+ Post New Thread
Results 1 to 6 of 6
  1. #1
    Full Member level 6
    Points: 2,927, Level: 12

    Join Date
    Feb 2014
    Posts
    355
    Helped
    126 / 126
    Points
    2,927
    Level
    12

    GAN (Gallium Nitride) Fet Discussion

    I want to spark another discussion to try and get people's experiences and impressions on GAN's right now. My understanding of the market is this:

    Transphorm
    • Only 600/650V, only cascode (silicon fet in series with higher voltage depletion GAN)
    • Cascode means gate drive is really driving a silicon fet -> easy
    • Standard packages (TO-220, TO-247, PQFN)
    • Standard gate drive + standard packages make this the easiest to try
    • Available on digikey


    GAN Systems
    • 100V/650V enhancement mode
    • Custom but straightforward packages, top side or PCB cooling (10W is realistic)
    • Medium difficulty gate drive - wants 6V max is 10V
    • Available on mouser


    EPC
    • 20-200V with a couple 200-300V devices, enhancement mode
    • Tiny specialized packages, only PCB based cooling possible (probably ~5W or less)
    • Tricky gate drive, wants 5V, but absolute max is 6V
    • Perhaps the most established?
    • Dozens of parts available on digikey


    TI
    • One 80V part (half bridge) and one 600V part
    • Integrated approach, gate drive, regulator and enable fet is all in the same package
    • 80V part available on digikey, 600V part is in 'preview'


    On Semi
    • Rebrands transphorm?


    Infineon
    • Has some parts, keeps them secret for now



    Any other comments thoughts or impressions?
    Anyone using these or know companies who are?
    Which brand and why?
    What they being used for?

    •   Alt12th February 2018, 16:51

      advertising

        
       

  2. #2
    Advanced Member level 5
    Points: 19,887, Level: 34
    Achievements:
    7 years registered

    Join Date
    Jan 2011
    Posts
    3,294
    Helped
    1172 / 1172
    Points
    19,887
    Level
    34

    Re: GAN (Gallium Nitride) Fet Discussion

    I have lots of experience with EPC devices for research projects. Their specs don't overlap much with Transphorm and GAN systems parts, and they're going for different applications, so a direct comparison is difficult.

    Gate drive isn't too hard for the EPC devices so long as you follow their recommended layouts. That means using at least four layers, and small drill sizes. But the packages are not difficult to work with, I've had them assembled a couple times with no problems.

    Thermal management is a bit tricky for EPC devices, since you can't withstand too much force without cracking the dies (especially if the dies aren't perfectly level). If you use a soft TIM and some spacers under the heatsink then they can dissipate as much power as other packages (per unit area, anyways).


    1 members found this post helpful.

    •   Alt12th February 2018, 18:17

      advertising

        
       

  3. #3
    Full Member level 6
    Points: 2,927, Level: 12

    Join Date
    Feb 2014
    Posts
    355
    Helped
    126 / 126
    Points
    2,927
    Level
    12

    Re: GAN (Gallium Nitride) Fet Discussion

    Thanks for the comments. Would you have any concerns putting them in a production design (reliability, availability, end-of-life concerns)?

    The 100V Gan Systems devices do go head to head with EPC at that voltage. I just did a quick comparison and for the same Rdson EPC and Gan systems were within about 30% or less on most specs (input/output C etc). Though EPC covers that space with many more voltage and Rdson choices.



    •   Alt12th February 2018, 18:44

      advertising

        
       

  4. #4
    Advanced Member level 4
    Points: 6,571, Level: 19

    Join Date
    Dec 2015
    Location
    Madrid, Spain
    Posts
    1,180
    Helped
    284 / 284
    Points
    6,571
    Level
    19

    Re: GAN (Gallium Nitride) Fet Discussion

    GaN is starting to be widely promoted for DC DC converters (see EPC's YouTube channel) and now with all the Wireless Power Transfer thing, is taking advantage of that as well.
    Because GaN allows higher switching frequency in converters, new ferrite materials are starting to being developed as well (50% less core loss is claimed compared to standard high frequnecy ferrite at "Next generation ferrite material for SiC & GaN applications ").
    Tricky gate drive, wants 5V, but absolute max is 6V
    GaN gate drivers implement a clamp feature apart from the usual UVLO one.


