The AM Forum
December 11, 2024, 09:07:45 PM *
Welcome, Guest. Please login or register.

Login with username, password and session length
 
   Home   Help Calendar Links Staff List Gallery Login Register  
Pages: [1] 2   Go Down
  Print  
Author Topic: SiC FETs latest?  (Read 41248 times)
0 Members and 1 Guest are viewing this topic.
n1ps
Contributing
Member
*
Offline Offline

Posts: 338


websdr http://sebagolakesdr.us:8901/


« on: March 20, 2016, 08:59:36 PM »

So where are we in testing these FETs to implement into class E/D RF decks?  I believe there has been testing for 160/80 and 40. There has been some experimenting with several different FETs I think.  Steve QIX posted some results last year that looked very encouraging on a specific FET running about 250W.  I think some people went ahead and started fabricating 2 or 4 FET decks using those FETs. 

Inquiring minds want to know the latest Smiley Smiley

TNX

Peter 
Logged
steve_qix
Contributing
Member
*
Offline Offline

Posts: 2598


Bap!


WWW
« Reply #1 on: March 23, 2016, 11:29:49 PM »

They work pretty well, but the peak current rating is an issue.

The documentation for the SiC FETs shows a high peak current rating, but I do not believe the reality meets the theory.  Or the gate internal resistance is causing the devices, during transitions, to start to come out of saturation.

Either way, the result is the same.  You have to run lower current than the theoretical norm or the devices will be damaged.

Otherwise, they are excellent.  Efficiency is better than standard MOSFETs, and they are significantly easier to drive.

Still experimenting :-)

Regards,  Steve
Logged

High Power, Broadcast Audio and Low Cost?  Check out the class E web site at: http://www.classeradio.org
n1ps
Contributing
Member
*
Offline Offline

Posts: 338


websdr http://sebagolakesdr.us:8901/


« Reply #2 on: March 24, 2016, 08:46:34 PM »

OK great.  We'll talk more on air abt it.  Maybe Saturday if you come north to the fester.

TNX
p
Logged
WD5JKO
Member

Offline Offline

Posts: 2002


WD5JKO


« Reply #3 on: March 25, 2016, 12:32:54 PM »

the gate internal resistance is causing the devices, during transitions, to start to come out of saturation.


   Although not specific to SIC Fet's, I use an IXYS DE-475 series FET in repairs of commercial Class E amplifiers. The manufacturer had a packaging defect leading to a sudden rise in the gate internal resistance, Rg. They have recently modified their manufacturing process, and things are now much improved. Rg would shift just from soldering the device. Later at some point with gate current, the gate connection would go high impedance, and yet still function at DC.

   I think the high gate current charge / discharge pulses is leading to a degradation in Rg, something Steve suspects of the SIC devices.

   See failure analysis photo below.

Jim
Wd5JKO



* DE-475 Fet Failure.jpg (81.61 KB, 933x799 - viewed 1283 times.)
Logged
VE3ELQ
Contributing
Member
*
Offline Offline

Posts: 258


« Reply #4 on: September 14, 2016, 09:31:54 AM »

Getting back to AM stuff after a year building linear amps and was looking at the specs for the Cree C3M0280090J in the 7 lead D2pak compared to the 1200V 10A FETs that I have had good results with.  900V 11A, Fast turn on/off, lower gate input C and charge, but offset by a higher Gate R, and lower cost. They looks like good candidates for 7.4Mhz and possibly higher. Anyone try these yet??
Have 6 on my next order list so will experiment and see what they will do.


Nigel
Logged

A "Tube Free" zone.
M0VRF
Member

Offline Offline

Posts: 177


« Reply #5 on: September 14, 2016, 12:12:25 PM »

Hi Nigel, Yes do keep us informed.

A PCB for 4 devices with 2 drivers is still on my 'To-Do' list.

I'm sticking to the T0247 package size but should be fine for FET testing.

Cheers from the UK.

Stretchy.
Logged
KA6MOK
Member

Offline Offline

Posts: 22


« Reply #6 on: September 15, 2016, 09:41:51 PM »

A lurker comes out of the woodwork...

Hello to the group...  Been reading around on here for a while,  and went over to the Cree site a couple weeks ago and had seen the new C3M line of FETs.  Was going to finally register and post, but Nigel beat me to it.  See what procrastination will do to you?  Roll Eyes

They do look good....  and that alternate package with the lower inductance is interesting... A little harder to deal with, but I'm wondering how much difference that might make for a try on the higher bands...

