Testing a Cree XHP50.2 J4 3A led

Great test again, djozz!

The best improvement is the much better beam, I think. steel also posted some beamshots here: The new LED XHP50.2

I hope I can make some with a deep smooth 52mm reflector and a Ledil Crystal optic next week.

What about using a current limited driver like the LD3 set up for 2S with a max amps of 6 or 7?

Simple enough to use an FET Zener modified and use something less than a top cell so that 10.5A isn’t surpassed. Should be pretty easy with the GA and MJ1 and such out there. Or the 20700B.

Dale, perhaps you know this: what I have with the few zener modified FET drivers that I made (diode replaced with 220 Ohm, Zener parallel to D1) is that the low voltage cut-off is too high (at 3.6V or so, edit: I mean double that of course). I know that it has to do with the voltage divider resistors that are wrong for the altered circuitry, but do you have a suggestion for other values that should at least be closer to the correct voltage cut-off?

The LD3 would melt if you tried to use it with this low Vf LED. It will need a buck driver.

With the Vf that low I am betting that you would have to drop down to laptop cells to keep the current low enough and most of those are only rated for ~5A at most so you would be asking them for twice what they can handle.

I have run into this issue with various low Vf LED’s. You simply can’t get the current low enough even with super thin wires, low drain cells ect with an FET driver without either melting the wires, over driving the LED or pulling more then the cells are rated for.

The only practical option I can think of going forward for FET drivers are custom MCPCB’s with pads for some SMD resistors. The star will heat sink the resistors to deal with the heat and they will reduce the power.

The downside is they are very inefficient and will make the light even hotter.

Or another rather interesting option would be to have a pad for an FET on the mcpcb for heat sinking and then use an opamp LD3 style driver to control it. By having it on the MCPCB the heat sinking should be good enough to allow it to handle it.

Still lots of extra heat and inefficient but at least it is variable and as such the efficiency will improve as the voltage drops.

With LiFePo cells you may be back in business with FET drivers. They provide a load of current but the capacity is way below li-ion.

This is an option, although what sizes are they available in and what prices?

https://www.fasttech.com/search?lifepo4%20battery

This one should do an adequate job :innocent: :
https://www.fasttech.com/products/0/10007811/1683800-a123-anr26650m1a-3-2v-2300mah-rechargeable-lifepo

Thanks for the test. The lack of donut hole could make this a game changer, but the low Vf is an issue.

From HKJ’s data, the A123 maintains right about 3.25V at 5A, which is right around the XHP50.2 forward voltage (6.5V). With circuit resistance factored in the resulting current will be a bit less than 5A. A good solution for some applications, and since the discharge curve for these cells is nearly flat, it will be nearly constant current through the discharge. Probably not bright enough for some, though.

A possibility I just thought of is using one LiFePo cell in series with one standard Li ion 18650 (close to capacity matched). My estimate is that this would result in 8-10A.

Oh HELL no! Don’t mix LiPo and Liion! I sure ain’t gonna do it, y’all are free, of course, to take your own chances!

Jos, I use values in my firmware of 245 and 228 with a 22K resistor paired with the 4.7K normally used. YMMV

So those are the common values, and you adjusted the firmware, that is fair enough. For me changing firmware is out of the question within the foreseeable future, so I hoped to solve this by adjusting the resistor values to match the existing (105C) firmware. But have no clue in which direction to start, I need an educated guess.

I have seen people use a 36k for a 3s setup and 48k for 4s IIRC. Although they still needed fine tuning in the firmware IIRC. Still it shows you need to increase R1.

This test was done on the J4 looking pretty spectacular. Cree list a K4 that does 350 more lumen at 3 amps than the J4. So 5000 lumens is very doable, just not as pretty as this J4 A3 5000K LED. Now just sit back and wait for the Cree XHP70.2 cause the Heart of Texas is going to have scorch marks all over the place.

Thanks TA, exactly the ballpark information that I wished for. I will try swapping R1 for 33K and see what happens. :slight_smile:

You would need a fixed voltage drop of 3.0 Volts over the 220 Ohm resistor, so that when the Voltage goes down to 6 Volts the driver will get 3.0 Volts and low voltage warning will happen.
So i think you can replace the 220 Ohm resistor with a 3.0 Volts zener (in ‘anti-series’) and forget about the parallel zener (from the zener mod).
But, that means the MCU gets maximum 5.3 Volts minus the 0.6 Vf of the protection diode = 4.7 Volts maximum.
I don’t know how much Voltage the MCU can take though, and i haven’t tried this idea either…

When I tried using the zener without the resistor the zener fried in a matter of seconds.

Yes, but that’s when you use the zener in prallel, and that’s to be expected.
What i’m proposing is putting a 3.0 Volts zener in series (but the ‘wrong way around’) in stead of the 220 Ohm resistor so you use its reverse voltage drop.
It will then just eat 3.0 Volts of 6.0 to 8.4 Volts, so that when battery drops below 6 Volts, the MCU will think it’s 3.0 Volts and starts blinking in low mode.

[edit] i suppose “reverse Vf” is called Vr :slight_smile:

Yeah, and you still need a 10Uf on C1. The Zener likes having it’s own 10Uf for stability, but the designed C1 I always populate too.

I’ve used a 39K at R1 but got away from it by changing the firmware.