Introducing CobKits.com - specializing in DIY and Citizen COBs

Dreddd

Well-Known Member
Would you be so kind as to run these awesome tests on the same chips @ 90 CRI?

btw got my 10 x cxm-22's/bjb's, thanks cobkits!
 

caretak3r

Well-Known Member
@CobKits -Thanks for throwing the v22 into the mix. I was itching to see how well it did compared to the others. Do you still think the 1212gen6 is the way to go (vs the cxm22) for < 50W?
 

CobKits

Well-Known Member
depends on your budget, your space, what you have setup for drivers, etc. cxm isnt that much more expensive than a 1212, so once youre into heatsinks drivers etc it could be as little as $20 on top of your build for a lifetime of better efficiency. but there are applicatipons where more 1212s fit your space better, and when run super soft (<30W) most of the efficiacy differences are negligible.

i'd prob rec:
1212 >35W
cxm22 30-75W
1825 above 75W

in their range they all scale nicely in $/ppfd at a given efficiacy level, so it depends on your heatsinks, number of emitters, etc.

1212 would probably be best for those trying to emulate the efficacy of a quantum board, in an application where youre using extrusions or aluminum sheet to throw down a bunch of chips and run them under 25W
 

CobKits

Well-Known Member
Would you be so kind as to run these awesome tests on the same chips @ 90 CRI?

btw got my 10 x cxm-22's/bjb's, thanks cobkits!
i will be shooting all the spectrums i can, but not necessarily doing comparative output tests on all these. Spectrum is absolute, unlike relative output, so i hope to get us a good catalog of SPDs to work with

the only downside is that with the intensity in the sphere it saturates my meter, so ive been shooting them at low wattage (200 mA on 50V chips and 300 mA on 36V chips - those are the points that fall right about on the 10W mark in the graph. i know that spectrum can shift with temperature as well as current, id love to hear some suggestions on if i can improve this @Rahz @Malocan @Abiqua. For now im just trying to keep it somewhat standard and not saturate my meter and get goofy results

particularly as we get into 90 cri chips. how do we reconcile the overall PAR measurement that will be lower by the steeply filtered Li-Cor discarding the wavelengths above 700? is there some correction factor that needs to be derived to get a more accurate QER/YPF?
 

coreywebster

Well-Known Member
Cobkits when will you do the gen6 1825 test? Plus I noticed the data sheet on the gen6 1212 is 1211 as far as diodes, does that mean its now called a 1211 or will it still be referred to as a 1212 gen 6?
 

BuddyColas

Well-Known Member
another round of sphere tests. lets call this preliminary data as tests done on different days give different results due to variations in ambient temp.(shop is a warehouse with big rollup doors) Im getting antsy because it takes longer to mount the chips for testing than it does to run the test and i have all these fine specimens sitting here ;)

Today was more normal temps but the chips i ran the other day was a much colder day. i want to say -5C difference in ambient temps

"cold day" tests (will bias results slightly higher):
1216
1818
CXB3590 "A"

"warmer day" tests:
Vero 22C
Luminus CXM22
Luminus CLM22 (the alleged 36V counterpart of the CXM22, ~$1 cheaper)
CXB3590 "B" (same chip as "A" tested on different day)

View attachment 3922620

observations:
-wow CXM22 dominating across the board at all currents. this chip was new to me. I dont stock 4000k 80 (tho i initially screened that color from a pre-production sample - this was a regular off the shelf chip from digikey). 1818 is catching up with it at 130W but probably not really as 1818 was a cold day test

-I was interested in Vero22 as it is the same package size as the CLU048 and CXM22, it does well at high currents but lacks at low currents for some reason. On a same-day test its crossing CXB3590DB above 50W but still lags behind the similarly priced CXM22 and also the CLU048-1818. Definitely living up to Vero's reputation as an excellent high current chip with great thermal characteristics

