...All Things Vero...

[Airwalker16]

What is your avatar? Maple leaves?

 

[CobKits]

Cobkits why is the citizen spreadsheet giving such low effi% numbers ? I don't get It, your numbers are like 15-20lm/w higher then what Citizen is saying. I know real world data is different but still we are talking here about 10%+ difference and that is huge.
If I read the citizen pdf sheet It says there can be a 10% difference between a chip maybe you this explains it ? And you have the golden nugget pheno of the cobs haha in your tests.

yeah their simulator is very conservative relative to other mfrs. if i take one of their chips and operate it where they say 170 lm/W should be, and do the same with a cree, citi is much brighter.

example:

PCT says 2.0A on a db bin cree is 12203 lm/75.8 W = 161 lm/W. i get 601 ppfd/73.2W = 8.2 ppfd/W

citi tool says 1.4A on an 1818 4000K 80 cri is 11712 lm / 74.1W = 158 lm/W i get 623 umol at 72.6W = 8.6 ppfd/W

i only ever use the tools to compare products of a given manufacturer. yes they are off by up to 10% mfr to mfr

you can play around all you want with Tc etc but there will always be that bias

 

[SupraSPL]

The CREE datasheets don't bother with data at low dissipation levels, I assume that is because CREE could not conceive that customers would use CXB3590s at 25W or less.

I had problems getting reliable measurements at less than 4W/ COB @ 12", but I can confirm that efficiency continues to increase steadily as dissipation power falls below 25W, all the way down to 4W for both CXBs and Vero29s.

Based on that data I have no doubt that we can get well above 200lm/W with chips that are already available. Great for mobile setups, solar powered setups or during a power outage you can dim way down and run on batteries.

 

[CobKits]

another at a higher current where they should match.

PCT says 2.4A on a db bin cree is 14097 lm/93.5 W = 150.7 lm/W. i get 700 ppfd/89.86W = 7.79 ppfd/W

citi tool says 1.8A on an 1818 4000K 80 cri is 14738 lm / 97.5W = 151 lm/W i get 773 umol at 94.95W = 8.14 ppfd/W


if i bump Tc from 25 to 55 in each case
PCT says 2.4A on a db bin cree is 13212 lm/92.87 W = 142.3 lm/W
citi tool says 1.8A on an 1818 4000K 80 cri is 13953 lm / 95.4W = 146 lm/W

that could be a lot of it- maybe citis just have far better thermal management at higher, realistic case temps? maybe comparing data at 25C is useless for these purposes. in any case you can push paper to the cows come home but ill take real measurements any day of the week for my own purposes. how you choose to use that data is up to you.

 

[CobKits]

hey supra! welcome back...thanks for all your work over the years. trying to carry the torch here the best i can

 

[CobKits]

Based on that data I have no doubt that we can get well above 200lm/W with chips that are already available. Great for mobile setups, solar powered setups or during a power outage you can dim way down and run on batteries.

im not doubting that, but the vero simulator is saying 230 lm/W @ 25W, which.... well that would be something....

 

[CobKits]

iin my conversation with jerry today (where he said 38V veros were available next week in 4000K, with 3000 and 5000K end of the month)

i asked him when 50 and 72V would be available and he said "maybe months later"

hard to say what 'months' will mean - digi is saying october but not sure if that means all voltages

38V chip is nice but not getting me all that excited. the 50V tests at 2-3% more efficient per a trusted RIUer who has samples. i have no confidence in the 72V data they provide at this point so we'll just have to wait and see i guess.

 

[bobbuck]

Wondering what anyone thinks about the new 18s (29v)? I do a pretty small style grow and like the way the 6 18s (gen 6) provide even coverage in my space, plus they fit my driver well! Looks like I could go all 90 cri and still get a bump in output.

 

[alesh]

another at a higher current where they should match.

