185 volts at this power level. You will need to adapt the package or use as is and clamp the bottom down good. It has a 1.8 KW Pd rating.

Jim
Wd5JKO

I have tried these Carbide Mosfets in a experimental class E amp.
The first remarkable issue what I noticed was that the square wave voltage from the (IXDD609) driver was 'more squared' due the lower gate charge capacity of the Mosfet compared with other Mosfets.
The frequency what I used was relativily low, 1,8MHz.
Comparing with other HEXFET Mosfets, I use the IXFH44N50 the types from Cree what I have tested was the C2M0160120D.
In operational mode as a class E amplfier, I have found no difference between the IXFH44N50 and the Cree Mosfet.
Comparing the two types; draws exactly the same current, and gave same power at the output, you can say they give the same results.
The only difference between the two types that the Cree Mosfets require a lower shunt capacity at the drain in a class E amp.
The most remarkable issue is that these Silicon Carbide Mosfets are sensitive for connecting no load at the output filter.
I had four of these Mosfets and I blew all four up.
Then I used the types what I used before the IXFH44N50 and I have no problems anymore.

For high frequency's as 7 and 14MHz these Silicon Carbide Mosfets could be in a advance, comparing to normal HEXFETS.
They can result in faster switching with lower losses and lower gate drive requirements.
But they are more sensitive and less rugged than HEXFETS.

By the way the DE-475 is not an Carbide Mosfet.

73'Martin

Sorry for being late to the party. Kept forgetting to post. I played with some CREE C2M0160120D FETs last month in a Class D application. Definitely had lower input and output capacitance than the 11N90s. Efficiency at 7 MHz was significantly better than I had been able to get, about 90% for 100 watts (2 FETs in P-P). DC power consumption by the IXDD614 drivers went from 16 to 5.5 watts.

The catch was that when modulated, the THD readings were significantly higher on 75 and 160 meters than with the 11N90s. I didn't have time to pursue it then, but will be attempting to use the CREE FETs in another rig that I am slowly working on.

Didn't look for no-load blowups, though!

John, W1TAG

John your info is encouraging.   Puzzling why the THD went up, cant think why. Perhaps on the high mod peaks the FETs develop some parasitic oscillation which messes things up after all they are way faster than the 11n90s. Might be able to see this on a scope with triangle wave modulation at 100%.   
PDORT  Too bad yours blew but the 44n50 are much higher rated FETs current wise and no where near as fast so its really not a fair comparison. 

The C2M0160120D FETs are not that pricey so next time I make a Digikey order I will get 4 and play.

Trying to finish up an H bridge E rig and get it on the air before going too far off track on another project.
73s  VE3ELQ

Just a guess but the THD increase could be due to the SiC device improved switching characteristics.

That's interesting on the THD differences.  How are you modulating (what sort of modulator / technology, etc.) and at what percentage of modulation were you making your measurements?

Regards,  Steve

I have several of the C2M0280120D SiC MOSFETs for which I have measured s parameter data. I started a conventional class C push pull design. Cree offers Spice models. Although they are targeted at switcher applications, the models MAY provide insight into IMD. The FET I selected had a reasonable tradeoff in capacitance and gain for HF applications, 260 pF Ciss, 23 pF Coss, and a Qgs of 5.6 nC. The gm is quite good at nearly 3S. If you desire models for a specific FET, let me know.  

Regards, Alan

Steve,

The modulator setup for the 100 watt PA is a commercial audio power amp (Hafler P3000). It feeds single-ended through blocking caps to the top of a 100 mH power choke (Hammond 195T5) used as a modulation reactor. Since the modulated impedance is around 12 Ohms, no transformer was used. Very nice from 30 Hz to 15 kHz.

Measurements were made at 95% modulation. On 75m, I had 0.7% THD with the 11N90's, and 2.9% with the CREE FETs. The audio instrumentation was an Audio Precision System One, fed with audio from your mod mon pickup. I've put off further experiments until I get farther into the new rig under construction. That'll be a few weeks down the road, as I'm playing with audio processing right now.

John, W1TAG

An interesting article on development of SiC devices.  Looks like this is the future.
http://spectrum.ieee.org/semiconductors/materials/silicon-carbide-smaller-faster-tougher/0

I now have some C2M0280120D FETs and will testing them as RF switching amps shortly.
73s  Nigel

