[Review] Sofirn SP33 V3 - XHP50.2 3V eSwitch 26650

Just looks like the control loop is a little too twitchy. I doubt the temperature sensor is right on the LED’s board or shelf, but probably on the driver board. That lag is the gear equivalent of backlash, a delay. So throttling down seems to do nothing until the (reduced) heat works its way back to the driver. Driver then senses a cooler temp starts cranking it back up, and the delay again skews the readings.

Simple example of a(n oscillatory) control-loop that needs tuning.

The big disvantages of V3 is lack of output regulation. V2 have very good and stable 1000lm regulation, because is used CC boost driver.

I loved that but hated turbo disappearing after the first 4-6 minutes. Models shouldn’t advertise a light at 2500 lumens if it drops out like that. I understand that’s how a boost driver works but it’s still frustrating

The problem is not boost driver but thermal regulation of firmware or missing thermal sensor. For V2 it can withstand longer turbo times, but I think there is time stepdown. Also from gchart review we can see exactly the same V3 have awful thermal regulation. We need V4 with Anduril and very good firmware thermal regulation, they do that with new IF25. I really like 5000K option of V3 compared to V2 cold white.

Anduril has great firmware thermal regulation but I can’t sell it as it’s not user friendly. Sofirn also don’t like it . As a dealer for Sofirn I certainly won’t be encouraging it unless it’s a BLF edition and their own separate version. They did this for their SP36 by releasing 3 versions: SP36 - NarsilM, BLF SP36 - Anduril, SP36S - Simple UI. Too many people were struggling with the first versions so they released their own and I like it the most for selling.

I tested my V2 and it didn’t have timed stepdown, turbo lasted 8 minutes and couldn’t be reactivated due to the cell dropping below 3.7V.

No surprise turbo is much shorter as it’s much brighter. My lumen tube shows 2200 lumens on the V2 and 3420 on the V3 at 30 sec. That quite a bit more for the same host

You right, I have SP36 Narsilm. But if I compared Narsilm vs Anduril, Anduril is more user friendly and easy to adjust. For Narsilm every time I need to read manual. For boost driver of V2 is that the efficiency of driver is lower with low input voltages so it can’t regulate higher output power for longer times compared to 3V version of XHP50. Also I think is possible Sofirn to implement stripped Anduril like in first version of Emisars lights with Ramping OS. You compile Anduril with disabled most of it’s feature. I think for most users ramping UI is preferable over stepped one. So if you are distributor and can just ask from clents which one they’ll like most.

Ordered the V3 kit yesterday off of Amazon on the Sofirn discount, received it this morning. Really like the UI.

Only issues I see is ramping is too slow, and in mode levels, the levels are too few. Moon should be included in the cycle, maybe 1 more. I don't see any issue with 5 levels when they switch that quick.

Wish the 5000K was available on Amazon. The 6500K has the classic blue spill, white center, and yellowish corona.

Also wish they'd blink out the voltage level with that blue switch LED on power up - have it do something useful.

The original "v2" appears to be discontinued though, it uses the XHP50.2 6-volt LED and a boost driver, this model is stated as having "2500" lumens.

The updated "v2" still looks similar physically, but it now uses XHP50.2 3-volt LED and appears to use a similar driver as the SP33v3.0, except a bit lower lumens, this model is now listed as "3000" lumens -- it's basically just a slightly lower SP33v3.0.

beamshots comparing the original SP33v2 "2500" lumens with the updated SP33v2 "3000" lumens and also an SP33v3.0 "3500" lumens

(unfortunately not exactly apple-to-apple comparison since the SP33v2 '3000" lumens I got was 5000k tint while the others are 'cool-white' tint.

I have a couple old SP33's that have the 6V XHP50.2, but unmarked version. Think your V2's were unmarked. Was wondering if mine were V1 or V2. Maybe V1 was an entirely different LED?

Btw, the 3500 lumens rating is pretty honest. Use a 30T or 40T and you get a little better. Bypasses on the springs would probably do even better.

I have one of the 3000 lumen v2 sp33s. I can believe the turbo is 3000 lumens, or close to it. What I find strange are the lower settings. Moonlight was stated as 5 lumens in the manual but it is closer to 0.5 lumens. The low, mid, and high modes, which were supposed to be 150, 400, and 1000, respectively, are more like 80 and 250 and 800. It’s not that noticeable of a difference on high compared to other lights claiming 1000 lumens (Fenix UC30 and PD35). But those lower modes definitely are dimmer than other counterparts stating those numbers. Not that that’s a terrible thing, it just seems odd to have actual output on the lower modes be overstated.

The original SP33 v1 "2500" lumens (XHP50.2 6volt LED) will have a tailcap current of 6 Amps on Turbo, no higher. <= this will be a good identifier for the older SP33.

The SP33 v2 (XHP50.2 "3000" lumens) will have around 8-10 Amps on Turbo (tail current from what I remember, will find my notes...)

The SP33 v3 (XHP50.2 "3500" lumens) will have higher than 10 Amps (probably 12-13+) tailcap current on Turbo, based on my measurements with a fully-charged 26650 battery (UT210E clamp meter) (tail current from what I remember, will find my notes)

Also, using the Zak.Wilson's "CeilingBounce" app and measuring the "ceiling bounce" (sorry I don't have a light integrating device as this time), gives the following graphs, which show different thermal control.

Ah, ok. They were all XHP50.2's. Thanx! I had been planning to swap in 3 channel drivers that CK designed here: https://budgetlightforum.com/t/-/55051, but never got around to it.

I believe these boards will work on all SP33 models, appears nothing changed in the basic driver layout, though not sure, haven't confirmed.

Guys, how do you remove stainless steel bezel of SP33? It will be interesting to be modded with Narsim/Anduril based driver.

Didn't try yet - maybe can try this eve.

It simply unscrews from the body. I used an elastic band on the bezel for better grip.

Pressing the bezel against the sole of a shoe while unscrewing should also work. Takes a little force but will come off.

I removed the bezel easily with my sticky gloves.

Interesting fyi, I added 22 AWG wire bypasses on the springs (short ones in the center of the spring), and got ~ 25% boost in lumens and candela.

Measuring amps at the tail, on a charged up 40T the amps went up from 14.7A to 16.5A, but that's only seeing the bump of the bypass on the driver side because my amp measurements are with the tailcap off. I did the before/after tests with both a 40T cell and a red Dolidada 5500 26650 and saw about the same percent bump. These springs are not high performance, accept maybe for their thickness but they are not low resistance.

"nnnn-nnnn" below is "at start - at 30 secs"

Before After

Sam 40T 4390-3890, 32.5 kcd 5340-4790, 40 kcd

Dolidada 3530-3270, 27 kcd 4450-4180, 35.5 kcd

25% bump from bypasses? Wowsers!

We've been get'n spoiled on the high qual springs and double springs in our lights as of late - these are standard springs. I'd be worried about melting them with a good cell without bypasses.

Confirmed - went from 2200 lm in kit spec (kit 26650 and everything out of the box) to 2820 lm with bypass (2x 22awg twisted ends, soldered to spring).

40T was getting 2635; up to 2965 w that bypass.