Jon
Logged
steve_qix
Contributing
Member
*
Offline Offline

Posts: 2598


Bap!


WWW
« Reply #7 on: September 17, 2016, 08:35:11 PM »

Hi !

I've used the C3M0065090D on 40 meters with reasonably good success.

And I still have my 4 device 1kw carrier RF amplifier for 75 meters using some of the 1200V devices.

The thing I see with these devices is that they are rather fragile.  If you look at the SOA curves, essentially ANY overload or current peak is going to cause damage due to junction temperature.

So, in order to have enough headroom, the devices have to be run in such as way as the repetitive peak current is well below the DC rating of the device.  Any operation at all in the "peak" rating area will be trouble.

If you look at the SOA curve for, say, an 11A SiC device and compare that to something like an FQA11N90 (an 11 A standard MOSFET), the curve for the FQA is _significantly_ better, indicating a much more robust device with respect to overload.

This does not mean in any way that the SiC devices are unsuitable.  It's just that any designs have to account for the vulnerability of the devices with respect to short overloads or current spikes.  These occur all the time in transmitters due to numerous factors.

I am working on a 6 SiC Fet RF amplifier which should be conservative enough (I think) to be able to replace a 24 standard MOSFET rf amplifier. However, I will be doing a lot of testing of the RF amp in the area of overloads (will force overloads) to ensure that the reliability is suitable.

The standard MOSFET designs are, at this point, rock stable (it's been literally a decade since I've experienced a MOSFET failure in any RF amplifier I've built, so we don't want to regress in this area.

Keep experimenting  Smiley Wink

Regards,  Steve
Logged

High Power, Broadcast Audio and Low Cost?  Check out the class E web site at: http://www.classeradio.org
KD6VXI
Contributing
Member
*
Offline Offline

Posts: 2692


Making AM GREAT Again!


« Reply #8 on: September 17, 2016, 09:32:56 PM »

Steve,

Six devices as in 3 parallel x 3?

Wow.


--Shane
KD6VXI
Logged
steve_qix
Contributing
Member
*
Offline Offline

Posts: 2598


Bap!


WWW
« Reply #9 on: September 17, 2016, 10:43:06 PM »

Steve,

Six devices as in 3 parallel x 3?

Wow.


--Shane
KD6VXI

In this case, 6 modules of 1.  My current kw SiC FET transmitter is 4 modules of 1.

Regards,

Steve
Logged

High Power, Broadcast Audio and Low Cost?  Check out the class E web site at: http://www.classeradio.org
KD6VXI
Contributing
Member
*
Offline Offline

Posts: 2692


Making AM GREAT Again!


« Reply #10 on: September 18, 2016, 11:52:36 AM »

I'm trying to wrap my head around it,  with the class e topology.

If I have it right,  it's 6 seperate ports off the splitter / combiner.   And the class E network goes after the combiner?   Or 6 seperate class e decks,  run into a 6 port and from same?

My foray thus far has been single fet decks,  I want to go further.

--Shane
KD6VXI
Logged
steve_qix
Contributing
Member
*
Offline Offline

Posts: 2598


Bap!


WWW
« Reply #11 on: September 18, 2016, 05:05:39 PM »

The layout is super-simple.

This is singled ended push pull.  3 modules in each phase (total of 6 modules).  In this case, each module has one SiC FET, but they could have more smaller FETs.  For instance, the 24 FET design has 4 modules of 6 FQA11N90 MOSFETs in parallel.

The secondaries are in series.  I tend to put the secondaries of the RF transformers for each phase in series, minus to plus (like batteries), creating two "banks".  Then combine the "banks" minus to minus or plus to plus (because the banks are out of phase with each other).  This will give you a pseudo-sinusoidal waveform, which is presented to the class E network.  One side of the series string goes to the network and the other side goes to the RF ground.

Due to this design, the harmonic output is EXTREMELY low, and the efficiency is better than a single module (or one phase with multiple modules).

There is a practical limit to how many FETs can be put in parallel.  First, there is the issue of long drain busses.  This will cause problems due to stray inductance,  6 FETs in parallel is practical and works very well.  More than that is trouble.  4 FETs in parallel is great.

2nd is the amount of current a single module can consume before parts such as RF bypass capacitors, shunt capacitors, core material in the output transformer and the like become large, expensive, on impossible to find.  7.5A DC at carrier per module is about the most I would ever run.

