[MOD] Armytek Wizard Pro XHP50 to 4x Nichia E21A

I've been really curious why AT use 12V configured XHP50 rather than 6V. From performance stand point, it's an inefficient route. Boosting from 3-4,2 V to 11-12V. The short runtime and heat produced confirm to this inefficiency.

Things get worse in it's Prime variant. Prime Pro now use XHP5 which by design is even less efficient than XHP50. And I assume AT uses the same driver for both.

Below is the chart from my DIY measurement. Couldn't measure anything reliably lower than Main 2 because of my DMM current reading limitation (no sensitivity below 0,1A). As you can see, the efficiency is normal for a boost driver up to Main 3. In Turbo 1 and 2, this Wizard turns into a heater, wasting 40% of the power to heat before it reached the LED!! No wonder it heats up very fast.

At max. power (Turbo 2) the LED only pulls ~1A at 4,2V before gradually lower the output. The driver just can't do sustained boost. This confirms Zak's and Maukka's finding why Wizard Pro and Prime Pro don't live up to the spec (rated at 2300 lumen) outputting only 1500ish OTF lumens. Cross checked with pct.cree.com also confirm this.

EDIT: In 6V test the Turbo 1 and 2 modes gave erratic then stayed at much lower output. The highest recorded input was 2,41A @ 4,2V. About half the 12V test (4,44A @ 4,2V). Something is definitely wrong with the driver. Looks like the test procedures damaged it.

As can be seen from the chart above. Boosting to 6V using the same XHP50 LED configured to 6V the driver efficiency increase. But at the expense of much lower output. Only ~0,6A at Turbo 2. More than enough for general uses. But as Zak said, it loses the WOW factor.

I reflowed the same XHP50 on Kerui's XM DTP board.

Summary:

  • This is a very good headlamp. Not the easiest one for modification but I proved that Armytek is mod-able. The design and machining is very well built. This is a very strong headlamp unlike others. I really like it for many qualities only Armytek has: The looks, toughness, magnet + charging, new super nice headband (very easy to replace/remove the headlamp when needed).
  • As for the efficiency. Most modes other than the Turbo are perfectly OK. Even the inefficiency in Turbo won't last long, it will switch down to next lower mode automatically due to it's excellent thermal regulation (throttled by the temperature read in the driver board, NOT by the LED temp). This is a fool proof design.
  • If you want a better efficiency in higher output modes, instant fix is to use H1-A or H2-C, depends on what LED you're using. This Wizard will fit 22mm driver with no problem. OEM Armytek use 12V configured XHP50. These drivers already tested by Jensen and their efficiency are very good for such boost drivers.

NOTE: I killed the driver. Dunno, looks like I shorted something unintentionally. Too many clicking with fluctuative input voltage perhaps? The inductor, MCU and e-switch seems to fail now.

UPDATE 180202:

Runtime test added:

https://budgetlightforum.com/t/-/48258/14

UPDATE 180204:

VR16SP4 installed:

https://budgetlightforum.com/t/-/48258/17

UPDATE 180206:

- Runtime and output results (XHP50 vs 4x E21A) added

- Firefly modes, CRI, and tint comparison between OEM and modded

: https://budgetlightforum.com/t/-/48258/21

UPDATE 180220:
- 4x E21A (6V + resistor mod) runtime

https://budgetlightforum.com/t/-/48258/29

UPDATE 180509:

- 4X E21A 6500K (12V) reached 1400 OTF lumens and beats the original XHP50 in both runtime and output.

https://budgetlightforum.com/t/-/48258/30

- Clemence

Maybe armytek is monitoring this thread and can provide an answer or give you a job

Thanks for the extensive test!
Your wires seem a bit long though. What AWG did you use?

I think in practise it doesn’t matter too much. The Turbo2 mode is not very useful for a headlamp. The runtime is too short and it dims too fast.

zak.wilson has showed that you can use this driver with a 6V Nichia LED and you will get more practical modes (highest mode will be 60min) with higher efficiency (Nichia 144A R9050 @840 otf Lumen for 55-60min).

