[[ GXB20 Driver – Homemade Constant Current Programmable XHP50 Single-Cell Boost Driver! ]]

It’s either the coil or the converter itself. It’s difficult to say because the converter works a little different and has like a “built-in diode”. The diode is replaced with a mosfet controlled by a deadtime
control logic.

It’s really annoying that I spent all the time fiddling around with the 17mm board (probably finished by tomorrow), but loneoceans did it already :person_facepalming: .
I hope my version isn’t completely wasted time :wink:

Very interesting, that is a nice IC. I was curious why I didn’t see a diode listed in the parts list.

Or triple xhp35.
Though my dream would be this driver on a 20mm DTP PCB. With a series od 3 XP-L2s on the same board.

I’m a driver but not a driver expert nor tweaked any LED drivers (resistor stacking doesn’t count) before. My question is: from all those components cramped in a tiny 17mm which one is the hottest one? I usually find the big square box is the hottest one, am I wrong?
Can we remotely place it somewhere else (near the tailcap perhaps) to save space and distribute the heat better?

- Clemence

Typically with how the drivers are placed (right behind the LED in the same thermal block), the major heat producer is the LED itself! However you are right that the next main heat producer is typically the inductor. However moving it far away from the LED (e.g. at the tailcap) isn't typically a good idea, since a significant amount of parasitic inductance is added, not to mention very large DC restive losses (since the inductor current is typically very high), as well as skin-effect losses at high switching frequencies. The ideal layout is to keep the inductor as close to the power circuit as possible.

Practically speaking, I think for most single-cell flashlights (esp. 18650), the maximum reasonable power one can expect is about 25-30W for short durations and <20W for any reasonable run-time. Anything longer than that would require a fairly large host with cooling fins, which I suppose defeats the 1-cell form factor (i.e. desirable to be small); not to mention DC power losses. For example, just 50mR DC losses from a single 4V cell outputting 30W would lead to 2.815W lost in just the wiring/switch/spring, a loss of almost 10%.

As a side note, the GXB17 is well underway - follow it here: https://budgetlightforum.com/t/-/46062

Is this efficiency measured for a particular input voltage or is it an integration over some range?

Hello; as mentioned earlier in the post,

"The test [of which the results were plotted above] was conducted as follows. The driver was hooked up to a constant voltage programmable Agilent power supply. In this case I conducted the test at 3.7V in across the range (driver was also tested at a variety of input voltage from 4.5V to <3V but less comprehensively). Input current and voltage was measured via Kelvin terminals to avoid errors due to lead resistances. Output current was measured across the load resistor as well as the output to find total driver efficiency. A total of 20 constant current levels were tested and measurements taken. "

I understand the efficiency may vary across different input voltages but a quick test from around 4.2 to 3.5V for a few different output levels seemed to yield comparable efficiency.

Thanks for the explanation

3.5 to 4.2V is nowhere near the real voltage window of a fully loaded li-ion cell for all of its discharge curve, and that's not to say the driver sees less voltage due to contact losses. As an advocate of worst case scenario conditions, for me the input voltage value which really matters is the one which makes the driver's life harder, so I'd also test with the possible voltages coming from a nearly depleted fully loaded li-ion cell in an overdriven driver: 2.7 - 2.8V.

Great work by the way.

It depend significantly on what cell is being used, how fresh the batteries are, and at what sort of operating limits one will tolerate the driver at. I designed the driver without really expecting to push LEDs to their absolute limits - this places not only significant thermal stress to the LEDs, but also the batteries etc and affects long term reliable operation. You are right to say that the driver input voltage will be closer to 3V- on a depleted overdriven cell - at this point the driver will try to draw even more current from the battery causing the voltage to drop even further, until low-voltage protection kicks in. That said, I'm currently limited in my bench power supplies to provide sufficient current at such low voltages, so it's also a limit to my test gear for proper characterization :)

As a side now, I'll be focusing my efforts on the GXB17 development, so expect to see less updates over here (since the GXB20 V2 is essentially complete and is working happily in 2 of my flashlights!), and I'll be focusing on the GXB17 over here: https://budgetlightforum.com/t/-/46062

I see the trend in FL manufacturers is keep pushing the limit until the efficiency is just a bit better than xenon bulb torches years ago.
Higher voltage LED and lower current applications is now becoming more mainstream to boost overall efficiency. Boost drivers have to keep up sooner or later.

Great job Loneoceans.

I’m greedy! I have some nice 4x UV 20mm boards and needed a17mm driver.

https://m.fasttech.com/products/1616/10002799/1208600-driver-pillar-w-emitter-slot-for-flashlight-diy

http://www.ebay.com/itm/311785443407

If you need cash for dabbing… Will donate/Kickstart.

+1

Hello all, just a quick update that I've a simple write-up (in progress) of this LED driver.

http://loneoceans.com/labs/gxb20/

It's mostly a rehash of stuff on this thread though. More details to be added there as I go along. Thanks for reading!

Great writeup, very informative.

Yes, very good, thanks!
Do you intend to extend the writeup to cover GXB17 too?
You mention that you use PWM because the driver won’t be used under 200 mA. That’s not true already. Is it still PWM?
In the writeup you mention having the lowest mode at 50 mA. I notice 1 mA in this thread. Which is the latest value? Also, it would be useful to add the range of currents supported by GXB20 v2, not just that there are 256 of them.

BTW, loneoceans, have you considered something like ATBTLC1000 for BT programming?

Yes the GXB17 driver will have its own writeup. Yes the driver is running in FPWM mode, but you're confusing that with the 'regular PWM' running in the kHz range which people use to modulate brightness. The Forced PWM mode is how the boost driver is configured and how it regulates its output. In this case that PWM is running in hundreds of kHz and is the switching frequency of the boost switches. The output driving the LED is true constant current. Finally if you look at the firmware diagram at the bottom of the page, the driver has a 'moonlight' mode which drives the LED at a very low current of around 1mA. The range of currents supported are 256 constant-spacing values, but the max is determined by its desired mode of operation, e.g. running at 6V out or 9V out, and configured accordingly to your desired maximum current.

I've worked on similar parts in the past but it really seems like far too complex work for a driver like this and a bit too much overhead as well as cost.

As far as I understand: You’re PWMing power into an inductor and you could be PFMing which is more efficient at low load. I have no idea what ‘low’ is, but seeing you mention 200 mA I assumed that this number is somehow related to the limit of lowness. Since GXB20v2 bottom 200 times lower than that, I assumed PFM could be preferable at the lowest modes.

I’m not sure if it’s suitable, but ATBTLC1000 starts at $1.71 at DigiKey. I believe that many would pay even several dollars to be able to easily have perfect modes.

Moodular and Lux-RC offer such lights already, though that’s well within the high end of the market.

The SOC is cheap but there are a lot of peripheral components that need to go alongside almost all RF SOCs including a chip antenna, internal regulator inductors and passives and crystals etc. That's not even including the very limited PCB space on a 20mm/17mm diameter board. Could I make it work? Sure, but this will require much more expensive PCB fab requirements, will complicate hand-assembly, and still requires more firmware overhead... Lux-RC's optical flashing programming does look interesting though but again requires additional infrastructure on both the host and client side, which is far too much work for this particular hobby project (which I only planned to make a few for my own flashlights). However anyone is more than welcomed to add more features on to the GXB driver.

Changing FPWM to PFM is just a disconnection of the mode pin so it's really not a big deal. It's a trade-off between audible PFM at low loads with higher efficiency, or just sticking with FPWM for always quiet operation.