LED test / review - Luminus SFT-70X-W (4000 K, typ. 95 CRI) - good tint and beam, but destroyed in test, better avoid currents over 4.5-5 A

Flashlight Modding and DIY Parts LED Testing
1

LED test / review EN


Luminus SFT-70X-W 4000 K 95 CRI


2026-10-25


UPDATE: Close-up images of the disassembled LED chip can be found here!


  • Type: lateral, quad-die
  • Bin: G3
  • Color group: FA3 (4000 K)
  • CRI: typ. 95

This LED tested here was bought from Kaidomain in 06/2026. It is labeled as “G3 FA3” for binning.

SFT-70 tested earlier had a metal frame around the LES, which holds the white silicon molding. This was changed to a white (silicone?) frame in this version, likely to save costs. This indicates a slightly changed production process.

There are bonding wires visible, although these are completely by white silicone. Bonding wires are likely to cause higher Vf compared to LEDs in flip chip design.

The SFT-70 fits on standard 5050 boards, like the earlier variants. With changing the solder pad layout, a 12 V config is possible. On standard boards, only 6 V is possible.



(Please note: the image of LES was taken from the LED after testing up to 5 amps where I interrupt testing to assess damage. There is a small burned area which was not there before testing. So no damaged or dirty phosphor was already there right from the factory.)


  • Maximum reached at 6.6 A, at this point 2118 lm @ 6.66 V
  • Power at maximum 44.0 W
  • Efficiency at maximum 48 lm/W



The SFT-70 4000 K 95 CRI tested here performs way worse than the other SFT-70 I tested, especially in terms of maximum current. With only 6.6 Amps the current is much lower than for cool and warm white SFT-70 alike, which topped out at around 9-10 Amps.

The slightly higher efficacy compared to the 3000 K variant is expected, since the binning of the 3000 K is even better (G4 compared to G3, although I don’t know how consistent the batches of these warm white SFT-70 are since May 2024).

Since the LES is exactly the same, luminance is also basically identical to the 3000 K variant, maybe some percent higher.


More problematic is the destruction of the emitter tested here. At around 5 Amps it started smoking, at maximum current of 6.6 Amps it smoked even more, and at around 7.5 to 8 Amps is was burned. The phosphor mix burned, which is “clearly” visible.

I would not recommend to drive the SFT-70 4000 K 95 CRI with more than 4.5 Amps. Maybe my sample was faulty in terms of thermal resistance (was the thermal image with the huge difference between chip temperature might suggest - 225 and 168 °C at 6 Amps is way too hot), but since I only one more sample left which is intended for another emitter swap I cannot confirm this.



The tint is very neutral and matches exactly the stated CCT. The CRI is very high.

There is a multi-phosphor mix used which fills the cyan gap quite well, which is likely done with BSON phosphor. Maybe also a mix of LED chips with slightly different peak wavelengths or modified chip is used.

From BSON aside, LuAG, YAG:Gd (maybe blend with SSN) and SCASN phopshors are likely used as well. The tint shift is low and is almost not visible, which is a huge improvement from earlier SFT-70 with very clearly visible tint shift.

  • Ra: 95.5
  • R9: 97
  • CCT: 3992 K
  • duv: -0.00169

The light quality is great, beam profile is good, and it fits on standard 5050 footprints (6 V only). But it’s performance is not as good, especially considering the low maximum current and the risk of burning above 5 Amps.

Keep in mind, that I was only able to test one sample, maybe there are better performing SFT-70 from other or same batches out there.

Apart from that, it is a nice emitter for using in flashlights.

Thank you for reading this test. :slight_smile:

Greetings, Dominik

v1.0.2
18 Thanks
2

Thank you as always for this extremely comprehensive test!

The lower power handling is surprising, especially since one would normally associate higher conversion efficiency with higher CCT. Perhaps this particular phosphor mixture trades performance for color quality more than the usual high CRI emitter, and it really shows in the spectrum, as reduced cyan dip.

Do you happen to have the R12 (deep blue) reading?

3

Except the Rf isn’t really good, especially compared to the old 3000K SFT40/70. Out of my handful of Ra95 4000K SFT70, the highest Rf I’ve measured was 92.

4

From what I’ve seen over the years, lower CCTs like 3000K generally have an easier time achieving high CRI readings, probably because the most problematic blue region is underrepresented. So I’m not surprised that the 4000K reads worse.

