LED driver question and plan

[Dr. Mario]

I have had a hard time finding a 22mm boost DC-DC LED driver for my upcoming quad Nichia 319AT Convoy S11 LED flashlight, and finding the LED drivers as nice as the Mountain Electronics' own version of boost DC-DC LED drivers which may or may not be coming back, so I don't have too much option but to whip my own driver to light up all four LEDs in series.

I decided to give Gallium Nitride transistor a shot to try and squeeze a bit more efficiency out of it, not to mention logic voltage gate driving which could potentially make it easier for the low power ARM microcontroller to switch the MOSFET directly with low part count. The chosen transistor is the GAN Systems GS-065-011-1-L as it comes in convenient DFN package, to keep the real estates as low as possible.

Hopefully it won't cost me too much to have OSHPark make the 22mm round PCB for that driver.

Whoever have used this transistor, is there a pro and con in using it in the low voltage, high power electronic projects?

It is tough with all the LED driver shortages, this time caused by Coronavirus...

Sent from my OnePlus 6T using Tapatalk

 

[DIWdiver]

Great idea to build your own. I've built a number over the years.

The PCB should be pretty cheap at OSHPark if you can keep it to 2 layers. If not try Seeedstudio or PCBway. The latter two offer assembly as well (OSHPark may as well), but I've not been as impressed by their assembly prices for hobby work.

The pros for the FET you selected are high voltage and low gate charge; the cons are high resistance.

I would think you could do a lot better with a silicon FET. Something like the CSD17313Q2T has 6x lower resistance, 7x smaller size, and 10x lower cost. Does that make it 420x better? The gate charge is a bit higher, but unless you are running very high frequency, that shouldn't be a problem.

If you want something in a similar package size, try the RQ3L050GNTB. It's only 2.5x lower resistance, but still 10x lower price (almost), and not too much higher gate charge.

If you are running lower frequencies, you can allow for higher gate charges, and a lot more possibilities open up.

 

[DIWdiver]

By the way, you might do better in the Home Made and Modified forum. The mods can move the thread if appropriate.

 

[Dr. Mario]

Mods - You can redirect to homemade and DIY electronics thread if necessary.

The reason why I chose this FET is because of lower gate voltage, that I can have the microcontroller to directly drive the FET at higher frequency due to 1 microhenry shielded SMPS-class choke (Vishay IHLP2525CZER1R0M01) being used to boost the battery voltage up to 6 - 12 Volts from the Lithium-ion batteries (Li : FePO4 cells included too), in order for certain LED arrays (four 319AT 5,000 k LEDs in this case) to light up.

I will do the driver board run at JCBPCB and OSHPark so I can see for myself who do it best.

I want efficiency and intelligence in that driver board, the only challenge is the real estates, whether I can fit some semiconductors on the board among the required passive components. The microcontroller chosen is the Silicon Labs EFM32TG11032 microcontroller as it's stupid cheap and have the hardware I need to make the boost converter out of it.

Sent from my OnePlus 6T using Tapatalk

 

[HarryN]

Thank you for posting about this. I hope that you will continue to update us on your progress.

I have been involved on the mfg equipment end (epi) with GaN for a long time.

I didn't realize that there were versions that could be direct driven form controller rather than needing a driver.

Are you going to try to run it in the MHz range? It can be a real learning experience to design in these higher frequencies, but the results can be quite impressive .

Traditional board materials might not be sufficient from a loss perspective, but done correctly you can sometimes make your own caps with a dual layer board.

 

[Dr. Mario]

Yep, aware of MHz phenomenons, I will have to design it carefully in the KiCAD and do some final calculation so I don't ring that transistor to death. They're expensive right now, so it is already important to know what to expect. I will have to research the inductance and capacitive nature of the chosen PCB materials and design accordingly.

As for the GaN MOSFET direct driving, it apparently depends on transistor construction and capacitance of the gate in question - this one has 6 Volts maximum gate voltage. Several hundreds Kilohertz tend to be more sane as for the SMPS. I am going to go in uncharted territory as far as the exotic LED drivers go.

EDITED: I wonder if I should put the RS232 pads right on the bottom of the LED driver board, where the positive terminal of Lithium-ion battery contacts with the soldered on brass button on that side, or should I keep it on the side where the electronics are (to update it, I will have to take it out rather than just remove the Convoy S11 head and plug it right in without removing anything in it).

Sent from my OnePlus 6T using Tapatalk

 

[Dr. Mario]

I decided that the RS232 contact pads will go on battery side of the LED driver so I just have to remove the head on my Convoy S11 flashlight containing this driver then just use the pong pins containing the USB to RS232 converter to flash new firmware onto it, adding or removing the flashlight functions easily without having to take the driver out, as it's best to take real estates seriously.

Done with ordering the semiconductors, now I will order some more passive components like resistors, capacitors and high speed rectifier (I wonder if the higher amperage microwave diode will work at ~1 MHz in the boost converter circuitry).

EDITED: Schoktty diodes apparently are capable of rectifying the current at very high frequency, even in the MHz range, so I may go with this diode for the output of the boost converter here to up the voltage.

Sent from my OnePlus 6T using Tapatalk

 

[HarryN]

It looks like mouser has that GaN device for ~ $6.

Amazing, I remember when GaN on Si was considered just a theory. Now it is a commodity.

