Top bin COB comparison

I have been meaning to do this test for awhile and finally got a chance last night. I used an Apogee SQ-120 PAR meter with the following correction factors provided by Apogee:

CREE CX 3000K 80 CRi (divide reading by .952)
CREE CX 4000K 80 CRi (divide reading by .97)
CREE CX 5000K 70 CRi (divide reading by .962)

estimated
CREE CX 3500 80 CRi (divide reading by .96)

Each COB was hung at 12", from the PAR meter, as precisely as I could manage. The PAR meter was centered directly under the COB. The readings were pulsed, ambient was 70F (21C) so Tj was 21-25C.

I used adjustable constant current drivers and monitored Vf and If with Fluke and Amprobe multimeters. The dissipation Wattage was divided by the PPFD reading to arrive at the awkward but useful "PPFD/W @ 12"

These curves show the effect of current droop and how it affects each COB differently. The photon counts should be accurate enough within reason that these curves can be compared absolutely. There were some surprises in the results:

Some of the takeaways:

-The efficiency gains from running at low power are real/measureable and they continue accumulating even at very low dissipation wattage.

-The CXB3070 3K AD performed surprisingly well at low-mid power, especially if you factor in the cost difference. It outperformed the CXB3590 3000K CB. I will perform the CXB3070 test again with a different COB to see if I get the same impressive performance. If so, it disagrees with the CREE datasheets.

-The CXB3590 5000K CD had a great photon count compared to all the others, but since a larger portion of its photons are blue, I am not sure if it could outyield the 3500K or 4000K in practice.

Overall, I think it could make some sense to mix the 3500, 4000 and 5000K in a flowering room to widen the spectral curve.

SupraSPL said:

Some of the takeaways:

-The efficiency gains from running at low power are real/measureable and they continue accumulating even at very low dissipation wattage.

-The CXB3070 3K AD performed surprisingly well at low-mid power, especially if you factor in the cost difference. It outperformed the CXB3590 3000K CB. I will perform the CXB3070 test again with a different COB to see if I get the same impressive performance. If so, it disagrees with the CREE datasheets.

-The CXB3590 5000K CD had a great photon count compared to all the others, but since a larger portion of its photons are blue, I am not sure if it could outyield the 3500K or 4000K in practice.

Overall, I think it could make some sense to mix the 3500, 4000 and 5000K in a flowering room to widen the spectral curve.

Click to expand...

Nice man. Great testing. Thank you. I'll send you a cxb 3070 3500k bb so you can play with it.

We got to remember bins...ranges min to max. So might be one closer to max against more typical or lower ranged.

I'm still trying to get accurate spds of them. 15nm accuracy was a joke. Buying a stellar net for my Christmas gift to myself. Give us some more spds to compare to the data sheets.

Thanks man (I still owe you one ). The results definitely raise the CXB3070 3500K BB question, very curious to see how it will do. True it is possible that I have a smokin 3070 3K AD performing near max and the 3K 3590 could be near min. It could also be a measurement error caused by the positioning of the PAR meter, so I will definitely do some more tests to get a better idea what is going on. I left the room set up for COB testing so it will be easy to add to the data set and make direct comparisons.

You considering the BLACK-Comet "SR"? Bummer about Apache robbery!

So the 3590 isn't as great as we thought? Seems like the 3070 is just as good yet cheaper correct?

SupraSPL said:

Some of the takeaways:

-The efficiency gains from running at low power are real/measureable and they continue accumulating even at very low dissipation wattage.

-The CXB3070 3K AD performed surprisingly well at low-mid power, especially if you factor in the cost difference. It outperformed the CXB3590 3000K CB. I will perform the CXB3070 test again with a different COB to see if I get the same impressive performance. If so, it disagrees with the CREE datasheets.

-The CXB3590 5000K CD had a great photon count compared to all the others, but since a larger portion of its photons are blue, I am not sure if it could outyield the 3500K or 4000K in practice.

Overall, I think it could make some sense to mix the 3500, 4000 and 5000K in a flowering room to widen the spectral curve.

Click to expand...

It makes sense also if you just look at the spectral distributions...which is why i've done this since the beginning

SupraSPL said:

Some of the takeaways:

-The efficiency gains from running at low power are real/measureable and they continue accumulating even at very low dissipation wattage.

-The CXB3070 3K AD performed surprisingly well at low-mid power, especially if you factor in the cost difference. It outperformed the CXB3590 3000K CB. I will perform the CXB3070 test again with a different COB to see if I get the same impressive performance. If so, it disagrees with the CREE datasheets.

-The CXB3590 5000K CD had a great photon count compared to all the others, but since a larger portion of its photons are blue, I am not sure if it could outyield the 3500K or 4000K in practice.

Overall, I think it could make some sense to mix the 3500, 4000 and 5000K in a flowering room to widen the spectral curve.

Click to expand...

Here is my question. I wonder how the smaller cxb3070 fair temp droop wise to there cxb3590 brethren. Also I wonder about Les size being larger at 12" is the spread exactly the same? Are the larger LES putting down the same exact footprint. Makes me scratch my head for sure. I'm guessing that center is a good measuring tool but I think the spread of the bigger chips will be 16% bigger. I could be dumb as a rock on this though. So basically you have roughly 16% more phosphorus area and blue led area underneath. So if cxb3590 3500k should be roughly 12% more efficient or 20 something % more light but spread 16% further than these findings are spot on. I'm at work and I'm not doing the math right now lmao.

