SiC JFETs

I'm using SiC JFETs in a test circuit - only a small number needed so no cost considerations - and it works perfectly.

It would be a bit large for a production circuit, as I only need to switch about 5A at about 100V fully on or off, so not much dissipation. But all the SiC JFETs seem to be big and very high voltage. All the Silicon ones seem to be small and quite low voltage and relatively high on resistance.

Why is this, fundamentally? Is there nothing in the middle?

(In this particular circuit, a MOSFET doesn't work because of the body diode, there's a small AC on top of the DC. I'm sure I could design round this, but the JFET works very well.)

Cheers

--
Clive
Reply to
Clive Arthur
Loading thread data ...

What SiC parts are you using? The Crees have a (slow) body diode. They do condduct in both directions when ON.

The GaN fets don't have a real body diode, but they turn on at a couple volts negative on the drain. I don't know if negative gate voltage inhibits that turn-on.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

United Sic UJ3N065080K3S

Where do you get them?

Cheers

--
Clive
Reply to
Clive Arthur

Check Digikey; they have a good search engine.

We mostly use the EPC parts, which are great and dirt cheap, but they are BGA-style chip-scale packages, which are harder to assemble and cool than packaged parts.

Lately I've been designing tiny mouse-bite (crennelated?) PC boards with the GaNs and stuff on them, glob-topped. We use them as components on bigger boards.

formatting link

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

Damn expensive!

--
 Thanks, 
    - Win
Reply to
Winfield Hill

Back-to-back MOSFETs, e.g., AoE-III Fig 3.107, page 205, lots of other more simple circuits, depending on usage.

--
 Thanks, 
    - Win
Reply to
Winfield Hill

How small can you get the impedance bump from those castellated connections? I assume you don't use them for real fast path stuff, but have you TDR'd them and could share insight on whether they could possibly be used at 100MHz, 300, 500, 1GHz? Assuming you can optimise them and the parent board for that, of course.

Reply to
Clifford Heath

Not really. Ten test boards to evaluate a principle. Even in production, if we used them, the cost would be a very small percentage.

Back to back MOSFETs is good, but more complicated drive, so more components, so more failure mechanisms. This is highish temperature stuff.

The main problem with these parts is the physical size. Maybe I should x-ray the packages and grind them down to nearer die size. (That was a joke.)

But the question remains, why is there such a huge difference between SiC and Si JFETs, one being heavyweight high voltage and the other being lightweight low voltage?

Cheers

--
Clive
Reply to
Clive Arthur

Isn't that like asking why germanium & silicon diodes are different?

NT

Reply to
tabbypurr

What the hell would anyone want a, say, 200V 1A JFET for, in any material?

:-)

The reason there are more diverse types (JFET, MOS, IGBT, BJT, various cascode combos) in SiC, is it's still somewhat new and speculative, and JFETs were one of the first released (probably?). I don't know if they're still considered as preferred for new designs, by those that use them (military??), but MOSFETs are the better deal for switching, in any case.

Si JFETs are almost completely dried up, because they stink for almost all applications, except the few applications that they're just too good for. There were odd types back in the day, which were basically "tube replacers" and it seems just didn't catch on for anything more.

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Design 
 Click to see the full signature
Reply to
Tim Williams

I use a lot of optoisolated bidirectional SSRs in test sets and stuff, and in production products. They are fairly slow, but have zero switch offset and are small and really easy to drive. We use some Ixys self-protecting SSRs, really slick.

--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin

Some of the GaN parts are depletion mode, which sort of resembles a jfet. Idss of a few amps maybe.

--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin

Dunno. Like;t to grab the AC offline market. My question is why the SiC and other guys use such huge packages, not beneficial to their customers. Heat flows vertically, for the most part. OK, yes a broad surface is helpful to the insulating layer. But hey, we want smaller packages!

--
 Thanks, 
    - Win
Reply to
Winfield Hill

I haven't TDRd them, but 1 GHz should be easy. Light goes a foot in 1 ns, so those tiny features would be invisible. A 1 pF lumped load along a 50 ohm line is just a 25 ps time constant.

Below 1 ns starts to be interesting. Below 100 ps starts to be hard.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

I suppose it is a bit, but the difference seems to be much bigger.

Silicon magically and reversibly transmutes to Germanium at high temperatures - Vbe decreases and leakage increases :-)

Cheers

--
Clive
Reply to
Clive Arthur

Good point. I do remember when SiC FETs were still being developed and the company I worked for had some tie in with a university somewhere or other. That was maybe 12 years ago, so early days yet.

There are SiC JFETS with Si MOSFETs in a single package wired in a cascode arrangement as you mention - not sure why that's better than just a MOSFET, but it does make the JFET easier to use in some configurations.

Yes, I remember now being excited about a Wireless World article about FETRONs when I worked at BBC radio with its acres of valves, so they must have been highish voltage.

Cheers

--
Clive
Reply to
Clive Arthur

The thing about the modern parts we're looking at, is they have very high capacitances. Unlike tubes.

--
 Thanks, 
    - Win
Reply to
Winfield Hill

Mosfets have a lot of Miller (Cg-d) capacitance. A cascode greatly reduces that. SiC and especially GaN non-cascode fets also have lower capacitances per watt as compared to silicon fets.

GaN is amazing. I've done things with them that simply couldn't be done with any mosfet. Ditto SiC.

I think maybe the silicon-GaN cascodes are monolithic.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

Actually, power MOSFETs do have an intrinsic JFET cascode structure, ignoring this makes them hard to efficiently model in SPICE. See papers by RCA engineers Wheatley, Ronan, and others, and a set of app notes AN-7506, 09, 10, 32 and 33.

This is what causes Crss capacitance to rapidly diminish above 15 volts or so, or 50 volts for Super-Junction parts. The latter types feature capacitances as low as 1pF above 100 volts, and give SiC a run for its money.

--
 Thanks, 
    - Win
Reply to
Winfield Hill

They're amazingly sharp. Though I doubt the datasheets. I have been unable to reproduce curves quite so sharp. These curves fit the data, but not the datasheet:

formatting link

This is easily enough modeled with a few junction capacitors (which may have to be written out, because some SPICEs don't accept large M figures, for, absolutely no reason, thanks..). The formula doesn't quiiiite fit (it wanted M ~ infty for this series, very slow convergence on the curve fitting), but still just works.

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Design 
 Click to see the full signature
Reply to
Tim Williams

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.