Nichia E17A/E21A Skilhunt H04RC Group Buy


The E21 has an XP-G / 219C size die, and the E17 has half the mm² surface area die of the E21, so E17 is comparable with XP-E die size.

The E21 can safely handle max. 2 Amperes and the E17 can safely handle max. 1 Ampere.

So a quad of E21 can handle 8 Amperes and a quad of E17 can handle 4 Amperes.

So i think a quad of E17 would be more than enough to handle the power of the H04.

A quad of E17 is also much smaller so it will be less floody (tighter beam) and probably have less ‘donut hole’ effect.

But the donut will be evened out by the optics anyway (i assume)

Probably the E21 quad will have better efficiency because they will have an easy job, below half of their maximum recommended current.

But they cost twice as much as the E17, but the LED boards for both are equally priced and not cheap, so for the total price of this modded Skilhunt, the difference is negligible.

The driver of the light will determine the current and thus the run time, wich will be about the same, but maybe a bit longer for the E21 because it will work on a slightly lower Vf because they run at less than half the max. recommended current.

So in all, it seems the quad E21 is the most attractive option for buying this light.

Well said Jerommel, thank you :+1:
For CW and WW, I suggest to use E21A unless you need throwy confined hotspot. Or if you need pure RGBAL color headlamp.


I decided to go with E17a. The slight drop in efficiency/output is more than compensated by the much higher intensity in the beam. I think it is easy to get a floody beam using a diffused wide optic, but its harder to get a throwy beam. I think the E17a quad board will give me a better range to chose between throwy and floody. The higher intensity in the beam means I will likely use the headlamp in a lower mode, another reason why I won’t worry about the efficiency difference.

If I had more money, I would probably get one of each :slight_smile:

Awesome, exactly what I was looking for, thanks guys. I’ll look into them a bit more before committing.

Quote from another thread:

Is that how they will behave in this H04 as well? (I’m not sure what power levels were being referenced and how that compares to this H04).

Do we have any sense on what the optical efficiency of these two quad die setups are? Is this something maukka can measure? I’m under the impression that the larger the LES, the more losses there are OTF in these TIRs. If this is true, does the gap between the 17 and 21 close?

Yup, just like any normal LED. But 4P setup means 4 times less tint shift

Quadtrix E21A gap = 0,2mm. Total LES = 6,22mm
Quadtrix E17A gap = 0,1mm. Total LES = 4,95mm
I’m sure Maukka will do a thorough test about this


I received several questions about my Skilhunt H04RC modification. But today I received an interesting question:

  • Incandescent lights generates the heat outwards towards the emission direction, unlike LED, which generates most of the heat downwards. This is why you can make powerful incandescent lights body from plastic.
  • Skilhunt limit the max current (then output) to only 2,5A and then stepping down to 800mA in 2 minutes for very good reason: this is one of the most compact 18650 headlamp with powerful output, there’s just not enough surface area and mass (for temporary heatsinking) for continuous max mode. At 800mA (mode 7, next lower mode) the headlamp is already very warm after 10 minutes of continuous use. Restarting the max mode every time it stepped down makes it scorching hot in few minutes.
  • LED efficiency is inversely proportional to the current. The higher the current, the less efficient it becomes. By using 4 LED instead of one, we reduce the input and increase efficiency. Using 4 LED in 4P like on this H04RC also lowers required voltage - longer regulated runtime (about 30 minutes extra in case of H04RC). This headlamp uses constant current buck driver. A buck driver can only regulate preset current as long as there’s still enough voltage uphill. When the battery voltage gets lower than the voltage needed by the LED at preset current, the current also goes down - no more regulation. The output will now follow the battery voltage until there’s no meaningful voltage to drive the LED. Any CC buck driver requires certain minimum voltage to work, in H04RC there’s internal protection at ~2,8V to save the battery from over discharge.

I don’t know about those Xenon bulb but, my gut feeling tells me it’s still less efficient than most high CRI LED. As for the output, make sure it’s real measured out the front after all optical losses.