    1 members found this post helpful.

  5. #5
    Advanced Member level 5
    Points: 35,873, Level: 46

    Join Date
    Mar 2008
    Location
    USA
    Posts
    5,767
    Helped
    1670 / 1670
    Points
    35,873
    Level
    46

    Re: GAN (Gallium Nitride) Fet Discussion

    You could actually drive the EPC FETs from a 3.3V source
    with only a trivial Ron penalty.

    My EPC half bridge driver used an internal LDO (times two)
    to ensure that the high side and low side devices saw the
    same domain voltage, so consistent timing (shoot-through
    control, not amenable to a lot of feedback at 80V common
    mode offset, need fixed timing). And of course, to stay
    well away from the 6V limit.

    EPC is playing games with how they rate current on the
    newer products. I called them out on it and got the
    response that the rated current is pulsed peak, not
    continuous.Caveat emptor. The current they rate is
    simply not reliable (one solder ball, a hundred amps?
    Not in this universe, unless you are force-cooled to
    room temp. My regular customers aren't so lucky).

    For hand built stuff the GaN Systems devices will be
    a lot easier / more familiar to handle. But for ultimate
    compaction, "no package" wins out.

    Intersil / Renesas has just announced some co-packaged
    EPC FETs and drivers (low side only as I make it out).
    However the press release is aimed at the HiRel market
    and Intersil is the kings of rape-the-customer pricing
    in that segment. Maybe there's a plastic version that
    is not insane, but I have not seen that announced. I'd
    expect more in this vein from other vendors.


    1 members found this post helpful.

    •   Alt13th February 2018, 07:03

      advertising

        
       

  6. #6
    Advanced Member level 5
    Points: 19,887, Level: 34
    Achievements:
    7 years registered

    Join Date
    Jan 2011
    Posts
    3,294
    Helped
    1172 / 1172
    Points
    19,887
    Level
    34

    Re: GAN (Gallium Nitride) Fet Discussion

    Quote Originally Posted by asdf44 View Post
    Thanks for the comments. Would you have any concerns putting them in a production design
    If the application would benefit significantly from them, sure.
    (reliability, availability, end-of-life concerns)?
    I'm not worried about these, though obviously I wouldn't use any devices in engineering status for production (they actually sell engineering status devices via digikey, which is pretty neat IMO.)

    The 100V Gan Systems devices do go head to head with EPC at that voltage. I just did a quick comparison and for the same Rdson EPC and Gan systems were within about 30% or less on most specs (input/output C etc). Though EPC covers that space with many more voltage and Rdson choices.
    Really? Which devices are you comparing? I'm comparing EPC2045 to GS61008P and seeing equal Cout, but less Cgtotal and Cgd for the EPC2045. They're both referred to as "7mohm" devices but EPC refers to the maximum Rdson (typical is 5.6mohm), while GaN system refers to the typical Rdson.

    - - - Updated - - -

    Quote Originally Posted by dick_freebird View Post
    EPC is playing games with how they rate current on the
    newer products. I called them out on it and got the
    response that the rated current is pulsed peak, not
    continuous.Caveat emptor. The current they rate is
    simply not reliable (one solder ball, a hundred amps?
    Not in this universe, unless you are force-cooled to
    room temp. My regular customers aren't so lucky).
    I always see two ratings in their datasheets, one for pulsed (usually 300us) and one for continuous. For some applications (Lidar, for example) such a pulse rating is useful, and I don't see anything strange about it.


    1 members found this post helpful.

--[[ ]]--