-CXB3590 day 1 and day 2 tests were similar trends, and you can see that ambient temp difference definitely having an effect (selling point for ducted hoods? ;)). Again, you can see why i dont like to compare chips tested on different days, were seeing a 1.5-2% difference in relative output

at 50W were seeing a whopping 7% difference between top and bottom chips - so really any of these chips will do fine... chip cost really comes into play at that point. I would definitely say all these chips are in the same "class" of package size and relative die count. It will be interesting to see what happens when we bring out the big guns with larger packages and higher die count. The chips on deck include gen 6 CLU058-1825 and CLU058-3618, as well as Vero29 V7 and Luminus CXM27 and CXM32.

will also be running 1212 gen6 mostly in comparison to the similar 1216. i still have a warm spot for these chips as economical low-current champs.

Once we get to the formal test I hope to run dupes on most of these to check for output/bin variation, i think I have 2 of all of the chips except for CXM27 and CXM32

I am happy with the sphere performance, it seems to be giving more reliable low current data than the open-air tests
So let me see if I got this straight. Would the chart say at 50 watts a V22 is 4-5% less efficient photon generator than a CXM22?
 

CobKits

Well-Known Member
Using the passive heatsinks from you should I clean up the area where the cob is mounted down to bare aluminum?
no! anodizing is a surface treatemnt (its actually an etch and subsurface treatment, the "color" goes into pores in the metal

those heatsinks are flat and a good surface for thermal paste. too smooth/polished is actually worse for thermal compound
 

CobKits

Well-Known Member
So let me see if I got this straight. Would the chart say at 50 watts a V22 is 4-5% less efficient photon generator than a CXM22?
zooming in:

upload_2017-4-11_13-15-43.png

42.5/40.7 = 1.044 -> 4.4%

again thats why i want to try to test 2 chips from each model, we could have pulled an above-average cxm22 vs a below average V22. 2 or more chips are needed to really make a conclusion. im familiar with the CXM22 and that looks consistent with past testing - as for the vero- thats the first V22 ive laid my hands on (i have another as i bought a pair), so i had no idea what to expect. per datasheet luminus says +/-6% on flux, vero says +/- 7%

in my experience (for citi and luminus and the limited number of vero ive tested at least) - they are much tighter, +/- 1-2%

like look at these two separate Vero29Cs i tested (this was open air @ 12", not sphere):
upload_2017-4-11_13-27-2.png
 
Last edited:

sixstring2112

Well-Known Member
i will be shooting all the spectrums i can, but not necessarily doing comparative output tests on all these. Spectrum is absolute, unlike relative output, so i hope to get us a good catalog of SPDs to work with

the only downside is that with the intensity in the sphere it saturates my meter, so ive been shooting them at low wattage (200 mA on 50V chips and 300 mA on 36V chips - those are the points that fall right about on the 10W mark in the graph. i know that spectrum can shift with temperature as well as current, id love to hear some suggestions on if i can improve this @Rahz @Malocan @Abiqua. For now im just trying to keep it somewhat standard and not saturate my meter and get goofy results

particularly as we get into 90 cri chips. how do we reconcile the overall PAR measurement that will be lower by the steeply filtered Li-Cor discarding the wavelengths above 700? is there some correction factor that needs to be derived to get a more accurate QER/YPF?
I know my names not on the list but.....
What about mounting each cob in the sphere then turn it on at say 1000ma for 15 to 30 minutes to really get that heatsink up to temp,then back it down to your test current (300ma or whatever it is) and then shoot the tests with the licor.never been a fan a flash testing on cold heatsinks because in our rooms they are almost never cold after 20 minutes. 75w cree on a 140mm x 80mm sink is always around 122f in my room with 80f as the ambient.think my 35w cree run about 105f on the same sink.
That and i would love to see the vero 29 gen 7 68v cob included in those graphs if its not too much trouble.
What color temp were those last set of sphere test?
 

caretak3r

Well-Known Member
zooming in:

View attachment 3922912

42.5/40.7 = 1.044 -> 4.4%

again thats why i want to try to test 2 chips from each model, we could have pulled an above-average cxm22 vs a below average V22. 2 or more chips are needed to really make a conclusion. im familiar with the CXM22 and that looks consistent with past testing - as for the vero- thats the first V22 ive laid my hands on (i have another as i bought a pair), so i had no idea what to expect. per datasheet luminus says +/-6% on flux, vero says +/- 7%

in my experience (for citi and luminus and the limited number of vero ive tested at least) - they are much tighter, +/- 1-2%

like look at these two separate Vero29Cs i tested (this was open air @ 12", not sphere):
View attachment 3922922
I really appreciate all of the testing you do.
 