There is a lot of documentation about subjects such as these on the class E web site.

Sounds like a new project in the wings!
Logged

High Power, Broadcast Audio and Low Cost?  Check out the class E web site at: http://www.classeradio.org
KD6VXI
Contributing
Member
*
Offline Offline

Posts: 2692


Making AM GREAT Again!


« Reply #12 on: September 18, 2016, 09:14:10 PM »

I want to build something for 40.

The antenna farm at the moment is a 40 inv v.   I can load it on 75, but it's good for local only. On 40, I've talked with Brandon,  IIA  and farther.....   So I know it's my best bet until I get my 80 foot tower up.

Funny,  it's 40 that scares me,  and I've got a working 10 meter deck :-)

--Shane
KD6VXI
Logged
VE3ELQ
Contributing
Member
*
Offline Offline

Posts: 258


« Reply #13 on: October 08, 2016, 07:13:04 AM »

Cree/Wolfspeed add another one to their family. Looks like a good candidate for the high power guys. The IXDN630 looks like a good driver for it.
http://media.digikey.com/pdf/Data%20Sheets/CREE%20Power/C3M0065100K.pdf
http://www.ixysic.com/home/pdfs.nsf/0/6AC7CC624179BDA185257873005D6BE4/$file/IXD_630.pdf

More info here:  http://www.semiconductor-today.com/news_items/2016/oct/wolfspeed_061016.shtml

73s  Nigel
Logged

A "Tube Free" zone.
VE3ELQ
Contributing
Member
*
Offline Offline

Posts: 258


« Reply #14 on: October 08, 2016, 05:23:24 PM »

I just tested the new Cree C3M0280090D 900V 11.5A TO247 pack FETs in my old 4 FET test deck.  At 7.4 Mhz they function beautifully with IXDN614 drivers at 15V, one per side, drive current is 310ma. It took 100PF drain caps to get the pulses just right.  Input at carrier was 70V at 4.2A for a power of 294W, output was 251W as best I can measure it, this in class D with a 3 pole butterworth output filter.  Full modulation to 145V on peaks was no problem with RF output peaks past 1KW.  The deck barely gets warm. Class E was not tried but should result in even higher efficiency.

I was hoping for 20 meters so tried a different output transformer with lower primary inductance. Unfortunately the drivers are not up to the task as the gate waveforms looked closer to a sign than a square wave and the FETs responded accordingly with poor performance.  With 250W input the output was 130W and the deck got pretty hot within a few minutes.  But it did make some RF at 14.4 Mhz.  From this data though I think these FETs could work on 20 meters with the right drivers.  So the search continues as 20 meters still alludes me.

These are now my "go to" favorite FETs for this application. Am starting build on a 40 meter 6 FET 350W deck to match 2 other 6 FET decks already built for 80 and 160 using the 1200V 10A FETS.  A good buy at about $4. This stuff just keeps getting better.  Next is GaN if the price ever comes down.

73s  Nigel
Logged

A "Tube Free" zone.
WD5JKO
Member

Offline Offline

Posts: 2002


WD5JKO


« Reply #15 on: October 08, 2016, 05:42:52 PM »


Nigel;

Glad to see you back after your LDMOS experiments.

Consider an alternative driver for class D, E gate drive:

http://www.ixyscolorado.com/index.php/ixys-rf/mosfets-drivers-diodes-power-modules/drivers-sic-diodes/item/ixrfd630

The rise and fall times into a 1000pf load is under 4 ns......
The rise and fall times into a 4000pf load is under 7 ns......

That is screaming guys!

Here is a pasted email from an IXYS application engineer:

"I would like to suggest that you look at our IXRFD630 driver on our driver page that is well suited to high frequency operation and 13.56MHz. The 630 replaced the 420. The IXD_614 from the IXYS IC division and all of the other drivers there are primarily limited by power dissipation Pgate=Vcc*Qgate*Freq. Their driver die are small in comparison to ours. The ones in the small SOIC packages are the ones really limited to 2MHz as they use epoxy attachment for the die to substrate, the ones in the TO-220 package I believe use solder die attachment with larger surface area and so you may be able to operate them at higher frequencies. But our driver is in a bigger package (more surface area) with a bigger die using solder attachment for the die so we can dissipate a lot more power than all of the others."

