The AM Forum
November 09, 2024, 04:46:03 AM *
Welcome, Guest. Please login or register.

Login with username, password and session length
 
   Home   Help Calendar Links Staff List Gallery Login Register  
Pages: [1]   Go Down
  Print  
Author Topic: getting on the CREE bandwagon  (Read 19330 times)
0 Members and 1 Guest are viewing this topic.
KQ6F
Member

Offline Offline

Posts: 124


« on: June 08, 2015, 05:20:49 PM »

I built up a 75M RF deck using four C2M160120D MOSFETs running current-mode Class D.  Am not getting superlative efficiency (about 82% @ 350 watts) but nothing is getting warm thanks to the copper baseplate and oversize heatsink.  Picture attached.

Rod
KQ6F


* ClassDRFdeck1.JPG (1016.83 KB, 3648x2736 - viewed 1455 times.)
Logged
W3GMS
Contributing
Member
*
Offline Offline

Posts: 3048



« Reply #1 on: June 08, 2015, 05:37:54 PM »

I built up a 75M RF deck using four C2M160120D MOSFETs running current-mode Class D.  Am not getting superlative efficiency (about 82% @ 350 watts) but nothing is getting warm thanks to the copper baseplate and oversize heatsink.  Picture attached.

Rod
KQ6F


Congrats on getting it up and running and also on the quality of construction.  Superb indeed and I am sure you will find what not making the efficiency what it should be.  So loss somewhere for sure.

Joe-GMS
Logged

Simplicity is the Elegance of Design---W3GMS
n1ps
Contributing
Member
*
Offline Offline

Posts: 338


websdr http://sebagolakesdr.us:8901/


« Reply #2 on: June 08, 2015, 08:12:14 PM »

Very nice.  A couple thoughts:
1. I see you are driving 2 FETs with 1 driver.  Steve indicated it worked best 1:1 in his previous posts.  Not sure if this affects efficiency or not.
2, What are you using to drive the deck?  Watch the pulse widths.  It is typically best to set each phase at no more than 45%. 

Quick thoughts from here....

Nice to have the site back on line.

Peter
Logged
VE3ELQ
Contributing
Member
*
Offline Offline

Posts: 258


« Reply #3 on: June 08, 2015, 08:15:51 PM »

A beautiful job Rod, well done.
Class D has many advantages with only a small efficiency cost with these FETs. But its only a number, if running cool thats what matters.  My 160 80 and 40 meter decks are all now running class D at 350W and using the same FETs as yourself. Im getting between 83 and 85% efficiency so I would say you are in the ball park.
Best 73s  Nigel
Logged

A "Tube Free" zone.
KQ6F
Member

Offline Offline

Posts: 124


« Reply #4 on: June 08, 2015, 08:38:01 PM »

Peter - am driving with 50% input pulses but am getting about 45% duty cycle into the MOSFET gates.  This is because driving two MOSFETs with one driver slows the rise-and-fall times to about 20ns.  An accidental serendipitous event.

Nigel - have improved efficiency by switching from a Chebychev LPF to a Butterworth.  Both have 50-ohm input impedances at my chosen frequency but the Butterworth has a much higher input Z above the passband.  Apparently there are a number of output harmonics.  The Chebyshev was shunting them to ground while the Butterworth goes more-or-less open circuit at these frequencies.  The result is lower supply current and less overall loss.  Efficiency is now about 90%.

Rod
Logged
John K5PRO
Member

Offline Offline

Posts: 1026



« Reply #5 on: June 16, 2015, 02:05:06 AM »

Proper termination of a power device for harmonics is a way to raise efficiency, been done for decades in some broadcast rigs (RCA and Marconi) and now rewarding wireless designers at microwave frequencies. This is how class F is built. There are numerous ways to do it, such as sticking a short circuit for second harmonic right at the active device output, or having a third harmonic trap. Its entirely possible to do waveform modification of the plate (or collector or drain) current or voltage, and force an approximately rectangular shape, where the transitions are fast and no currrent flows same time that voltage is switching off.

One can imagine that your butterworth filter is doing a bit of this as well, as those harmonics are open circuited at the load.
Logged
Opcom
Patrick J. / KD5OEI
Contributing
Member
*
Offline Offline

Posts: 8296



WWW
« Reply #6 on: June 17, 2015, 12:55:30 AM »

That is a beautiful build well done!
No harm in tweaking for efficiency but no shame in 85% when using switching power supply parts either!
Logged

Radio Candelstein
K6IC
Contributing
Member
*
Offline Offline

Posts: 737


« Reply #7 on: June 17, 2015, 03:35:18 PM »

Hi Rod,

That IS beautiful work!   Have fun with it.    Vic
Logged
K4RT
Contributing
Member
*
Offline Offline

Posts: 507



« Reply #8 on: July 04, 2015, 11:23:01 AM »

Very nice work.  Can you post a schematic (or a link)?