OLD MOTHER SATIVA

Well-Known Member
re CXM22

i attached 4 of them to an aluminum bar using the[for 52w] Mingfa[affordable] heatsinks i had used on the Nichiaj130b @33w

[they are pretty darn good..gonna take the refekectors off em]

the CXM22 were running at 74w and were bit hot..so was the driver..seeing as its a hlg-240-48 and running at 295w

with air movement they are in the cool enough zone

hence the song:

"what a difference a breeze makes"

they are growing very well..
 

CobKits

Well-Known Member
I know my names not on the list but.....
not intentional brahski, cant catch everyone around here. that was the spectrum data guru shortlist. def appreciate you opinion as youre putting the rubber to the road consistently so fire away.

What about mounting each cob in the sphere then turn it on at say 1000ma for 15 to 30 minutes to really get that heatsink up to temp,then back it down to your test current (300ma or whatever it is) and then shoot the tests with the licor.never been a fan a flash testing on cold heatsinks because in our rooms they are almost never cold after 20 minutes.
its not flash testing its an active iceled mounted outside of the heatsink. i had to mod them all to clear the sphere. lemme take some pics.

i measure 0.1, 0,2, 0.3.....0.9, 1.0, 1.2, 1.4.....3.0A in succession. the test takes about 3 minutes to sweep from low current to hi current. i thought about videotaping the meters and just ripping thru the setpoints in 60 seconds and replaying the vid to get the numbers. as of now the 3 minutes is me writing down voltage and PAR for each step.

so its a real chip on a real heatsink. the heatsink is defintiely overcooled for the low current measurements but its the same from chip to chip to chip, just trying to eliminate variables here.

phase 2 would be to do something like youre suggesting. take 2 models of chips that perform similarly and put them on a passive fixture and test them in a tent or big sphere. i could also turn the fans off on the actives and let them hit some kind of steady state

again ambient temp is a bitch were in the mountains where its 30-40 at night and 70-80 in the day

That and i would love to see the vero 29 gen 7 68v cob included in those graphs if its not too much trouble.
it will happen, i have 2 of those as well as the lower voltage ones. i just have to recycle heatsinks at some point so id like to eliminate some before i unmount them (theyre all soldered). v29 is a bitch as its a large chip with the plastic casing and just barely fits the sphere opening

What color temp were those last set of sphere test?
every chip in this test will be 4000k 80 and also will shoot spectrums for us to haggle over.

they are the most readily available for samples and the 4000k 80 doesnt have too much above 700nm and plays nice with the par meter. once we get done with teh relative output test ill shoot spectrums on 3000k and 90s and whatever just for reference but thats a separate deal
 

BuddyColas

Well-Known Member
zooming in:

View attachment 3922912

42.5/40.7 = 1.044 -> 4.4%

again thats why i want to try to test 2 chips from each model, we could have pulled an above-average cxm22 vs a below average V22. 2 or more chips are needed to really make a conclusion. im familiar with the CXM22 and that looks consistent with past testing - as for the vero- thats the first V22 ive laid my hands on (i have another as i bought a pair), so i had no idea what to expect. per datasheet luminus says +/-6% on flux, vero says +/- 7%

in my experience (for citi and luminus and the limited number of vero ive tested at least) - they are much tighter, +/- 1-2%

like look at these two separate Vero29Cs i tested (this was open air @ 12", not sphere):
View attachment 3922922
Thanks for your prompt and detailed reply. I know your sample size is small, but real-life actual comparisons are great for making buying decisions. You really back your products. Good on ya!
 
Top