Jim
Wd5JKO
Logged
VE3ELQ
Contributing
Member
*
Offline Offline

Posts: 258


« Reply #16 on: October 08, 2016, 09:01:06 PM »

Jim thanks for the info on those drivers.  Certainly looks like they would do the job for the higher bands.  I was aware of them but a $30 price tag up here kept me away.  May re-visit my decision in the future.

I included 2 of the newer IXDD630CI drivers with my FET order.  They have twice the punch at nearly twice the speed of the IXDD614s.  These were installed initially on my test deck.  The scope waveforms looked super good then I connected them to the FET gates and with no high voltage even connected powered them up with 15V. They both failed within about 2 minutes.  I have no explanation as to why.  I replaced them with IXDD614s and had no problems.  Go figure. 

73s  Nigel
Logged

A "Tube Free" zone.
KA6MOK
Member

Offline Offline

Posts: 22


« Reply #17 on: October 10, 2016, 05:22:06 AM »

Hmmm... that sounds a bit dissapointing about the driver.  I was also looking at the IXxD630 parts,  they looked much better than IXxD614 parts, and for nearly the same price...  as opposed to the really fast and expensive IXRF630 parts.

I even got my DigiKey order in this Friday....    Grin   along with a few familiar FET's both Silicon and Silicon Carbide...  got a few of both driver types. Huh Guess it was a good idea to get both!
Not sure when I'll have the spare time to start playing with these, and need some other parts to get started,  but we'll see.  Anyone have some vendor reccomends for fixed and variable caps, for instance?

On the driver failures, I think it's worth gettiing a hold of the Cree application engineers.  Doesn't sound right that any reasonable number of these kinds of FET's should cause them to blow.  Maybe some defect problem?  They'd probably be very interested to know the details to figure things out, and get some good ones to you, I'd imagine.

Otherwise,  awesome results Nigel!  Very encouraging...   And, once the IXDD630 driver fail mystery is solved, they should do much better on 20....
Logged
VE3ELQ
Contributing
Member
*
Offline Offline

Posts: 258


« Reply #18 on: October 10, 2016, 08:43:57 AM »


On the driver failures, I think it's worth gettiing a hold of the Cree application engineers.  Doesn't sound right that any reasonable number of these kinds of FET's should cause them to blow.  Maybe some defect problem?  They'd probably be very interested to know the details to figure things out, and get some good ones to you, I'd imagine.

Jonathan, I dont think its a Cree FET problem, they are working FB with 614 drivers, or perhaps you meant IXYS, a driver quality issue. My thoughts are that with a 30A punch these drivers need some serious bypassing.  No way is a single puny 7815 regulator going to supply that kind of current, and that's what I was using at the time with .47mfd bypass cap at the driver Vcc pin. Not smart in hindsight. I have now installed seperate 7815 regulators up close to the drivers and a 10mfd 25V cap with a .1mfd and a .001 cap all in parallel on the Vcc pin of both drivers, nice and close.  This works great with the 614s so now need to get some more 630CI drivers and try again. Cant think of what else to try.

Good luck with your experiments, please keep us up to date.

73s  Nigel

EDIT:
Got to thinking given the higher speed of these 630CI drivers its possible that they went into parasitic oscillation which quickly destroyed them.  I was looking at the drive current meter not the scope when I powered them up so could have missed it. I never did see gate drive so it could have happened right away. If this was the case then they will take some careful mechanical design to keep things very short with low inductance possibly limiting these to 1 driver per FET right up close.
Logged

A "Tube Free" zone.
KQ6F
Member

Offline Offline

Posts: 124


« Reply #19 on: October 11, 2016, 03:02:27 PM »

There are some new SiC devices out from ROHM.  At first blush, this one looks pretty good:
http://http://www.rohm.com/web/global/datasheet/SCT3160KL.  But wait...it's internal gate resistance is 18 ohms.  That sounds to me like its gate structure is silicon as opposed to metal and as such is vulnerable to overheating from RF gate drive.  However, I'm not familiar enough with how much gate resistance is allowable...Steve QIX can probably comment on that.  To get a feel I looked up the venerable FQA11N90 but Fairchild doesn't publish that parameter..
Logged
WD5JKO
Member

Offline Offline

Posts: 2002


WD5JKO


« Reply #20 on: October 11, 2016, 10:37:16 PM »



The company I work for will be installing several ion implant machines for the SIC industry. This whole industry is growing leaps and bounds, and I expect big changes to come over the next year.