Brad
Logged
K1JJ
Contributing
Member
*
Offline Offline

Posts: 8886


"Let's go kayaking, Tommy!" - Yaz


« Reply #9 on: July 04, 2015, 12:09:07 PM »

Rod,

Nice, clean looking rig!

Question, I'm curious:  Looking at your picture above...   I realize the large toroids are low impedance and they are supposedly "self-shielding," but does mounting them directly on a copper sheet and also putting steel washers on top for mounting cause any kind of reduction in Q, inter-coil currents or other disturbances to the transformers?

In my class E rig I mounted them on Plexiglas, but maybe this wasn't needed?

T
Logged

Use an "AM Courtesy Filter" to limit transmit audio bandwidth  +-4.5 KHz, +-6.0 KHz or +-8.0 KHz when needed.  Easily done in DSP.

Wise Words : "I'm as old as I've ever been... and I'm as young as I'll ever be."

There's nothing like an old dog.
KQ6F
Member

Offline Offline

Posts: 124


« Reply #10 on: July 04, 2015, 12:38:49 PM »

Tom - I initially wondered about that myself.  But placing the toroids down against the copper baseplate and clamping them with the steel washers had no discernible effect.

I'd put up a schematic but at this point I'm less bullish on the CREE MOSFETs.  I've had two unexplained blowouts.  In both cases neither voltage or current conditions exceeded ratings.  Huh

Rod
Logged
K4RT
Contributing
Member
*
Offline Offline

Posts: 507



« Reply #11 on: July 04, 2015, 03:47:45 PM »


I'd put up a schematic but [ ... ]


Understood.  It's an interesting project to follow, so keep us posted.
Logged
KF1Z
Member

Offline Offline

Posts: 1796


Are FETs supposed to glow like that?


« Reply #12 on: July 04, 2015, 04:23:49 PM »

Tom - I initially wondered about that myself.  But placing the toroids down against the copper baseplate and clamping them with the steel washers had no discernible effect.

I'd put up a schematic but at this point I'm less bullish on the CREE MOSFETs.  I've had two unexplained blowouts.  In both cases neither voltage or current conditions exceeded ratings.  Huh

Rod


I was wondering about the toroid mounting as well.....

But,
As for the "blow outs."
If you are not monitoring voltage/current with a scope, but only VOM, then you could very
easily miss the spike in voltage.
It takes only a short duration of over voltage, or more likely current to destroy the mosfets.

May be wise to add some TVS protection to gates and drains.

First thing I would be sure of is the waveform to the drivers,
If there is even a little hiccup in the RF drive, then mosfets go away fast.

Logged

Opcom
Patrick J. / KD5OEI
Contributing
Member
*
Offline Offline

Posts: 8296



WWW
« Reply #13 on: July 04, 2015, 04:48:30 PM »

If the amp runs for a while before a failure it could be incremental damage. What I mean is that some structure may be overheating and deforms until eventually the total failure occurs, destroying evidence of the root cause that might otherwise be detectable through some parametric measurement or another as an out of spec measurement.
Logged

Radio Candelstein
KQ6F
Member

Offline Offline

Posts: 124


« Reply #14 on: July 04, 2015, 04:55:45 PM »

Regarding the toroid mounting - I should mention that the toroids are wound with coax and are interconnected to make a 1:4 transmission line transformer.  The transformer is followed by a 1:1 current balun that chokes off current flow on the outside of the coax shields and forces all currents to flow on the inside of the shields.  I believe this arrangement makes the transformer insensitive to how its mounted.  Before I added the 1:1 balun I was getting weird effects due to capacitive coupling from coax shield to ground.

And KF1Z - you're right about an interruption in gate drive causing failures.  My first blowout occurred when I stupidly shut down gate drive while maintaining drain supply.  The energy collapse in the tank circuit caused a large positive spike that took out one of the FETs.  However, I modeled the event in LTSpice that told me the spike was about 500v.  The FETs are rated for 1200v so there shouldn't have been a problem.  (Of course, my model may not be totally accurate although it does amazingly well in terms of output power vs input power.  So go figure....)

BTW - I watch everything like a hawk using two scopes running simultaneously..

Rod
Logged
KF1Z
Member

Offline Offline

Posts: 1796


Are FETs supposed to glow like that?


« Reply #15 on: July 04, 2015, 05:59:56 PM »

That is why I like Steve's over-current protection built into his
PWM modulator.
Along with adding the TVS diodes,

It has saved more than one mosfet :-)


With that set-up, you have to really TRY to waste mosfets.


