99.9% of “GaN” AC adapters do not have Gallium Arsenide power switching transistors in them; some (not even a majority) have just a GaN gate driver to drive the silicon mosfet that regulates voltage on the charger.
I didn’t realize GaN had become just a marketing buzzword for chargers, I had thought they actually used gallium arsenide transistors in them to output power.
edit: GaN= gallium nitride NOT gallium arsenide
Linus did a recent review of some of these GaN charging devices, a sponsored video so there is that.
I think he’s part of the problem. I just looked up the anker 737 charger he was touting in the video and it uses silicon mosfets for all the power conditioning, except in an intermediate buck step down before it hits the silicon mosfets for the actual (the bulk of) voltage conversion to take place.
-That exception isn’t as impressive as it sounds, one of the reasons they’re using the GaN transistors only for an intermediate stage is that they are ~140mΩ RDS(on) parts, while the silicon mosfets the charger uses for final stepdown are 1.8mΩ RDS(on) parts. Now I’m sure the GaN FETs have better Qg than the silicon mosfets to get out of the ohmic region faster, perhaps 2-4 times better, but that doesn’t matter when you have 75 times more steady state resistance.
I’m not saying the engineering isn’t impressive, but it had little to due with GaN FETs used in it.
GaN is gallium nitride. (GaAs has many uses, particularly solar cells.)
oops apparently all the III-V compounds just blended together in my head!
might be worth a L1 list of the known good ones?
The only charger I found that had GaN FETs for all the important switching is EPC’s EPC9171… which is only a reference design and the kit costs ~1400USD.
It is the only 240w “extended power range” USB charger I’ve ever seen though.
Maybe, but it could be a fool’s errand. Manufacturers change designs on a whim, most recent I remember is with NVMe SSDs where there was huge performance difference between same models because they just swapped controller and memory for inferior ones and calling it the same Super Mega Gamer whatever marketing bull crap.
Government bureaucracy might actually play to our favor in the post-launch gimping of products matter. If major components were replace in the power supply the manufacturer would have to get it recertified with the bajillion compliance agencies again… as opposed to the SSD manufactures that swap out the memory and controllers on their SSDs without consequence once they get good reviews under their belt.
Where did you get this conclusion?
I saw one teardown. I got the impression that this unit has at least three GaN transistors in two chips. One for DC boost in front of the primary side, two for switching voltage on the primary side (perhaps in a full-bridge topology…not sure).
I’m not a fan of GaN charger yet. Haven’t owned one. But seems GaN chips aren’t expensive to include in consumer chargers.
The 99.9% isn’t thoroughly empirical, its just a number I threw out because I couldn’t find a single charger that used the GaN FETs for the important switches on the power adapters… it seemed like they are only being used for the relatively unimportant switching.
The Anker 737 is a hybrid flyback design (plus a power factor correction circuit) to get to the highest nominal output voltage (plus some smallish offset to account for ripple under high load); Then after that there are buck converts to actually deliver the appropriate voltage.
Anker’s only using GaN FETs for the PFC boost and half-bridge of the flyback primary, I would have expected them to use GaN for the rectification prior to flyback, for the synchronous rectifier on the secondary side and in the buck converters downstream that actually handle the bulk of the voltage conversion (atleast by losses).
Using GaN for rectification is actually a place where they are very beneficial because they effectively have zero diode reverse recovery charge unlike silicon. I could accept no GaN rectification on the primary side because it’s only 50-60Hz, but it absolutely should be there for the secondary side synchronous rectifier. it’s like Anker put GaN in the places it is least effective to improve efficiency.
Now that I’m thinking about it, I’m not sure what kind of ringing the PFC boost circuit creates, it may actually sink alot of current into the Qrr of the primary side rectification.
All the more advanced charger designs seem to be going to an resonant LCC converter topology instead of hybrid flyback so the intricacies of those designs are going over my head.