Last week I was at one of those companies trying to address why some of the wafer sensors work with silicon wafers but not with silicon carbide wafers. If you hold a SIC wafer up into the light, a dim yellow light comes through. The wafer sensors use infra rad light, and that goes right through the wafers. I had several things to try in my bag of tricks which includes red, green, and blue emitting LED's, and the hope of just throttling back the current in the IR led. I got lucky finding that reducing the LED current by 2/3 did the trick.

It will be fun watching this niche technology blossom.

Jim
Wd5JKO



Logged
steve_qix
Contributing
Member
*
Offline Offline

Posts: 2598


Bap!


WWW
« Reply #21 on: October 12, 2016, 12:03:26 AM »

There are some new SiC devices out from ROHM.  At first blush, this one looks pretty good:
http://http://www.rohm.com/web/global/datasheet/SCT3160KL.  But wait...it's internal gate resistance is 18 ohms.  That sounds to me like its gate structure is silicon as opposed to metal and as such is vulnerable to overheating from RF gate drive.  However, I'm not familiar enough with how much gate resistance is allowable...Steve QIX can probably comment on that.  To get a feel I looked up the venerable FQA11N90 but Fairchild doesn't publish that parameter..

The 18 ohms into the 390 (more or less) pF of gate C is typical of what I've seen for SiC devices of similar size  The cree devices of a similar die size (current / voltage) are in this range as well.

For this device,  I think the capacitive reactance is going to be around 100 ohms (more or less) at 4 mHz.  The 18 ohms series resistance should be taken into consideration, but will not stop the device from working, at least on 75 meters.
Logged

High Power, Broadcast Audio and Low Cost?  Check out the class E web site at: http://www.classeradio.org
KA6MOK
Member

Offline Offline

Posts: 22


« Reply #22 on: October 12, 2016, 03:36:32 AM »

Jonathan, I dont think its a Cree FET problem, they are working FB with 614 drivers, or perhaps you meant IXYS, a driver quality issue. My thoughts are that with a 30A punch these drivers need some serious bypassing.  

No, not the FET,  I was thinking the drivers might have had an issue.  They should be pretty robust, and handle 10X the load your FET's amount to.   Now, what they do without a lot of bypassing,  that is a good question.  I even called to find some tech support folks to ask about that and what might cause failures,  haven't heard directly back yet.

For the faster IXRF mega part,  there is some detailed layout and bypassing suggestions for it to perform fully.  There's an application note or two out there on running those,  might be an idea to read and follow. Probably not quite as much is needed for the IXDD's,  but you want them every opportunity to A. be stable, and B. be able to crank gates up and down as fast as they can.  

Your definitely on the right track adding the different sized and types of caps...  that's pretty much the approach in the app note, just more of different values, and really low ESR ceramics...

Jon
Logged
KA6MOK
Member

Offline Offline

Posts: 22


« Reply #23 on: October 12, 2016, 03:47:22 AM »

There are some new SiC devices out from ROHM.  At first blush, this one looks pretty good:
http://http://www.rohm.com/web/global/datasheet/SCT3160KL.... 

Heh..  you on the DigiKey mailing list too?   Grin  I got a new parts email the other day with these in the group...

Yep, that does look like a resonable candidate,  as well as a few others in that same family of parts.  Maybe a bit more capacitance than some, but with a higher current rating.  There are a few larger that should be pretty usable, too... and typically the gate R goes down in larger FET's, but of course the capacitances goes up.
Logged
steve_qix
Contributing
Member
*
Offline Offline

Posts: 2598


Bap!


WWW
« Reply #24 on: October 12, 2016, 09:00:29 AM »

The SiC fets work well when all is well, so to speak.  They don't have the SOA characteristics of a standard MOSFET, and therefore are much more intolerant of any sort of current overload.  So do be wary of that when designing.

I've been able to achieve VERY high efficiency due to the low R D-S on in experimental implementations.

The drive is more critical for sure. The devices don't saturate like a silicon MOSFET, so absolutely as much drive as is possible will be critical.  Without that, you will notice nonlinearity under modulation, and also random failures on peaks due to die heating.

Just my experience..
Logged

High Power, Broadcast Audio and Low Cost?  Check out the class E web site at: http://www.classeradio.org
Pages: [1] 2   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 0.05 seconds with 18 queries.