Logged

VE3ELQ
Contributing
Member
*
Offline Offline

Posts: 258


« Reply #16 on: July 04, 2015, 07:31:48 PM »

Rod sorry to hear you had some FET failures.  
Without a schematic its hard to look for a weak link.  A couple things come to mind however to check for.  You stated that you shut down the gate drive with B+ still applied and had a failure. That should not occur unless your drivers, or one of them, goes to high output if you shut down drive upstream and that drives the FETs or one side FETs on continuously. That will over current a FET very quickly.  Another possibility is gate pulse overlap where both sides are briefly driven into conduction which will result in very high current spikes.  The last area I can think of is a short in the OP circuit.  Looking at the 2 small toroids that appear to be wound with light gauge enameled wire and mounted on the copper heat spreader, hard to see for sure. If that is the low pass filter the RF voltage can get pretty high there and arc through the enamel.  The FETs will run fine with no load but a dead short will kill them quickly from over current.  Just some thoughts.

BTW I have 3 decks running at 70V 350W on 3  bands with 3 different sized SiC FETs and never had a failure, but I did have one arc through its isolation pad and blew a hole through the pad.  I suspect a speck of foreign material caused it.  It placed a dead short across the modulator O/P and the 4 11N40 mod FETs survived.
Good luck with it.

73  Nigel
Logged

A "Tube Free" zone.
KQ6F
Member

Offline Offline

Posts: 124


« Reply #17 on: July 04, 2015, 08:16:54 PM »

Nigel -

Regarding one of my failures - you've hit it spot on.  To generate the two-phase gate drive I'm feeding the drivers from a flip-flop.  I inadvertently shutdown the input to the flip-flop which of course left one side high.  Boom - that blew one FET on that phase.  Up til now I had been thinking it was an overvoltage condition from the collapsing tank circuit field, but now realize it was an over-current situation.

Thanks for pointing that out.

No problem with overlapping gate drive pulses.  They are 45%/ea.

Also no problem with the toroids in the LPF.  You can't see in the photo but they are stood off the baseplate with thick stick-on tape.

Thanks again...

Rod
Logged
N2DTS
Contributing
Member
*
Offline Offline

Posts: 2303


« Reply #18 on: July 04, 2015, 09:19:34 PM »

Really nice looking work!
Logged
steve_qix
Contributing
Member
*
Offline Offline

Posts: 2597


Bap!


WWW
« Reply #19 on: July 05, 2015, 07:47:12 AM »

I have run into what I believe is a brick wall with the SiC MOSFETs.

Everything works fine up to about 100V total supply voltage.  This is on 40 meters.  When I reach about 120V, I will experience failures after a while.  This is repeatable with two different layouts, and with different types of SiC MOSFETs.  At 100V, the RF voltage will rise to about 350V, which is normal for class E.

There are *NO* excessive high voltage spikes on the drains.  The drains are protected with 540V transzorbs anyway, and the transzorbs do not fail.  The drive is 100% consistent.  This is the same drive I use with all of my standard MOSFET class E rigs.

Now, this is another noticeable anomaly:  As I increase the voltage, the positive peaks start to roll off.  This is also consistent in the SiC MOSFETs.  John W1TAG was seeing something similar.  This happens on both 75 meters and 40 meters.

I do not have a good working theory as to the cause of these anomalies, but I have done a lot of experiments and made a lot of measurements.

Anyway, interesting..
Logged

High Power, Broadcast Audio and Low Cost?  Check out the class E web site at: http://www.classeradio.org
WD5JKO
Member

Offline Offline

Posts: 2001


WD5JKO


« Reply #20 on: July 05, 2015, 09:34:47 AM »

I have run into what I believe is a brick wall with the SiC MOSFETs.

Everything works fine up to about 100V total supply voltage.  This is on 40 meters.  When I reach about 120V, I will experience failures after a while.  This is repeatable with two different layouts, and with different types of SiC MOSFETs.  At 100V, the RF voltage will rise to about 350V, which is normal for class E.

There are *NO* excessive high voltage spikes on the drains.  The drains are protected with 540V transzorbs anyway, and the transzorbs do not fail.  The drive is 100% consistent.  This is the same drive I use with all of my standard MOSFET class E rigs.

Now, this is another noticeable anomaly:  As I increase the voltage, the positive peaks start to roll off.  This is also consistent in the SiC MOSFETs.  John W1TAG was seeing something similar.  This happens on both 75 meters and 40 meters.

I do not have a good working theory as to the cause of these anomalies, but I have done a lot of experiments and made a lot of measurements.

Anyway, interesting..

