Nichia Optisolis 2700-6500K color and CRI testing

Update 21.7.2018: More color temperatures tested in post #7.

Clemence sent me the Nichia Optisolis in 3000K and 5000K flavor for color and CRI testing.

The exact model numbers are:

There was four samples of both and they came already reflowed on Virence VR16S1 DTP MCPCBs. Each board had two emitters wired in parallel. The other pair will continue its journey to Koef3 for further testing. Also check out Djozz’s output test here.

The emitters are specified for up to 150mA (3000K) and 100mA (5000K) continuous current, but according to Djozz they seem to handle larger currents fine.

CRI and color testing was done at Nichia’s recommended current of 65mA (per emitter). For the tint vs. current test I ran my samples up to 200mA.

All tests were done with bare emitters with no reflectors or optics used.


There’s practically no tint shift at all. Not with varying current or measurement angle. The tint is also excellent landing on or near the BBL.

The point of the Optisolis is color accuracy, which is excellent. I measured CRI (Ra) at 97 (3000K) and 99 (5000K) and R9 at 94 (3000K) and 93 (5000K). CRI also doesn’t change with current. The 3000K version seems to over saturate colors according to its TM-30-15 Rg value of 104. The 5000K one is more faithful at Rg 99.

Data for 3000K sample #2

Data for 5000K sample #2

The two different samples are very consistent. There’s very little difference in the spectrum between them.

CRI Data at different currents (per emitter)



Forward voltage up to 400mA (200mA per emitter).

The Vf is about 0.1V higher on the 5000K model.

1 Thank

Very nice color and tint, and extremely good CRI which are equal or even better than the Ledengin Gallery-white 3000 K (stated CRI 98) emitters! Well done Nichia!

Thanks for your testing, I'm very excited to see this LEDs and it's color for myself! Extreme testing would follow if I got some of these.

Excellent review. This LED is really something else. And some great measurement and testing hardware as well :sunglasses:

Well done Maukka, great test and setup. I’m drooling at those Rigol (can’t afford those of HKJ’s :disappointed: )
Your results matched the results Nichia sent me earlier. 200mA continuous use seems fine. It’s been almost two months they’re inside my light bulb with almost 24 hours daily use.


Naw, not me. I’d take care of them and love them and give them a good home.

Thanks for testing.

Hmm, 3000K use blue ~440nm emitter, but 5000K use ~420nm violet emitter.

from datasheet :
This product uses a unique red phosphor that has a slow response time; when compared with the blue die and green phosphor in the product, the red will fluoresce with a delayed pulse of about 5msec and have an afterglow for about 5msec after the blue die
dims. If the LEDs are operated with a pulse current, ensure there are.

Got some more Nichia Optisolis, this time the whole shebang: 2700-3000-3500-4000-5000-6500 K.

All the measurements can be found here:

Nichia Optisolis 2700-3000-3500-4000-5000-6500K - Album on Imgur

The tint and CRI on these is amazing. The 5000 K is truly a daylight simulator with practically perfect color rendering. You can also see that the peak on the 5000 K and 6500 K emitters is significantly lower at the purple wavelength. The dip above it is also much shallower than the typical cyan dip in a LED spectrum.

Just look at the spectrum!

Here’s a 4000 K with the low peak at a typical higher wavelength and a clear dip above it which reduces the color rendering indices just a bit. But in turn their Rg value is above 100 which means they oversaturate some color a bit, which might be preferable to some.

I also tested the output on the 3000 K and 5000 K variants up to 400 mA. Somewhat surprisingly the 3000 K is more efficient on lower currents but as the current is increased the 5000 K flies past. Forward voltage is higher on the 5000 K one as expected. I found it also amusing that I happened to get a perfect average 3000 K emitter, since its output was bang on 24.1 lumens at 65 mA just like Nichia specifies.

The tint was only tested with bare emitters. There’s barely any tint shift as can be seen from the OP, but Clemence tells me that these don’t work very well in clear optics as a quad at least but a frosted one is fine. It’s could also be quite difficult to find a suitable reflector. My go to Pineapple OP one did exhibit clear shifts in the hotspot. However, a beaded TIR optic looked excellent.

Thank you so much Maukka for such invaluable data. Too bad the 5000k doesnt have perfect duv or slightly negative duv. The hint of green kills it for me.

I guess I’m not the only one skipping the 2700K, 3500K, 4000K, and 6500K spectrum tests. I waited and kept waiting for the other tests. Until Maukka sent me an email and told me it was there already…. :person_facepalming:
Very nice Maukka. Big thanks. I hope your Carclo 10623 works with the quads.

- Clemence

As Clemence posted on his sales thread, the L757 (2700-4000 K) can’t be installed in parallel with the W757 (5000-6500 K) series because the forward voltage difference is too high. Here’s a comparison:

I’m so happy I have automated this. Would have taken ages to put down those numbers manually.

Yup, I know how it feels inputting those numbers. I did exactly the same 100mA increments too……manually. Does your Rigol comes with PC UI? Or did you build an add-on DAC hardware? Really like those sweep automation.

- Clemence

The Rigol has some software from the manufacturer but it’s utter crap. This Excel file is just using some modified VISA VBA examples from Keysight IIRC which interfaces with the PSU over ethernet. The same thing is much easier to do in Python, which I use when doing the output measurements in the sphere, since it’s simple to integrate the spectrometer executables. Everything just gets written in a CSV file like this:


Where Y is the lux reading from the meter. This has to be converted to lumens of course, but that happens automatically when I paste the data into Excel.

Thanks for the info! Now I know testing multiple LEDs is not that hard for you. More incoming Nichias for you then…. :wink:
Slowly but sure, you’ve becoming the top one stop tester for us in WWW.

BTW, do you think it’s possible to spectra testing a male firefly? I’m serious here…

- Clemence

I have done everything manually in my led-testing years so I’m glad for you too that you automated it :slight_smile:

My new spectrum-toy (a Mavolux Base) needs a minimum of 10 lux on the (7mm diameter) sensor to produce a spectrum, you may get that if you tape the firefly on top of the sensor :partying_face: . I do not know about maukka’s setup?

A hotel manager asked me if I can find a suitable LED to mimic firefly in the hotel garden. He wants it to be as realistic as possible. Weird request but very interesting! Finding/programming a driver with random natural blinking is another challenge.

We can place several fireflies in a small light bulb diffuser in front of the sensor. But I don’t know how Maukka would get those creatures in Finland.

- Clemence

I found this article from 1964

Wow Djozz, very informative document you got there! Big thanks. I’ll try to find the suitable LEDs for it. Different species different spectral distributions too. 500nm - 700nm seems to be the range. But the peaks are species dependant. Now I’m thinking to lure fireflies……

- Clemence

It looks like a phosfor converted green has about correct width of the emission peak but has the wrong peak wavelength (520nm for the new E21A converted green). The Rebel PC-lime however has its peak at 545nm which comes close to some of the firefly species tested in the 1964 article.

PC lime… interesting….
Let’s see if the next E17A green samples I will be receiving looks good. Want some Djozz? :wink:

- Clemence

Of course I’d love some. Brand new leds, hmmmm :heart_eyes:

A quickndirty spectrum of my flaslight with Rebel PC-lime: