Convoy S21D w/ legendary Nichia 219b :) - Review & comparison w/ Nichia 519a, E21a & other lights with Nichia LEDs incl. Emisar D4V2, Convoy S21F. (Summary & measurements on P. 1)

How do you like the light so far? Would you recommend it over the S21D?

I don’t know how mixed emitter lights work. the pic you posted in the middle, is it a 50/50 amp mix between the 5700 and 2700 emitters in the S21F? and you say it is at 3700K?

So does the formula I wrote above actually describe reality? “I have noticed in beam shots in the past that the lower temperature LED affects the finished mixed temperature a lot more than the higher one, at least perceptually. So rather than a mixed temperature being (2700 + 5700) / 2, which would be 4200K, it is perceptually more like (2*2700+5700) / 3, which is 3700K.

Assuming that the higher and lower CCT channels have similar output levels, the lower output one will be weighted more in an averaged CCT calculation. On the CIE chart, distances between points 1000K apart become longer toward the low end and shorter toward the high end. You want to take the average of the two points on the CIE chart, not the average Kelvin values.

To calculate the resulting CCT of a mix it’s best to convert to “mireds”, do the average then convert back to CCT.

Mired unit: Mired - Wikipedia

Essentially, CCT is not linear (e.g. 3000K -> 3500K isn’t same amount of change as 4500K -> 5000K)
Mired is linear. I learned that in this thread: Preferred mired (colour temperature)?

So, to convert the tint mix of 2700 and 5700:
2700K = 1000000/2700 = 370.4
5700K = 1000000/5700 = 175.4
Average it: (370.4 + 175.4) / 2 = 272.9
Convert back to CCT: 1000000 / 272.9 = 3664K

More general formula: T1 * T2 / ((T1 + T2) / 2)
T1 * T2 / average(T1, T2)

Tint mix 2700K and 5700K = 2700 * 5700 / ((2700 + 5700) / 2)
2700 * 5700 / 4200 = 3664K

Edit: Oh neat, found the mathematical name for it: “harmonic mean”

Nice discussion. It’s interesting that the calculated result is similar to my estimate from drawing lines between the 2 CCT and take the mid-point. I wonder if it is coincidental or the line method is expression of the mathematical calculation.

I have three 219b D4V2’s with different CCT’s, and their measured outputs:
4500k = 1670 lm
3500k = 1500 lm
2700k = 1350 lm

From the above I suspect the 519a 2700k LED would have lower output at same current as the 519a 5700k. Wonder if it affects the resulting CCT at same current for both LED’s a little bit.

Well, in that case you’d want a weighted harmonic mean. Found an online calculator: Online Weighted Harmonic Mean Calculator

Lets throw out an example and use 2700/5700 with 2700 at weight 1 and 5700 at weight 1.3 (i.e. 30% more output than 2700)

Weighted harmonic mean = 3843K

I prefer fixed CCT lights so I still hope Simon would eventually release S21F that way.

But I also would buy this S21F again based on what I see so far. The mixing of CCT’s gives a unique low-Duv 3700k beam color. The Duv is actually lower than 219b 3500k.

Output at 3700k is low for my taste. I have changed the optic to 10 Degree Clear and now the improved central brightness is much more useful to me.

So, does the 21F have the same buck driver as the 21D….or none at all?


Both are 8A Buck

Yes, according to Simon, but has this actually been confirmed (with pictures of the driver)?

I’m curious how a dual-channel buck topology looks like.

It didn’t arrive today, so no delivery until at least early next week :frowning:

Beamshots on the “wall of truth,” my living room, with white balance of DSLR fixed at 5000k. This is white wall hunting at its best, for me more revealing than the close up shot above. (Next would be real scene of plants and rocks in my front yard. But not yet because I’d like to give my neighbor a break from this craziness.)

I hope you could see the very subtle rosy tint of S21F vs S21D 519a, which is very slightly more yellow. A reflection of the 40 point change in Duv.

Is it worth it? How nutty do you want to be with this hobby? If you are reading this thread, IMHO the answer is a resounding yes, maybe.

. yes, I do see that subtle difference. It is so subtle that I would not buy S21F just to get a rosier 519a.

I would, however, consider dedoming a 519a at 5700K if that were my goal.

But group modes are quite different.

Yeah they are. I slightly prefer the 12 group mode on the S21D, I wish it was the same on the S21F. I’m not a fan of the ramp mode so I use the stepless mode.

Or 219b 4500k, or 519a 5000k, etc. (it doesn’t have to be a de-domed LED). If you are starting a collection or this is first Convoy Nichia light, then I agreed, this may not be the best choice; the S21D is, IMHO. And just to be devil’s advocate for the tint-ramping crowd and pointing out the obvious, a de-domed 519a 5700k doesn’t have tint ramping.

OTOH, for people with multiple Convoy lights already, and the 30 or so lucky people who got it, I wouldn’t advocate selling it either. It’s Convoy’s first tint ramping light, it’s only 30 bucks with a buck driver (!), and it has a beautiful tint, so far (I haven’t tested the other CCT’s).

Simon curiously said something to the effect that he doesn’t have enough 519a 5700k LED (?? pls correct me as needed), so who knows when it’s going to show up again. “Curiously” because I thought he would buy LED in the thousands.

Hmm interesting…even Hank doesn’t have 5700k listed as an LED option….maybe people are running low on stock

I was debating on de-doming the 2700k LEDs in the S21F for a more broad range of the spectrum but was afraid the mixture of domes will have a tint shift. If Simon can’t produce more for a while, maybe I’ll hold off lol

Wondering why you like stepped mode….what’s the thinking? On the most popular group #2, the final step goes from 35% to 100%….I don’t get that idea.

I think if the 21D had ramped mode I already would’ve bought it…part of the reason I was waiting to see the 21F.


I’m not a fan of the switch feel for brightness ramping and the ramp speed isn’t consistent for me. I like it more on Hank lights and with anduril.

Also, I usually like to start my lights off from low to high. With the S21F, the brightness in moonlight mode cannot be adjusted or memorized. If the S21F has the ability to turn off memory mode like the S21D, I wouldn’t mind using brightness-ramp. It seems weird to go from 40% to 100% but in person, it doesn’t get that much brighter for me lol

A trick I like to do with the S21F is while leaving it on moonlight mode, I un-twist the battery cap to reset the stepless mode back to lowest mode.

We don’t perceive light/brightness in a linear way - some eyes are slightly more sensitive than others but generally it takes a lot of extra lumens to perceive a lot of extra light. For most people 3-4 times increase in lumens will look about twice as bright…ish. Remember that Simon is listing current (amps), not lumens, in these percentages, and that emitters are also not linear in their input vs. output, so that compounds it a little more. When those modes were settled upon years ago, the idea was having good visual differences in the light output, so usually you’re going to see at least double between modes, sometimes triple. This will look different to your eye when the beam is outdoors where you can shine at something in the distance…indoors on a nearby wall or whatever, it may not look like as much of a difference.

This is a shortcoming/flaw of the driver and firmware programming. Again with the non-linear thing here. The ramping is made of many many little steps, spaced much more closely together than the few that are in the stepped mode option. If the many little steps are linear/evenly spaced, then we get this weird fast-slow ramping effect. When TK worked on Anduril she figured out that an algorithmic approach for the step spacing rather than a linear approach was best for getting a nice visually-smooth ramping action. This has been brought to Simon’s attention but I don’t know if or when he might get his driver engineer to improve it. He may not think it’s worth the effort, or maybe he wants to sell through stock first, or who knows. With any luck he might actually adopt Anduril after all but that remains to be seen.