LED test / review - LatticePower CSP2323 (≈ 6000 K 90 CRI) - LED used in Wurkkos TS10 Mini

LED Test / review EN


LatticePower CSP2323 6000 K 90 CRI


2024-01-22


The very popular Wurkkos TS10 Mini uses relatively unknown emitters, the CSP2323 from LatticePower. As there is currently virtually no reliable information on these LEDs, here is a first test.

The emitters tested here in the cool white version (according to Wurkkos 6000 K 90 CRI) were taken from a red TS10 Mini (purchased from Amazon DE/EU in July 2023).


Technical data


Tj 85 °C, If 700 mA
  • Type: single die, CSP
  • Bin: unknown
  • Color group: unknown ( ≈ 6000 K)
  • CRI: 90
  • Rated voltage: typ 2.85 V V
  • Max. Forward current: 1,400 mA
  • Max. Peak current: 2,000 mA
  • Viewing angle: 120 °
  • Thermal resistance: unknown
  • Max. Temperature Tj: max. 135°C

Datasheet can be downloaded here: Datasheet (newest version, LatticePower)


Design, LES, Performance and Luminance


In contrast to most other emitters, the CSP2323 is very small and flat. This is achieved thanks to the chip scale package design (CSP). In this design, the LED chip is embedded directly in the housing without a substrate and is soldered directly onto circuit boards, which massively reduces costs and size.

According to the definition, the housing in the CSP design may have a maximum surface area of 20 % more than the LED chip, for which the connections for SMD assembly must obviously be connected to the die without bonding.

The CSP2323 are extremely sensitive. Extreme caution is required when reflowing manually, as the yellow phosphor layer may fall off or break off as the temperature rises, or the white housing around the LED chip may simply break. These LEDs should therefore only be handled if this cannot be avoided.



As is usual with CSP LEDs, the footprint is special and does not conform to any commonly used standard. This makes it difficult to use such LEDs for flashlight modding. There is no dedicated thermal pad, which limits the maximum performance as only limited heat can be dissipated from the LED chip into the PCB.

As expected in the CSP design, the illuminated area takes up the majority of the LED package. Light is emitted from the side, which reduces the luminance. The light color of the luminous surface corresponds to the specified CCT.



The luminous area is 3.2 mm² in size. It is not possible to determine the exact size of the illuminated area here because of the light emitted from the side.



Within official parameters, as far as known:

  • at 1,400 mA (official maximum current): 455 lm @ 3.09 V
  • Power at official maximum: 4.32 W
  • Efficiency at 1,400 mA: 105.3 lm/W
  • Maximum reached at 3.2 A, at this point 683 lm @ 3.41 V
  • Power at maximum 10.9 W
  • Efficiency at maximum 62.5 lm/W


The maximum performance is limited due to the lack of a thermal pad. From just 2.4 A, the luminous flux drops significantly within a few seconds, and at around 3.2 A there is no further increase in luminous flux. This is a problem with all LEDs without adequate heat dissipation via a dedicated thermal pad.

Wurkkos specifies 1400 lm from three CSP2323 as the maximum luminous flux for the TS10 mini, which indicates that the LEDs are operated at a maximum current of exactly 1.4 A each. The latter is essential for such LEDs to ensure a long service life. Overcurrent of CSP-based LEDs is only recommended in exceptional cases.

In comparison, it is easy to see how the older XP-E2 has no chance, despite its high bin and cool white CCT with low color rendering. The Vf of the XP-E2 is also very high. The SST-12 has a similar characteristic curve to the CSP2323, despite the dedicated thermal pad. The Nichia 219BT-V1 in the high-CRI version is also older and is also less efficient.



Data for 25 °C Tsp (at 85 °C the luminance values are around 13 % lower).

Due to the light emitted from the side and the already reduced efficiency due to the high color rendering index, the luminance is low. This LED is not suitable for throwers, although LEDs with a dedicated thermal pad are preferable here anyway.


Light quality/beam, tint and spectral data



This beamshot was taken with a Wurkkos TS10 mini brass with 4000 K CSP2323. The TS10 mini has a triple lens without texturing. The beam is more on the floody side. There are very slight color shifts around the flood spot, but these are not noticeable in either the cool white or the neutral white version. The same applies to use with reflectors (OP/SMO).