The GaN charger designs in consumer space seem pretty much dictated by a couple of GaN chip suppliers. Most of the designs are probably from their application notes. In all the teardown’s I’ve seen Navitas chips are common place. So likely I would think the current “state of the art” is where the sweet spot is between technology and cost.
I was told one advantage of GaN power transistors are low heat, higher frequency, and meaning an even smaller transformer. That implies internals of a charger can be packed even more densely, giving an overall compactness of the end product. So from this perspective, current GaN chargers do what it promises. The trend likely to continue for a while until the market hits the next level of evolution.
I think your right, the lack of reference designs are why we don’t see better implemented chargers.
The GaN FETs to make significantly better chargers exist, the thing that started this whole train of thought for me was buying some GaN FETs to possibly replace silicon ones in a design.
I can get a 0.8mΩ RDS(on), 25nC gate charge @ 5v GaN FETs commercially, they are ridiculously good for high frequency applications at high power; a similarly optimized silicon mosfet of the same RDS(on) would have at least triple the gate charge and would require 10v to get the gate fully working.
Since my engagement in this thread, I watched a few more reviews about GaN chargers. One is by a hugely popular youtuber, sort of geek type. He teared down his GaN charger but was disappointed to find a discrete silicon MOSFET not GaN MOSFET.
From my brief survey, discrete GaN MOSFET is expensive and perhaps way over spec’ed for consumer chargers which top out around 150W. That’s not considered high power application by any measure.
What the industry adopts, as the current “state of art”, is an integrated chip which include a driver, some sort of control logics, and the GaN MOSFET. E.g. the common Navitas chips. Why they could integrate a GaN MOSFET inside a SMT chip is interesting. I’m guessing GaN MOSFET generating little heat is perhaps the primary reason.
So for all intents and purposes, almost all relatively new GaN chargers have GaN MOSFET and the benefit of it.
The aforementioned youtuber perhpas bought a very early design or a cheap design (the video is >1 yr old). He seems like a good tech tuber in general but that particular video is misinformation as of now. So I won’t link here.
One interesting observation from his particular sample of GaN charger is that its efficiency is actually not as good as old non-GaN chargers at the very common 5V and 9V. It might be because of the poor design of that particular GaN charger though. But I won’t be surprised good old non-GaN chargers do better at efficiency simply because of more than a decade of optimization in the designs.
So while I’m not abandoning my good old chargers anytime soon, if people need compactness for travel or absolute need for a new charger, GaN chargers are the way forward. Otherwise, no rush to replace good old work horses.
Those Navitas chips, the ones with the integrated drivers, are exactly like the powerstages that motherboard manufactures use for their motherboard VRM functionally; they’re half bridges with gate drivers, they’re also ~2 orders of magnitude more resistance than what a motherboard manufacture would use which isn’t very impressive and would explain why they might not be as efficient as silicon under certain circumstances (high load).
Perhaps when USB goes to 48 volts we’ll start seeing GaN FETs used on the load side of the converters instead of only on the primary side; reason being that the load side of the adapter will then need to be able to provide anywhere from 5 volts all the way up to 48 volts which necessitates higher switching frequency, which is where GaN shines.
Perhaps we’d also see the de-adopting of GaN on the primary side at that point too since the primary side’s step down conversion ratio will be reduced.
Ironically enough after evaluating the highest end power GaN FET I could commercially find I’m going to only adopt a GaN gate driver and stick with silicon MOSFETs. The GaN FETs had some packaging disadvantages and just too much resistance to be practical, even though the GaN FETs had a 50% better RDS(on) x Qg figure-of-merit than the silicon MOSFETs did.
Here’s some die shots of the FET, they are supplied as bare passivated dies as opposed to encased in polymer like most discreet components:
physical size ~6mmx2mm
the top of the FET is silicon as can be told by the bluish-grey. GaN is yellowish
in this side profile (chip is only about 500 micrometer thick) you can actually see how the GaN epitaxial layer sits of the silicon
This topic was automatically closed 273 days after the last reply. New replies are no longer allowed.