Condition Tj = 85°C
Bare LED
At 2,5A a single XML2 - U2 CRI 70:
LED output = 919,2 lm
LED voltage (4P) = 3,24 V
LED power = 8,089 watt
LED efficiency = 113,64 lm/watt

Condition Tj = 85°C
Bare LED
At 2,5A (625mA/LED) 4x E21A D320 CRI 70:
LED output = 1094,3 lm
LED voltage (4P) = 2,8575 V
LED power = 7,144 watt
LED efficiency = 153,12 lm/watt

Condition Tj = 85°C
Bare LED
At 2,5A (625mA/LED) 4x E21A D220 CRI 9080:
LED output = 744,5 lm
LED voltage (4P) = 2,8575 V
LED power = 7,144 watt
LED efficiency = 104,2 lm/watt

From the calculation above, you can see why I don’t trust advertised Skilhunt (and most manufacturers) number for 1200 lm output out from a single XML2. Even using XML2 with U4 bin (the brightest available), at 3A (with 2x 18350) you can only get 1197 lumens before those plastic lens transmission losses. And that at junction temp 25°C, which is very unrealistic. Your LED can still overheats in –16°C air environment


I don’t know much about it either, but according to wikipedia it’s between 30 and 55 lm/W.

Looking at the charts I’m surprised the output loss is so severe. According to TA tests, E21A CRI9080 at 2.5A has the same output as XM-L2 U2 (912/910 lm). At 1A XM-L2 is 7% better.
Skilhunt uses U4 bin and you used D220 which is a little lower.
This would make me expect ~18% lower output at peak and ~23% lower at 1A. I’m not sure what’s the current at the steady state but it must be quite close to 1A.
At peak I see ~22% lower output which is can be accounted for by test variation.
But at the steady state the difference is ~43% and that’s definitely too high.

Any explanation?

IMO, TA measurement is too optimistic. Brightest E21A is D320 at R70 = 320lm at 700mA (Tj=25C). It’s comparable to XML2 U4, 320lm at 700mA (Tj=25C). Skilhunt use U4 bin for the CW version, mine is NW version. And I pretty sure it’s U2 bin. I measured H04RC using D220 E21A R9080 which means 220lm at 700mA (Tj=25C). Thus, 4x (4P) E21A D220 at 700mA produce only 245,4 lm - still much lower than a single XML2 U4.

I can’t explain the big difference in output either, my measurement device is not as sophisticated as Maukka’s. Let’s wait for his result.


Thanks for the reply Clemence. I’m patiently awaiting maukkas results, and really hope he does efficiency measurements of the TIRs.

My “unofficial” test showed ~80% - 81% TIR lens efficiency for quadtrix E21A. Pretty normal for uncoated 16mm TIR optic. Quadtrix E17A should be somewhere around 84%


The E21A was tested by TA before he had his lumen calibration corrected with the Maukka lights so those numbers re inflated.

All these numbers are kind of confusing to me. Are they right or wrong? Are we waiting on some more results to come through?

I kind of just want a high CRI headlamp that can output like 400-500 lumens indefinitely (not throttling down because of heat) and I can keep swapping batteries into as they run out. Is this what I’m after?


clemence I think E21A may be is too big for this optics. Could you measure flux without host and optics ? or at least e21 vs e17 flux?

I’m really glad that maukka started selling his calibration lights and that so many of us have them. Before that those were available, people with real numbers would hesitate publishing them because they were a lot lower than the rest (I have noticed easily a variation of 30-40% among calibrations of different people), and the temptation was there to match your calibration to what was generally accepted (many flashlight manufacturrers did that too). My led tests always gave low-ish output numbers compared to “the rest” (BLF members and flashlight manufacturers), until maukka’s calibrated test rig showed that they were actually 7-11% high (depending on spectrum, my imperfect sphere coating generates limited but measurable spectrum errors).
But even apart from calibration differences, that since maukka’s lights are for sale have decimated, the fact that most people (including TA) use cheap luxmeters with very imprecise optical filters, is still a souce of error that can cause a variation that for some tints can be as high as 20% (see my luxmeter tests).

Sure that’s what I’m gonna do.