___________________

I have used this company for LED drivers before:

https://www.4pcb.com/blog/tag/colorado/

And believe it or not had some "machined" in Australia instead of the traditional photo chemical approach.

Having them machined allowed construction of the board using essentially a copper "board" the thickness of normal FR-4 and then the layers were just attached on top of that normally. It was great for moving heat from the devices to the flash light walls. Single sided of course.

 

[Dr. Mario]

Yep. Around $6 a pop. I bought it for ease of driving as I don't think there's room for the MOSFET driver so I have to wing it here owing to the impressively low gate voltage (as spec'd out in the datasheet).

I also bought it out of curiosity as the microcontrollers and processors in general will be using Gallium Nitride (and other LED semiconductors materials depending on who will be making them) soon due to Silicon running out of steam as transistors are getting stupid small, going past 5 nanometers gate width.

Sent from my OnePlus 6T using Tapatalk

 

[HarryN]

Yep. Around $6 a pop. I bought it for ease of driving as I don't think there's room for the MOSFET driver so I have to wing it here owing to the impressively low gate voltage (as spec'd out in the datasheet).

I also bought it out of curiosity as the microcontrollers and processors in general will be using Gallium Nitride (and other LED semiconductors materials depending on who will be making them) soon due to Silicon running out of steam as transistors are getting stupid small, going past 5 nanometers gate width.

Sent from my OnePlus 6T using Tapatalk

Yes, pretty soon we will have to use a silicon interposer like AMD uses on their processors instead of PCBs.

 

[Dr. Mario]

Yep, I heard of it, that may bring around the long-awaited Threadripper APU (or not), as that APU could have more powerful GPU so HBM memory could be required (I wonder if Radeon DNA is capable of out-of-order execution pipelining within the vector processors in the shaders to deal with photorealistic games which are basically code spaghetti at this point, while I know it is already superscalar - there are multiple vector ALUs within a vector processor in the shader engine cluster, while superscalar out-of-order GPUs already exist for a while especially in the Snapdragon processors in the phones, and more recently, the Arm Mali G75 / G77), along with having more options for whatever AMD thinks up of. Exciting time ahead.

As for the LED, I wonder if Nichia would plan on pumping out powerful LEDs like the 219 and 319 with steady increase in performance (some newer Nichia LEDs regressed on brightness capabilities).

Sent from my OnePlus 6T using Tapatalk

 

[DIWdiver]

What supply voltage were you planning on using? The silicon FET I suggested has lower resistance at any gate voltage above 2V or so.

 

[Dr. Mario]

3 to 4.2 Volts input, from a single Lithium-ion cell. I will look at that FET - ideally, I would like to keep that transistor as compact as possible as 22mm LED driver board isn't that much space. (Gallium Nitride transistors have some edges against the Silicon version, and that's the component size.)

EDITED: That Texas Instruments logic power MOSFET isn't bad, however, I needed 10 Amps current capability to stay within the SOA (safe operation area) of that transistor because the whole thing could consume around 6 Amps or slightly more for 12 Volts current at 2 Amps to light the Nichia 319 LED array at full brightness.

* What about that one, CSD17551Q3A?

Sent from my OnePlus 6T using Tapatalk

 

[DIWdiver]

I guess I never realized the 319 could handle that much power.

That FET looks good too, but you are going to need a lot more drive to push the gate at 1 MHz. It would take 60 mA to charge or discharge that gate in 100 nS.

I wonder if this would work:

If you had 2 processor pins, you could use one to drive a 2-transistor totem pole, and one to drive the gate directly. The totem pole would get you most of the way there fast, and the other pin would get you the rest of the way.

 

[Dr. Mario]

Totem pole is a very good idea, I considered going that path as totem pole IGBT driver chips exist and they can be found in a very small SMD package like QFN. I may probably design two separate driver boards with two different transistors and test them altogether.

And yes, I have a single LED Convoy S11 flashlight with the Nichia 319AT LED at 2 Amps, and it's quite bright. It's also excellent with 26650 Lithium Iron Phosphate cells which I bought some from Bioenno Power.

EDITED: Rohm Semiconductors BD2270HFV looks like a good MOSFET driver for the Silicon version of MOSFET that I picked, I will use it in Silicon version of LED driver board - it's too much for GaN MOSFET I bought for this flashlight, as it charges up to 13.5 Volts at Vcc of 5 Volts.

Sent from my OnePlus 6T using Tapatalk

 

[DIWdiver]

You know, I think I looked at that driver and discounted it because it was a high-side driver, and you need a low-side driver. I see now that that was a mistake.

However, I was not wrong to discount it, only in my reasoning.

If you look at the gate rise and fall times, they are specified at over 100 uS. And that applies at very low gate charge that we have been looking at. It would be worse at higher gate charge. This is at least 1000 times slower than necessary to operate at 1 MHz. While this part might be useful in some very low speed switching applications, it is completely unsuitable for switching regulators, even the very lowest frequency ones. Operating at 1 kHz, this thing would be too slow.

 

[Dr. Mario]

Ah, good point - it makes sense as the charge pumps tend to take their sweet time charging to high voltage. I will have to do it the hard way, using dual bipolar transistor array to drive the gate.

Sent from my ONEPLUS A6013 using Tapatalk