I see. I take it you get an exact performance value of a bin, or sub-bin, with each delivery? And that should also specify its behaviour in relation to forward voltage and temperature? In such case an honest calculation should come somewhat close, isn't it?
You been previously hinting that even sphere measurements aren't 100% exact. What exactly happens here, and how big could a distortion occur?
Hi mate, this post will be longer than usual, so please bear with me . . .
Yes, when we order Nichia diodes we get bin ranges within the upper or lower half or quartile. For example, when we were negotiating with Nichia to purchase their new Vitasolis range a couple of years ago we had them test a bunch of production samples to see the average output of what we would be getting.
The diode we wanted was NF2757GT-V3F1 Rvg0b – Vitasolis in 3000K. We were flexible with the colour bin, as colour accuracy is not as important (and can be a small advanatge in some respects, as it broads the spectrum). But we needed matching voltage bins and we wanted the highest efficiency production bins we could source.
This is not as easy as simply saying "Give us the top bin". You must compete with other companies for those bins, while some bins cannot be guaranteed for larger orders as there is simply not enough yield. This means that if you have 4 bins (hypothetical) and in every batch of LEDs the yield is spread: Bin1 = 20%, Bin2 = 40%, Bin 3 = 30%, Bin4 = 10% then you can see there is only 10% of production to go around for the top bin, while it is easier to be guaranteed lower bins – especially Bin 2 and lower, which makes up 60% of production.
If everyone wants Samsung LM301H diodes in 3000K, then everyone is competing for the top bins and there are fewer to go around.
However, we use Nichia diodes, so there is less competition. Plus, we have a long relationship with Nichia so they look after us. Plus, Nichia diodes are better than Samsung. For us, it's an easy decision to use them. Samsung are good at marketing, but those in the industry know that Nichia is the better company.
Getting back to our first Nichia Vitasolis order . . . The datasheet states that the top efficiancy bin in 3000K is P12 and that P12 is 36-42.8lm. That is the top, top bin and only a few percent of all diodes manufactured will be in this bin. We wanted a realistic figure of what to expect. So Nichia tested a batch for us and provided the results:
Nichia said:
Attached is the random sampling data for the Vitasolis LED for each lot of the LED. Generally all LED are at the rank of P11d43 as I saw from the sampling data, it is over 33 lumen all of them. I roughly calculate the brightness rage for each sub ranks as below, general guidelines:
P11d41 = 30.3-31.725lm
P11d42 = 31.725-33.15lm
P11d43 = 33.15-34.575lm
P11d44 = 34.575-36lm
The "d" designation is a quartile ranking. d41 = 1st quartile; d42 = 2nd quartile etc
We ended up with P11d43. It is important to remember that this is the MINIMUM efficiency ranking in each order. That means, NO diodes will be below P11d43 (less than 33.15lm) and that all of them will be at or above this range, including P12.
We paid a premium for these LEDs, and this was the highest efficiency bin that Nichia could promise us for the quantities that we needed in the matching voltage bins that we needed (within 0.1V). If we required larger quantities, then of course we would have to accept lower bins to make up the numbers.
Rememer also that Samsung and Nichia still have to find customers for those lower ranked bins – they can't just sell the top-ranked bins.
What do you think happens to those lower ranked bins?
In the end, our order had the designtaors: sm303, sm3050a, sm3050b, sm3050c, sm3050d, sm3050e, sm3050f/P11d43/K22/Rvg0b
sm3XXX = 3000K CCT bin
P11d43 = luminous flux bin (33.15+lm)
K22 = voltage bin (2.7-2.8V)
Rvg0b = overall CCT (3000K)
I hope that is not too much information for you!