Just a quick report on the C2M0280120D SiC FETs.  First I tried two in a 200 volt half bridge at 4 mhz and got excellent results. The OP was a beautiful 200V peak to peak square wave which cleaned up nicely to sign wave with a series resonant LC OP network into a Cantenna load. Best efficiency obtained was 93%.  I was impressed.  So I resurrected a 6 FET parallel push pull amp built 4 month ago with 11n90 FETs and four IXDD614 drivers, pretty much as QIX Steve designed it only with toroid OP transformers which run cold. It worked well but driver current was 1.2 amps at 12 V at 4mhz best efficiency obtained was 92%. I exchanged the 11n90 FETs with six C2M0280120D FETs and removed two drivers so the 3 FETs on each side had one driver. Drive current was now 205ma with great looking gate waveforms.  At 60 Volts it was making 300 watts into the load without breaking a sweat. OP tuning was set for Class E best efficiency and was not overly critical, the power stayed up there just the current dropped a little. Now here is where it gets really interesting.  My first efficiency measurement gave 98%, cant be I thought, so checked meter cal against my fluke, bang on, measured the load at 46 ohms with 2 meters and used that in the calcs the scope is right on and the waveform was sign wave.  Ok try again at 200 watts, 97%, 300 watts 98%. So if its that good it must be running cold right, and yes it is.  Barely noticeable with a finger on a FET after 20 minutes at 300 watts no fans. My Cantenna on the other hand was smoking hot bubbling oil out the vent so I terminated the heat test.  My VFO driver is set up for 3.6 to 4 mhz so will rig something up to drive them at higher freqs and see what they will do.  Oh yea, when you stick your finger on a FET while operating at 300 watts watch out for the drain buss, that really hurts.  I love these FETs.  Try some yourself.
73s  Nigel

Thats great! So just to summarize: in half bridge you got 93% efficiency, and in class E topology you have 98%, is that right?
Do you use one or two of these transistors in balanced class E topology, or only a single one - or more of them in parallel?

73'Martin

As I said, I studied all the Cree MOSFETs at the time I looked at what they had to offer. I did some simple modeling and s parameter measurements. These were at the top of my list.

Alan  :)

I'm starting a project with some of these devices.  1kW on 80/160 meters.

There is a larger device they have (60A at 25 degrees C).  In theory, 4 of them will replace the 24 MOSFETs in my current 80/160 meter RF amplifier.  Furthermore, the gate charge is QUITE a bit less, so 1 driver will be sufficient for each device, and the rise/fall times will be faster.

The _calculated_ efficiency is between 97 and 98 % efficient (MOSFET efficiency - does not account for output transformer losses and tank circuit losses).  The calculated efficiency with standard MOSFETs is around 90%.

The cost is the interesting calculation.  To build the 24 MOSFET transmitter with FQA11N90s, the devices cost between $72 and $96 for all. 12 drivers at around $5.00 each are required, along with the support circuitry, extra power supplies, bypass caps, etc. etc. etc. for 12 drivers.  Looking at $150.00 to $175.00 for the MOSFETs, Drivers and power supplies for the drivers.

The silicon carbide FETs are around $32.00 each (need 4).  I need 4 drivers at $5.00 each, and ONE power supply for all of the drivers.  So, it's pretty much a break even, or possibly a cost savings.  If the efficiency is greater, the heat sink can be made smaller - further savings.

I'm big on price/performance, so of course I had to look at that factor before starting anything.

Will report back!

The other thing I forgot to mention is this:  if only one device is used per module, the possibility of parasitics is DRASTICALLY reduced because nothing is in parallel.  Won't THAT be nice!

Wow....4 FETs to do 1KW. :) :) :)  Steve are you building one of these now? That will save a lot of fabrication time.

p

Steve that should be one COOL project, (pun intended).

I've had very good results with the IXDD604 DIP 8 drivers, one per FET, which are inexpensive and a little faster than the 614s. They are on a surface mount PCB strip with the chips on the bottom (through the hole) held in contact with the heat sink with three 256 screws.  Can provide photos and PCB layout. If you mount the FETs up on a 3/8 inch alum spacer you can get the drivers up nice and personal to the FETs with very short gate leads. Just an idea.  We look forward to another of your excellent designs.
73s  Nigel

Steve, are your referring to the Cree FETs for the 4// units ?

Alan

These are my amplifiers, the only thing what I have to do is connect them into the transmitter cabinet.(http://k105.PNG)

Yes - from Cree, but the next size up.  I don't remember the number.  The .04 ohm device is the one I'm trying first.  They cost a little over $30.00 each (at least at Mouser).

Have you had any issues with the - 3V gate bias requirement?

Phil - AC0OB

Steve as you are working with high current devices, are you using a heat spreader?

p

If that's successful, then one FET providing a 240 watt carrier can do full legal limit (1500W PEP) when modulated to +150%. That pretty much makes my current 8x11N90 project obsolete.

1 FET should be able to do 250 watts or thereabouts, but don't build a 1 FET transmitter.  Use at least 2 modules (even if each has only 1 FET) for SIGNIFICANT harmonic reduction.