PCT says 2.4A on a db bin cree is 14097 lm/93.5 W = 150.7 lm/W. i get 700 ppfd/89.86W = 7.79 ppfd/W

citi tool says 1.8A on an 1818 4000K 80 cri is 14738 lm / 97.5W = 151 lm/W i get 773 umol at 94.95W = 8.14 ppfd/W


if i bump Tc from 25 to 55 in each case
PCT says 2.4A on a db bin cree is 13212 lm/92.87 W = 142.3 lm/W
citi tool says 1.8A on an 1818 4000K 80 cri is 13953 lm / 95.4W = 146 lm/W

that could be a lot of it- maybe citis just have far better thermal management at higher, realistic case temps? maybe comparing data at 25C is useless for these purposes. in any case you can push paper to the cows come home but ill take real measurements any day of the week for my own purposes. how you choose to use that data is up to you.

Well, comparing point measurements (BTW what sensor do you use, what's it calibrated to and do you apply any correction to the taken data?) in order to estimate a COB's efficiency is not an exact science. I'm sure there's a quite large margin of error.

Anyway, I expect Citizen's 1818 to indeed have a better thermal management. There's no R̶_̶t̶h̶ R_j-c in the Cree datasheet but it seems that at binning current, T_c = 55 °C corresponds to T_j = 85 °C, while the input power is ~87 W of which 45-50 W is heat. It makes R̶_̶t̶h̶ R_j-c ~0.63 °C/W_heat or ~0.35 °C/W_input. Citizen states R̶_̶t̶h̶ R_j-c = 0.17 °C/W_input.

 

[CobKits]

Well, comparing point measurements (BTW what sensor do you use, what's it calibrated to

licor li190 w/factory calibration, could prob use to go in to them for calibration but i only ever use it for relative measurements anyway

and do you apply any correction to the taken data?) in order to estimate a COB's efficiency is not an exact science. I'm sure there's a quite large margin of error.

you bet. which is why i disclaim. its all relative. in the end plants dont care how efficient a chip is but they sure do care about the raw ooohmoles 12" below it (and yes we take for granted all have similar beam angles- datasheet suggests this)

Anyway, I expect Citizen's 1818 to indeed have a better thermal management. There's no R_th in the Cree datasheet but it seems that at binning current, T_c = 55 °C corresponds to T_j = 85 °C, while the input power is ~87 W of which 45-50 W is heat. It makes Rth ~0.63 °C/W_heat or ~0.35 °C/W_input. Citizen states R_th = 0.17 °C/W_input.

which woudl explain how they run out so far on current. does that number already consider package size? the clu058-1825 are 0.14 Rth and the -3618 are 0.10, both same package

same for 1212 @ 0.34 and 1818@ 0.17 in same package size

youd think cramming more dies in a package would reduce efficiency but the opposite seems to be true

0.10 explains the astonishing performance of the 3618 - 134 lm/W @ 300W on a $60 chip thats cheaper than an HPS bulb

but back to the subject
Vero29 V7



is R j-c only a component of overall Rth?

also we see its not constant but varies with current. could the different citi data be partially a result of differnet nominal current/wattage points for each series?

 

[alesh]

licor li190 w/factory calibration, could prob use to go in to them for calibration but i only ever use it for relative measurements anyway

Seems like a good one but still has different response for different spectra.

you bet. which is why i disclaim. its all relative. in the end plants dont care how efficient a chip is but they sure do care about the raw ooohmoles 12" below it (and yes we take for granted all have similar beam angles- datasheet suggests this)

Yes, but small difference in beam angle could make noticeable difference. Size of LES also affects the measurement.

which woudl explain how they run out so far on current. does that number already consider package size? the clu058-1825 are 0.14 Rth and the -3618 are 0.10, both same package

same for 1212 @ 0.34 and 1818@ 0.17 in same package size

youd think cramming more dies in a package would reduce efficiency but the opposite seems to be true

More dies = larger area in contact with the case
It makes sense. If you cram more dies into a package you run each die softer.

but back to the subject
Vero29 V7

View attachment 3782016

is R j-c only a component of overall Rth?

also we see its not constant but varies with current. could the different citi data be partially a result of differnet nominal current/wattage points for each series?