I’m curious: do you find Rf and CRI subjectively good measures of color quality, in the sense that higher CRI/Rf always corresponds to greater visual similarity to the reference source? I’m asking because the SFT70 3000K tested significantly lower in both CRI and Rf (94, 92) compared to the SFT40 (98, 96), but with every sample I’ve tried, the SFT70 has subjectively greater visual similarity to a halogen lamp compared to SFT40. This holds for every color other than dark indigo-blue.

Colorimetric comparisons among spectrometers (taken from reddit)
5

Every cooler CCT (≥4000K) Nichia I’ve ever measured is capable of Rf ≥94, even Simon’s 4000K NV1010, 4000K HT35, 3700/4000K FFL351A etc. have higher Rf than these new 4000K SFT70. I know Bob has some of these new SFT70 in 3000K, and if I’m remembering correctly, they’re also worse than the older ones.

Personally I really don’t look at Ra at all anymore, and give it very little value. I’m not sure why koef3’s 3000K SFT70 measured so much lower than the SFT40, but all of mine have Rf 96.

1 Thank
6

Unfortunately not, I didn’t save the measurement before starting the performance test :confused:

I tested the 3000 K SFT-70 over two years ago. (Man, time flies :open_mouth: )
I also used a simpler (and maybe more inaccurate) spectrometer. But still I got Rf 92 on my sample.

3 Thanks
7

Since the SFT-70 4000 K tested here was destroyed I wanted to take a closer look on the LED chip being used. (Keep in mind this could be different to older SFT-70, which I haven’t disassembled yet, mainly due to lack of samples.)

The LED chip looks very interesting. Instead of being an array of four individual chips arranged in a 2x2 pattern (which could be expected), it is one single LED chip instead, which is visible on the monolithic design on the outer edges, especially where the bonding connections/solder pads are.

There are four independent connected junction areas on a single sapphire (or silicium carbide) substrate, which are electrically seperated by the thin lines in the middle.

this type of chips seems to be quite rare since most multi-die emitters (XHP50, LHP531, Daydream and many more) are built as array of independent small LED chips, like shown in this disassembly of the LHP531 (look here in the test). I never saw this type of LED chip before.

Maybe Luminus/Sanan choose this design to eliminate the gap between the chips since it is basically non-existant on the SFT-70 which improves the beam a lot even in SMO reflectors.

4 Thanks
8

Wow, it is so unexpected that this new sft70 is a single die emitter. So a smaller LES for a little higher lumen density, compared with other high cri emitter

1 Thank
9

Thanks for the awesome test!

is the 3000k version the same generation with the white frame or is the 3000k version with the gold frame?

or

10

I’m unsure if one should classify this as single-die, or just single-chip. For me, a white LED cannot be single-die without also being 3V; having a voltage of 6V indicates serial configuration of at least 2 dies.

I think the throw increase that comes from reduced die gaps is less due to improved peak intensity, more due to the ability to use a non-diffusive optic without getting a horrendous beam.

I believe it’s the golden frame generation previously tested–the spectrogram looks identical.

11

This is an unusual configuration, and it might sound wrong to call it a single die LED, but a processor can have multiple cores in a single monolithic die, so i can see a LED having multiple diodes in a single monolithic die. It might be just a terminology problem.

1 Thank
12

My one 4000k SFT-70 (in a Convoy M21H) has pretty low R9 (peaks in the 60’s, but in some modes is as low as 39) and peak Rf is just 90.

Fortunately the tint is really nice!

13

I would call it is single die but multi-chip (if “chip” is used as term for the p-n junctions which are individually connected).
The comparison with a monolithic multi core processor is more fitting.

It is the old version with the golden metal frame. (I think two and a half years ago the newer version without the golden frame wasn’t available anyway, I think it started to appear at around last year or so)

2 Thanks
14

Interesting stuff! I wonder if the changes in the die setup is mainly for cost saving, or thinner birders or something else.

I got 4 of these from KD a few days ago, but won’t be able to test them with my Hopoo until the weekend.

15

I can confirm there’s a huge difference in beam quality in difficult lights compared to the old version. I was modding Zebralight SC600s with the old 3000K briefly, but they made a minor change to the reflector design that resulted in an obvious and completely uncorrectable donut hole in all lights manufactured since then. I was very pleasantly surprised to see the new version SFT-70s have perfect beam quality in the same reflectors.

1 Thank
16

Maybe I am able to find an old SFT-70 somewhere to disassemble it. Now I am curios on how the LED chip changed.

From my test of the low CRI cool white SFT-70 (the version with metal frame) from August 2023 the LED chip looks basically identical to the one of the new version. Maybe the height of the LED was slightly changed in the new version, or the metal frame has changed the optical properties (viewing angle) a little bit.

1 Thank