I agree regarding the Prime Pro XHP-35. I think it would make more sense to keep using a 3V LED. They could just use a higher Bin (XP-L W2) and maybe increase the current to 3.5A to get a bit more brightness.

Could you tell us the designations printed on the ICs on the driver? This could help us understand it better.

The original AT wire core was 0,4mm thick. I used 0,8mm core.
Will try to mod it later using the H2C or H1A boost driver from Kaidomain. Wizard will accept up to 22mm driver.
Here’s the clearer picture:

OP updated with 6V test

Your results do not match those of zak.wilson.
He measured an input current of 1.1A on the Turbo1 mode when using a 6V LED. You measured around 0.25A.
On Turbo2 he measured the actual runtime instead (55-60min). This should equate to around 2.8-3A.

His results make more sense to me because they basically show double the runtime compared to the 12V LED.

I wonder why yours are different. The one thing I notice is that your cables are very long. Inside the light they are only 1-2cm.

I also find it quite funny that you want to replace the sophisticated Armytek driver with a cheap & simple Kaidomain driver :). But efficiency is important…

This topic is interesting to me because I recently bought this Wizard to do the same mod that Zak did.

The problem with my driver is it showed erratic output by the end of the test. In 6V mode only in one brief moment it showed “correct” result (2,41A in turbo 2). Then it’s more like a lower random output before it finally stopped working.
I will test again using different driver. I had to use long wires, will try to cut it down as much as possible later

I’m guessing you opened yours with your bent screwdriver?

Yup. But for Wizard you have to be extra pppatient. I spent half a day during weekend just to get the bezel open without damaging anything. Faster route is to break the glass. The glass is there to protect the softer TIR optics (from scratch and dirt). Most optics available (at very cheap price) in Ebay/AE are too thick and don’t allow you to use the OEM glass anyway.
Next time I’ll just break the glass. It’s faster and make the headlamp more resilient for many uses. When the glass break, water will instantly seep in. PMMA TIR is softer but more resistant to breakage, and when the lens damaged, just replace it.
Note: without the glass, future bezel removal will be a lot easier too

I have to agree. Removing the bezel is quite difficult. I applied some leverage with long scredrivers and also tried pulling up with a bent screw driver, but it only moved maybe 0.25mm.

I managed to open mine today with minimal scratches and no cracked lens. It was easier than expected. I just needed to overcome my fears and apply enough leverage with a screw driver under the ring.

It’s only possible with OEM bezel. I did many bezel removal successfully with no cracked lens nor damaged. But when I ordered the black bezel version. For strange reasons, AT uses two O-rings, sandwiching the glass lens. That makes no room for any tools. Normal Wizard or Tiara have 1,5mm space for pry tools to work.
I think newer Wizard Pro uses two o-rings due to some cracked lens reports.

Well I just finished modding mine with a Nichia 144A and everything worked nicely. Hardly any scratches and no cracked lens. My light is from this fall, not very old. It only has a single o-ring.

Runtime test added:

- Clemence

Very, very good regulation! It’s very flat and steps down a the end to give you enough time to change the battery.
The prime runs a few minutes less before stepping down. The XHP-35 probably has a little bit higher Vf compared to the XHP-50 because of the smaller dies.

If you look closely the flashlights didn’t shutting down after the last step down. They continued to run in firefly mode for much longer, (didn’t want to wait) I predict, for another hour or so. Each tests started with 4,19V and ended at 2,85ish V
With Sony VTC the Turbo 2 will have longer flat runtime in Turbo 2. I prefer the 6V mod to get flatter Turbo 2 at much longer (900ish lumens).

- Clemence

UPDATE 180204:

Finally, after tedious drilling to avoid short circuit the ultra thin dielectric.






And to those doubting the white paint….

EDIT: based on Maukka’s test the Wizard must be using 70 CRI LED

- Clemence

Wow, I love it. I’m perfectly happy with my XM-L2 wizzard, but your mod wants me to buy the XHP50 version just for this modification

The difference seems way too big…
My only interest now is the throw.

I’ll post a video for you later

- Clemence