This CSP2323 has a very high CCT. Wurkkos specifies 6000 K for the cool white version tested here, with over 6500 K being achieved. As stated, the color rendering is very high and, including R9, is at a similar level to that of well-known Nichia emitters. However, the duv is well above 0 and the tint is slightly greenish. Although this is not a problem during normal household and outdoor use, it could be an annoying factor depending on your preferences.


  • Ra: 93
  • R9: 67
  • CCT: 6661 K
  • duv: 0.0033

Conclusion


Due to its design, the CSP2323 is not very powerful and can hardly be overcurrent protected. It is also very sensitive to handle and the special footprint requires corresponding circuit boards. On the other hand, it is extremely compact and cost-effective, the color rendering index is very high and the beam quality is okay for normal flashlight use.


Pro

  • extremely small in size
  • very cost-efficient
  • high color rendering index
  • good beam quality

Neutral

  • slightly greenish tint (sample tested here)

Contra

  • Low overcurrent capability, due to CSP design
  • Very sensitive during handling
  • Special footprint, no compatibility with existing accessories

Thank you for reading the test. :slight_smile:

Greetings, Dominik


v1.0.1
14 Thanks

Thank you always for you time in tests

1 Thank

Thank you very much for the detailed information. This was the information I was googling for last night, and I was glad to find it on taschenlampen-forum.de. I’m grateful you posted here in English so I can read without a translator.

Here, I would like to add to your post by providing links to relevant measurements and photos for comparison.

TS10 4000k measurements by thefreeman:

Measured spot on output level 3
CRI Ra: 97
R9: 96
duv: 0.0014

Here are photos for comparing beams of multiple lights on my hand (including TS10 5800K):

imgur album link to the photos

TS10 5800K is on the top left corner of a tile of 9 photos. Compared to my other high CRI and high R9 LEDs (E21A 5000k R9080 and 519A 5000k R9080), TS10 5800k shows fewer reds. I think your measurement of R9 less than 70 reflects this well."

Awesome thanks. With optical losses do you
Think it might be being driven at more than 1.4A? It’s hard to measure with the glued on tailcap. I had assumed we were looking at 6-7A. Pretty impressive little emitters though

Two points to prove this theory:

First, for such a mass-production light which is sold to non-flashlight-enthusiasts it would be simply stupid to overdrive CSP LEDs (which is in general a stupid idea, like I wrote in my test) as the manufacturer has virtually no influence on how the customer uses the light. The (financial) risk and loss of good image for customers returning their lights due to broken/overheated LEDs is simply too high. Also the junction temp of CSP2323 is 15 deg C lower as most other LEDs, with 135 deg C.

Second, the TS10 mini delivers around 1400 lumens, which eqauls to 1400 / 3 = 467 lm per emitter. This is simply close to my measurement of 455 lm at official maximum current. Since the TS10 mini has slightly lower light flux (mine in brass with 4000 K come to around 1300-1350 lm) this difference can be caused by loss in the TIR lens.

If we assume driver efficiency of 75 %, the TS10 mini would draw around 4.8 to 5.2 amps in Turbo with these 3 emitters. Of course this current cannot be held long since it throttles immediately due to heat.

thanks for all the time you put into testing
do you have a test of the LEDs installed in the TS10?

do you also have test data for the TS10 in the photo? :wink:
(not a bare LED test)

I have a color measurement for this exact LEDs/light.

CCT: 3956 K
Ra: 97
R9: 95
Rg: 99
duv: 0.0004

In general, for light quality these 4000K emitters seems to be better. My brass TS10 Mini does not have any annoying tint, just pure 4000K. (That’s the reason why I bought this TS10 second-hand)

1 Thank

thanks for sharing your tests :wink:

Thanks for the testing! I’ve been curious about the output vs current.

I personally find these little LEDs impressive! I have several TS10 samples and the CCT and duv does vary significantly, especially from the initial batch to the initial Ti version and now to the latest V2 versions.