I should've used R_j-c in the original post (corrected).
R_j-c is usually in °C per input W so it changes as a larger portion of input power is converted to heat at higher currents.

 

[CobKits]

More dies = larger area in contact with the case
It makes sense. If you cram more dies into a package you run each die softer.

ah yes, deg C per watt,

Seems like a good one but still has different response for different spectra.

unfortunately i only have 70 cri cree and only have 80 cri veros and citis so i had to make a compromise somewhere. i did some 3000k 80 cri tests with similar results tho

 

[JorgeGonzales]

Anyway, I expect Citizen's 1818 to indeed have a better thermal management. There's no R̶_̶t̶h̶ R_j-c in the Cree datasheet but it seems that at binning current, T_c = 55 °C corresponds to T_j = 85 °C, while the input power is ~87 W of which 45-50 W is heat. It makes R̶_̶t̶h̶ R_j-c ~0.63 °C/W_heat or ~0.35 °C/W_input. Citizen states R̶_̶t̶h̶ R_j-c = 0.17 °C/W_input.

That's what I said! If there is a 30C delta between Tc and Tj like @SupraSPL theorized to explain the PCT discrepancies, then Cree's thermal resistance is relatively terrible. But @guod insists that Cree's thermal resistance is as good as Bridgelux, and is he ever wrong? So there you go.

 

[ya bongo]

from the good old days, where SDS and Guod debating those things.

https://ledlight.osram-os.com/wp-content/uploads/2013/01/OSRAM-OS_LED-FUNDAMENTALS_Internal_Thermal_Resistance_of_LEDs_v4_09-07-11_SCRIPT.pdf

 

[alesh]

That's what I said! If there is a 30C delta between Tc and Tj like @SupraSPL theorized to explain the PCT discrepancies, then Cree's thermal resistance is relatively terrible. But @guod insists that Cree's thermal resistance is as good as Bridgelux, and is he ever wrong? So there you go.

Well it's what the data sheet suggests, I think. I have no means to verify it and TBH almost no experience with Bridgelux to tell. So there we go

 

[Grobek]

digi is saying october but not sure if that means all voltages

http://www.digikey.ch/product-detail/de/bridgelux/BXRC-40E10K0-C-73/976-1425-ND/6152435

 

[mr. Amsterdam]

http://www.digikey.ch/product-detail/de/bridgelux/BXRC-40E10K0-C-73/976-1425-ND/6152435

Thanks great, let's test some Vero's 70V

 

[ReddEyez]

They have 3000k and 4000k 80&90 CRI in stock, as well as 70&80 CRI in 5000k guys.

http://www.digikey.com/catalog/en/partgroup/gen-7-vero-18-array-series/62985

http://www.digikey.com/catalog/en/partgroup/gen-7-vero-29-array-series/62986

 

[Rahz]

Has anyone been playing with the new datasheets?

The current -vs- flux graph doesn't seem to be accurate when compared with the flux figures in the common drive currents chart, at least with the 3500/80 C variant. I'm pretty sure the chart is correct and there's such a large discrepancy the graph is pretty much useless.

 

[JorgeGonzales]

Has anyone been playing with the new datasheets?

The current -vs- flux graph doesn't seem to be accurate when compared with the flux figures in the common drive currents chart, at least with the 3500/80 C variant. I'm pretty sure the chart is correct and there's such a large discrepancy the graph is pretty much useless.

I saw some obvious typos in the flux charts last time I looked, so I stopped digging too deeply. The hand-drawn charts have never been confidence-inspiring either.

Stuff like this, I'm sure there's more:


Doesn't mean they aren't great LEDs, there are errors in other manufacturer's data sheets as well.