  Steve,

    A couple of things to consider. Maybe designing for a lower voltage and higher current final would work better? Also, the drain side transorbs will have a varying capacitance as the drain voltage rises above ground. Maybe this capacitance change, or changing 'Q' might be part of the issue. I was wondering if instead of a 540V transorb, to use an RF AC bypassed zener of similar voltage, and then couple that to the FET Drains with a SIC diode. Maybe use a .01 uf across the zener. If my thinking is right, this change will still protect the FET's, and at the same time, the capacitive loading Drain to ground will be reduced from what the transorb by itself would have provided.

   At work with some of the 3KW 13.56 Mhz amplifiers, we have had some failures of the big IXYS (DE-475 package) power FET's. They still pass the bench test with a DVM and a 9V battery, but it appears that the effective series gate resistance goes way up, like from a few ohms, to maybe 10K. With your Cree SIC Fet's are they failing with a short drain to source, or perhaps something to do with the gate?

Jim
Wd5JKO
Logged
VE3ELQ
Contributing
Member
*
Offline Offline

Posts: 258


« Reply #21 on: July 05, 2015, 10:26:37 AM »

I have run into what I believe is a brick wall with the SiC MOSFETs.


Now, this is another noticeable anomaly:  As I increase the voltage, the positive peaks start to roll off.  This is also consistent in the SiC MOSFETs.  John W1TAG was seeing something similar.  This happens on both 75 meters and 40 meters.

I do not have a good working theory as to the cause of these anomalies, but I have done a lot of experiments and made a lot of measurements.

Anyway, interesting..

Steve thats disappointing news.
If it were anyone but you I would be all over this with ideas.
But for fun lets try some logical analysis anyway.

You and others have been doing this for a while with 11N90 FETs and ferrite sleeve transformers.  That works.
So whats different.  The FETs.   You mentioned earlier that the cores get warm and the drain bypass caps get hot. That, and your above comments about lack of positive peaks and failures at high power are pretty good indicators of transformer core saturation. So why does it work with 11N90 FETs and not the Sic FETS???  Look at the difference in Rds On, 11N90 is 1.1 ohm, the SiC FETs you are using are .065 ohms, a huge difference, so possibly the 11N90s are acting as a natural current limiter resulting in soft saturation or none at all. As you know when the primary saturates it just becomes a short piece of #10 wire and current spikes can be huge exceeding the FET limits.

As we have discussed before a large core toroid transformer is probably a better option and may be the solution if you haven't already tried it. They certainly work for me, a T225A-2 core #12 wire on 40M at 350W carrier runs stone cold, and I mean that, no detectable heat at all. At the power levels you are looking for you may need 2 of them stacked or better yet a T300A-2 core, or two modules in series or whatever it takes.

Also these FETs are different animals.  They run better at high voltage low current with their 1200V spec.  Closer to a tube than a FET. I realize this requires some PS and Modulator re-design but worth considering.

Please hang in there, if anyone can solve this its YOU.

73s  Nigel
Logged

A "Tube Free" zone.
K1JJ
Contributing
Member
*
Offline Offline

Posts: 8886


"Let's go kayaking, Tommy!" - Yaz


« Reply #22 on: July 05, 2015, 01:26:04 PM »

http://amfone.net/Amforum/index.php?topic=35596.50

This is not the same application and device, but may shed some ideas about the failing SiC MOSFETs.

In this thread I tried a pair of IGBT FETS that were rated at 2KV as the pulse width modulator to modulate the four 4D32 RF tubes.  The IGBTs were about  $45 apiece, so I was careful.  The waveform looked great on the scope at lower power.

They worked fine until I got above 600 volts and then they went into a fast internal thermal runaway.  I tried many configs and precautions with no luck.  Bigger heatsinks and fans did not help. The damage was occurring on the chip itself, not able to dump the heat fast enough.  I went through about eight of them before I gave up. $$

I later went back to 4D32 tube modulators, ran the voltage up to 2200 volts and never had a failure.  

It's interesting that the cheaper 11N90s are so bulletproof compared to some of these newer devices.

T
Logged

Use an "AM Courtesy Filter" to limit transmit audio bandwidth  +-4.5 KHz, +-6.0 KHz or +-8.0 KHz when needed.  Easily done in DSP.

Wise Words : "I'm as old as I've ever been... and I'm as young as I'll ever be."

There's nothing like an old dog.
Pages: [1]   Go Up
  Print  
 
Jump to:  

AMfone - Dedicated to Amplitude Modulation on the Amateur Radio Bands
 AMfone © 2001-2015
Powered by SMF 1.1.21 | SMF © 2015, Simple Machines
Page created in 0.091 seconds with 19 queries.