Set at level 130/150…
With my Opple 3…
CCT has ranged from ~6000k to 6500k.
duv has ranged from -0.0050 to +0.0010. Most are in the -0.0020 range.

The 4000k version has had a similar variation.
CCT from 4000k to 4500k.
duv from -0.0050 to 0.0010 with an average closer to -0.0010.

Two 6000k and one 3000k mix gives a near perfect (for me) CCT and duv.
CCT: 4800k
duv: -0.0035

FB

2 Thanks

I was always under the impression we lost about 30% of the light out the front as it were. That’s for a reflector. As I understand it, optics are worse. And don’t almost all flashlight companies overdrive their emitters? I’ve never seen one under spec. I guess the csp style must be very sensitive. Without amp testing we won’t know though. My 4000k with included cell does 1250 lumens on turbo. Again, awesome test thanks. These are more efficient than I expected

Edit. So doing more studying on tir tech, they’re basically 100% efficient. Whereas a metalized surface, like a reflector, is only 80% efficient. And a non coated glass lens causes another 5-10% losses. So the ts10 could be over 90% efficient from led output to flashlight output. Nice.

Thanks a lot for sharing these tests, the results are pretty impressive for the 6000k.

Not many 6000k R9050 emitters with good tint, even if these have a low power ceiling.

AFAIK they can work on E21A MCPCBs (and the other way round), the footprint is unique but sort of interchangeable.

That’s true, but I think also because these CSP2323 are some sort of E21A clone (although a decent one). Tbf Lumileds also has a CSP emitter called HL2Z in the same size and with same footprint. It seems these CSP emitters are more or less the same, only the available CCT and CRI ratings are different.

No, I have some lights which not overdriving the emitters (Ledlenser, Fenix, Thrunite, Nitecore)

1 Thank

Would be interesting to see a comparison between the E21A, CSP2323 and HL2Z.
I’d never heard of the HL2Z but hopefully they’re not green, like the HL2X.

Maybe also the Cree XQ-E Plus HI, they’re 1616 but have a decently high max current, could be a throwier option. Probably would also fit in the same footprint too. It’s a neat little flip-chip emitter, almost like a miniature XP-L

I don’t think so, 2018 I randomly tested an domed XQ-E HD emitter (bin R2 5000 K), which was interestingly enough destroyed at only 1.4 A (burnt dome). Of course this could be just a bad sample, but I am afraid that this is still a problem with these emitters since these are not equipped with a dedicated thermal pad…

Samsung LH151B are the same size.

Datasheet

I swapped three of them in 2700k 90 cri into one of my TS10. They took the current just fine. I have nothing to measure them with (other than my green-hating eyes), but they’re certainly positive duv, unfortunately. I used minus green to help out.

Here’s a digikey search for in-stock 1616 LEDs, link The pics of these on digikey are distorted and they look rectangular, but they are actually square.

Edit: I got confused with other things I ordered. I must have used LH181B. Datasheet

The flip-chip version seems promising, especially the domeless model.

Comparing the real current limit between lateral and flipchip versions of emitters they mightve made some improvements elsewhere too. Like the XP-L HI or the new XM-L2.

I was thinking of putting together a TS10 using them, but found that 3535 mcpcbs were available and decided another route.

I have some of XQ-E HI 4500 K (old gen, bought 2018) indeed. Interesting, I don’t know of ordering some XQ-E HI - maybe a mistake, but at least a good one :smiley:

Maybe I will test it (in a very short test without much text and formatting)

We are talking about 2323 leds, 1616 is a lot smaller with less light flux (due to smaller LES and less heat dissipation)
But there a lot of them around, even Cree has some, with the XD16 (and XD16 Premium white, whatever that means).

For this I had to buy some, but for now I did not plan to order LEDs at Mouser (at least I have to pay 50 euros to get over the free shipping threshold), since I have still a backlog…

1 Thank

Very interested in the results, it only to compare against the flip-chip version.

If the flip-chip versions are anything good, a TS10 mule with 14x of the domed version (and a high flux bin) could have quite a lot of output, maybe even a record for a 14500 light (2000lm maybe?)

I still don’t have the new gen XQ-E HI here.

*When that happens, I meant. Still an interesting small-size emitter though

1 Thank