Getgrowingson said:

So the 3590 isn't as great as we thought? Seems like the 3070 is just as good yet cheaper correct?

Click to expand...

Hard to say from one set of test results but I will definitely investigate more to try and figure that out.

@Stephenj37826 Yes it is possible that the difference in LES and beam pattern could skew the results for the 3070s. I will test a CXA3070 3K AB and see how that stacks up.

mc130p said:

It makes sense also if you just look at the spectral distributions...which is why i've done this since the beginning

Click to expand...

Regarding which one? The spectral distribution graphs in the CREE PDF can be very misleading because they are not relative to each other and do not take available bins into account (see Mr Flux CREE CXA Analysis thread)

Stephenj37826 said:

Here is my question. I wonder how the smaller cxb3070 fair temp droop wise to there cxb3590 brethren.

Click to expand...

True temp droop could change the picture when running them medium-hard. Cree combines temp droop with current droop in their graphs but if you compare their 25C lumens vs 85C lumens you can get an idea how 3070 fares vs 3590

I'm still puzzling over this, not just the red line. At 50 watts the difference between AD and CD looks to be about 3-4% and you have two samples of CD. Another AD being tested would be good but I agree with Stephen, it may have something to do with how the light is spread otherwise it seems like a big anomaly.

Another way of looking at it, at their respective nominal currents the 3070AD and 3590CDs have the exact same efficiency in your test, yet the datasheets indicate the 3590 should be about 7% more efficient at nominal.

Thanks for the data Supra.
The Spatial Distribution graph of 3070 and 3590 overlap. So their light spread should be similar. I think tiny change is LES size 30mm - 23mm= 7mm can be ignored given the reading is being taken at 18 inches (457.2mm)

I would assume 3590 is more efficient than 3070 at any given drive current. It would have more parallel groups diodes so lower current per diode.
3590 may have lost with 3070 in round 1 due to error. Waiting for the rematch results.
From Cree PCT i see 3070 AD, 3590 CB are very close in terms of lm/w. they do not have data for 3590 below 1.4amps

robincnn said:

Thanks for the data Supra.
The Spatial Distribution graph of 3070 and 3590 overlap.

Click to expand...

Hmm, they seem to be identical so we should rule out light distribution unless as an anomaly, though I'm not sure how that would work.

SupraSPL said:

Hard to say from one set of test results but I will definitely investigate more to try and figure that out.

@Stephenj37826 Yes it is possible that the difference in LES and beam pattern could skew the results for the 3070s. I will test a CXA3070 3K AB and see how that stacks up.


Regarding which one? The spectral distribution graphs in the CREE PDF can be very misleading because they are not relative to each other and do not take available bins into account (see Mr Flux CREE CXA Analysis thread)

Click to expand...

I saw the line-up and thats what I was going to ask you to do.
Great idea and thanks for taking the time to give us this info!

SupraSPL said:

True temp droop could change the picture when running them medium-hard. Cree combines temp droop with current droop in their graphs but if you compare their 25C lumens vs 85C lumens you can get an idea how 3070 fares vs 3590

Click to expand...

Are these temps measured directly on the LES?

bassman999 said:

Are these temps measured directly on the LES?

Click to expand...

I believe that junction temp is inside the package. If I understand correctly the LES can be higher than the junction temp. That is why it is difficult to directly measure junction temp. It is also difficult to measure LES temp because the huge stream of photons will interfere with the reading. You could click the light off and quickly try to get an LES temp reading with a thermometer that is pre-warmed to the approximate LES temp. I have never tried to do that but it would be an interesting test.

SupraSPL said:

I believe that junction temp is inside the package. If I understand correctly the LES can be higher than the junction temp. That is why it is difficult to directly measure junction temp. It is also difficult to measure LES temp because the huge stream of photons will interfere with the reading. You could click the light off and quickly try to get an LES temp reading with a thermometer that is pre-warmed to the approximate LES temp. I have never tried to do that but it would be an interesting test.

Click to expand...

In an attempt to see that all my cobs are seated well, I measured each, but results are up and down, and so I cant tell how well I am cooling them besides the reading from back of HS or on side of cob.
Any measurement on cob side when lit up is very hard to do right though.

I believe that different size of LES could change results slightly.

alesh said:

I believe that different size of LES could change results slightly.

Click to expand...

Does LES mean light emitting section or something like that?

So are saying the 3070 is stronger toward the center but has less side coverage of lumens; and the 3590 has better coverage with higher lumens so less in the center? I agree with that.

I'd like to see approximate 2x2 grid comparisons between the 3590 and 3070 now. That would be cool to see how each cob handles spreading the light and would possibly explain the anomaly above.

4ftRoots said:

Does LES mean light emitting section or something like that?

So are saying the 3070 is stronger toward the center but has less side coverage of lumens; and the 3590 has better coverage with higher lumens so less in the center? I agree with that.

I'd like to see approximate 2x2 grid comparisons between the 3590 and 3070 now. That would be cool to see how each cob handles spreading the light and would possibly explain the anomaly above.

Click to expand...

Light emitting surface. And yes, that's basically what I'm saying. It's similar to XP-E still being a popular choice for 'throw' flashlights despite its age and inefficiency compared to newer models.

The spread will not be better. We are talking about ~3-4mm "more spread" if at all measurable.
I know 100% that retrofitting an existing 3070 design with 3590 at the same current will not improve any coverage measurable by a meter.
I'm going to be without a lab for a week or two so new testing setups are on a halt. But can get one up Next weekend if someone doesn't beat me to it....I expect some to please.