Heat spreaders:  So far, I haven't planned on any heat spreaders.  Hopefully, I don't need them.  I'm only looking at about 10 to 15 watts of TOTAL power dissipation per device.  This is about half R D-S on losses and about half driver power, all of which is dissipated as heat.

We'll see how things work once it's actually built !!   :D

Hello Phil. No problems here with the drive requirement, the gate votlages are square wave of 12Vpp from the IXDD609 driver ic.

regards 73'Martin

Yesterday I was testing my 6 FET amp at 60V with 130V modulation peaks.  It was doing great for about 10 seconds when I heard a SNAP and saw an arc on the side of a FET, the transformer dug in then the fuse blew.  Unlike the 11N90s these FETs have an exposed drain tab on the back requiring insulation pads. I used mica TO-220 pads that I had available but they are about 1mm too narrow to cover the entire drain tab so that left a small gap at each side up from the heat sink by the thickness of the pads.  They are still insulated but at high power the drain voltage pulses are high enough to arc over this gap.  Fortunately nothing was harmed but the fuse and my nerves.  Have ordered some proper sized pads.

So lesson learned.  Fully insulate these things for high voltage.

73s Nigel

Glad nothing was blown up!!  If you're running class E, at 130V, I would expect you'd be generating peaks at the drain of around 400V - maybe even a little higher, depending on the value of your shunt capacitor and the tuning.

Still curious about your output transformer construction.  The output transformers I use do get fairly warm after a while.  If yours don't, I'd like to use that design.

Thanks and Regards,

Steve

Steve when I first built it with 6 11n90 FETs I used 4 rows of 3 (12 total) mix 43 ferrite sleeves for the OP transformer(s).  The primary was copper tubing secondary #12 wire.  It  worked OK but never got better than 87% efficiency and after about 10 minutes at 300W 4Mhz the cores were too hot to touch.  So I tried toroids wound 8 turns bifilar of #12 speaker wire from the HW store on 2 inch cores. Sorry I don't know what mix they are or even remember where I got them.  The 8 turns was a SWAG but they worked so well I never tried anything different.  They literally run cold at 300W and efficiency went up to 92% with 11N90 FETs and about 98% now.  I think Amidon -1 mix or perhaps -2 mix would work well.  A pic attached before it was modified for SiC FETs.
73s  Nigel

I think the key there is try to get by with the lowest perm possible. Get the L by an increase in the physical size and I have seen an approach where the core was epoxied to the heat spreader. The TLT implementations tend to run cooler while the magnetic coupled approaches run warmer as the core loss is an issue.

Alan

I have commenced construction of a 4 device, 1kW RF amplifier.

Heat sink drilled and tapped;  all devices (MOSFETs and drivers) mounted;  output transformers mounted and connected.

To be done:  RF bypass caps, shunt caps, driver bypasses, driver connections to the outside world, transzorbs (on MOSFET drains and driver power common point to eliminate any chance of a catastrophic failure), hookup and testing!

Steve whats it doing?  Did you make some watt things? Did it go bang?  Was on 80m AM with the crowd last night and we are curious. The silence is deafening.

73s  Nigel

Still constructing !  Will definitely report back when I know more.

One thing - the gate drive waveform is very good - one driver per MOSFET.  It's almost square and the driver current is LOW LOW LOW (Under 1 Amp at 15V for 4 devices - that's low!).

You guys are making me salivate about a solid state kW AM radio. The wide bandgap semiconductors using SiC (and Also GaN at higher frequencies) are really moving the technology to where a transistor = a decent tube.

4-5 years back, Microsemi was jumping into SiC static induction FET transistors, through one of their acquisitions in California. They sold me some devices (at work) on an eval board that could produce 1250 watts pulsed at 200 MHz. Single ended!! They were depletion mode so that a negative bias was required on the gate to stop drain current. In other words, with zero bias they would blow up! The drain voltage was 120-150 volts DC.

http://www.compoundsemiconductor.net/article/86791-microsemi-ups-the-power-of-its-sic-transistors.html

These transistors would make a LOT of power, but the bias requirement was fussy, had to have it on negative before the drain (plate!) voltage came on. Similar to GaAsFETs. 

Unfortunately they were arranged so that best topology would be in common gate, which made the power gain < 10 dB. This is typical for Lband radar transistors, being common base for years, and then when FETs came along, common gate. After blowing up the Microsemi part, and looking at commercial LDMOS devices, we decided that we would buy our amplifiers using Freescale parts, and only 8 would be needed for 5.5 kW amplifier, which was our goal. We have been successful with that project, but I have been wondering when and if SiC would find a new base, and it looks like its right here in HF communications.

Ok, I've got the RF amplifier constructed and hooked up.

So far, so good.  I am able to run 45V @ 22.5A with 4 devices on 75 meters.  This is a classic "AM Killowatt" like a KW1 or Johnson Desk KW would run.

