No mod that I have ever done was straightforward, there’s always something unexpected that has to be taken care of (i.e. the tiny screws were tight, I had to use a very well fitting screwdriver to not screw them up). But if you have dome some modding before and do the disassemby and assembly a bit gentle (it is tiny inside), there are no big problems.
The led leads are very short (1-2mm). I had to reposition the led because it is attached only by soldering points. Make sure it’s soldered deep enough to close the housing tight. I was holding the led in right position with one hand while the other was holding soldering iron.
As far as the screws are concerned, expect the loctite on the unloosen screws. That is probably #1 reason why it’s so difficult to unscrew for the first time.
I had a look at the position of the stock led before unsoldering it. The leads of the new led I bended sharply to the side very close to the plastic and cut off alongside the plastic. This gave enough “solder length” to attach the led just as well as stock led. I always pre-tin the leads of the led and then fuse the solder with the blob on the driver board.
Thank you gentlemen. Seems doable
Here’s a question. I can’t remember if anyone asked it yet, but I think not. What would the result be if I tint mixed these high CRI, low CCT emitters with some low CRI, high CCT cheapies? For instance, if I replace five of the emitters in a cheap 9-up light with these emitters, leaving four of the stock ones, which are probably 70 CRI, 6500K CCT, what could I expect? I know the resulting CCT could be calculated easily, but I’m wondering what the resulting CRI would be. Does CRI “blend” linearly like CCT does, or not?
It is very likely that the 9 leds cheapies use leds with a CRI far lower than 70 (i have some at home will do the test tonight)
The spectra are simply averaged, so the cyan dip may be halved, and far red too, I guess the CRI may be in the middle too.
One thing to note here is that as far as cheapo 5mm LEDs go, you might end up having the stock LED with 10000k or colder and godawful tint. There are no minimum to the quality controls there.
I am new to the CRI things and my setup (Colormunki Photo + ArgyllCMS) might be poorly missused but from what i see on my cheap lights they range from a CCT 11000K / CRI 74 (R9 -20) for the better to something that Argyll can't even understand telling me something like 999911K / CRI -1 for most of the others.
My guess is that the blue peak is so high compared to any other color that the calculation fails
As mentioned above, intermediate CCT and CRI. Duv will be lower than either emitter, which is a benefit.
The cheapo LED’s I have tried have a higher forward voltage than the Yuji’s, which means also than these replacements. The result will be your high CRI emitters are driven harder than your low CRI emitters. It might not be enough to cause a problem though.
The cheapo LED’s are also usually a ~15 degree beam. The much tighter beam will mean major tint deviation across the beam. It might be possible to resolve by sanding the dome of the cheap LED’s.
I have mixed equally numbers of 3200K and 5600K Yuji’s with similar beam angle and forward voltage in a test, and the result was really good, around 4500K. I only tested it a limited amount, but I actually liked it slightly more than the only Nichia 219B, SW45K light that I own.
Thanks for the replies. I knew the cheapies were cold, but 11,000K? WOW!
So, if I tint mixed like I suggested above, I would still end up with a very cold light, at a bit over 6700K average? Yikes!
Well, that might work in my favor, actually. Since djozz tested these LEDs at up to 100mA, they should do okay, and turn the overall light output a bit warmer than if all the emitters were putting out equally. Maybe, it might even get into the 4500K - 5000K range overall.
That would be a problem. I guess I could solve that with a bit of DC-Fix, or by sanding the domes like you suggested. Or both.
As a fair warning, 11000k is not the upper limit of how bad these can be.
The “6500k” cheapie samples I know was pretty darn cold 50000k. Djozz has tested it. Ewww.
You cannot expect any sort of consistencies as far as the el cheapos are concerned.
50000K? :-P Is that possible without the stuff emitting some nasty UV frequencies? Also, that cannot look very good.
50000k corresponds to X,Y of (0.227,0.228) which is not that far from 20000k (0.250,0.250).
Once you slip past 10000k in terms of x-y, it’s seem really easy to slip into 6-digit figures.
I have my theory that the light mixes by averaging the CIE x-y coordinates.
10000k on BBL: 0.280, 0.290
3000k on BBL: 0.431, 0.404
The resulting light will be: 0.358, 0.347
That corresponds to ~4500k and –0.0073 duv
If that works out the way I theorized, then you might want to throw in a few (and only just a few) green LEDs to pull the duv back closer to the BBL.
10000k (DL locus): 0.278, 0.293
3000k on BBL: 0.431, 0.404
The resulting light will be: 0.354, 0.349
That corresponds to ~4600k and –0.0048 duv
Ulrich - RO247700139TH
EricJ2190 - RO247700142TH
TBone - RR278755013TH
So, the CCT isn’t an exact average of 3000K and 10,000K, but an average of the underlying positional numbers? And those don’t scale linearly? That’s interesting. I really DON’T know much about light and/or color!
CIE coordinate itself is pretty linear.
It’s the CCT that is non-linear. You can look it up and see that the warm white part of the locus takes about 1/3 of the chart’s span. And the BBL locus is curved.
In all honesty, mixing unknown emitters is a bad questionable idea. Of course can be done, but you'll have to deal with a poorly predictable resulting tint and CRI mishmash. One of the most important factors to consider, if setting emitters in parallel, is you won't exactly know how they will share the driving current. This affects the amount of output from each emitter, and therefore CCT, tint and CRI. It will also affect the relative speed at which each emitter will age, and thus CCT, tint and CRI will also change with use.
When actual emitter tests are available you can at least know what to expect, i.e. you can more or less predict the final outcome.
For example, setting a LH351B 3000K CRI90 in parallel with an (MF) LH351C 5000K CRI90 at ≈1.5A would result in a 0.6A/0.9A current share given their Vf curves (planning to give this a shot sometime).