Efficiency appears to be amazing.  The RF amplifier is constructed on a 6 x 7.5 inch heat sink and the heat sink is just breaking temperature (I mean barely) with a low speed fan blowing on the fins because the fins are pointed down towards the table (pics and more data tomorrow).

Have not applied modulation as of yet - still making measurements.

The shunt capacitors I am using (Russian made micas - 1500pF at 500V) get too warm for comfort.  Am going to order some multi-layer ceramic caps for this rig.

Will probably modulate tomorrow.  I tend to bring things up rather slowly and make a lot of measurements, particularly when it's new design.

Regards,  Steve

   As a watcher of the class E Ham technology being developed I am getting more and more interested in building one.

That said, an area of concern is keeping the active devices alive in a fault condition where the SOA (safe operating area) is exceeded. I remember most distinctly the DC to AC inverters from the 1970's using paralled darlington transistors. With a massive lead acid battery bank feeding the inverter, a fault often resulted in what I called the "Gatling Gun Syndrome". Imagine a bank of 32 transistors in a TO-3 case where each fuses the PN junctions inside with so much heat that the metal cover over the transistor is blown away and followed by a split second 3 foot flame from all 32 devices! Those things could be used as a weapon.  :P

I fondly remember the IRF-100 FET from IR (International Rectumhorse) in 1979-1980 time frame. These as I recall were 80V 0.2 ohm RDS ON devices. These revolutionized the switchmode power supply, and inverter industry. Long gone are Baker Clamps, and IB2 suck out schemes to minimize the delayed turnoff, and then the ordinarily long fall time of a darlington transistor.

Watching and learning from others on this board, many mega-class E rigs have been built. Many have been reliable, and some not. The use of switching devices with low overall dissipation capability are fine so long as we keep things within the SOA of the devices. March outside the SOA too much or too long, and the ""Gatling Gun Syndrome" rears its ugly head. For example, I once took a pair of 8005 triodes in push pull, and played with drive level, bias voltage, and plate voltage. I could literally get over 1000 watts carrier out from a pair with no color on the plates. Those tubes deep into class C were approaching theoretical efficiency. The problem was, an out of tune plate tank allowed for no longer than 1 second to find resonance. I once took longer than that, and one tube instantly got so hot that the glass was sucked in.  :'(

This thread is about SI carbide FET's, so I am sorry about the somewhat off topic post. The IXYS device I mentioned earlier, although not SIC, they have been 100% reliable as a pair running 3KW out at 13.56 Mhz. These have a Pd rating of 1.8KW each, and in a class E circuit they run close to 90% efficiency on the commercial 21 meter band. Since I mentioned those devices, we had a power supply failure feeding the RF stage, and the power supply went open loop....I could hardly believe my eyes when the wattmeter read 4.8KW!! No failure to those FET's!.

In conclusion, how do we prevent catastrophic SOA violations causing melt downs?  It seems from what others have said, the SIC FET's are fragile in this respect.

Jim
Wd5JKO

Its often been said tubes have "ratings"  transistors have "specs".
Tubes will (usually) withstand short excursions outside their of their ratings. Transistors wont, even for a few milliseconds survive beyond their specs.  So the designs are, and always have been different.  Plenty of design head room is employed along with transorbs and other protection circuitry for over voltage, over current, high reflected power and anything else needed. My idiot proof 10 year old 1KW FET linear does all of this. It uses 4 SD2933 FETs when only two would do the job and still be under rated.  Sometimes its worthwhile to experiment to actually find the limits.  I did this with a pair of $1.50 11N40 400V 10A MOS FETs in a half bridge.  I drove them to destruction at nearly 1KW at 350V. I now know how far they will go and design accordingly.  Although SiC FETs are new to many of us I see no difference from what are the accepted design practices for solid state devices.

I highly recommend you build an E rig.  There is a lot of good info on how to starting with WA1QIX Steve's site. Start small at the 200 to 300 Watt level and go from there as I am doing.  Its a fun journey but it can become obsessive.

73s Nigel

VE6ELQ and all

i have enjoyed reading about these devices and the real-world tests.

thoughts on linear service as well as practical frequency limit?

my next deck will most probably use these.

jim: my class E deck has operated for around 10 years without failure. staying within maximum limits, having good headroom as well as proper overload detection can result in a very reliable transmitter.

thanks,
dan

Dan I thought the same, they may work well in linear amps.  Look at the transconductance curve fig.7 for the C2M0280120D FETs, its nearly a straight line once biased on a little. And with a Tr and Toff of 20ns should do 14 mhz OK, and the smaller ones that I am using are much faster.  We need a champion to work up a design so roll up your sleeves.

73s  Nigel