Cree CXA analysis

[Rooster99]

I may be interpreting the data incorrectly, but it would appear nothing comes close the efficiency of the 3000k 3070 z2 bin (apart from the z4's). Roughly 36% efficiency at 1.4a. Working out to US$0.82/w is spectacular for these sort of numbers! I have always considered the breaking through $1/w wall will be where we really see LEDs taking off as a viable alternative.

 

[Greengenes707]

Supra I liked it when you had the $/watts in your spread sheet. I know watts can all contain different par watts based on efficiency, but regular $/w was a great way to quickly calc production cost.
Just a thought

 

[SupraSPL]

Ya I have been trying to get used to $/PAR W because it takes radiometric efficiency into account, making a much more fair comparison than $/dissipation W. Maybe I should put both?

 

[SupraSPL]

I agree Rooster, out of all the choices I still prefer the 3000K CXA3070 Z4 (more will be available). The 2700K CXA3590 is awesome but I consider the cost of driving them a downside. Also the 3590 will concentrate the light into a super bright cone, maybe beyond the point of significantly diminishing returns, whereas a pair of CXA3070s would spread the light across the canopy more evenly without needing as much vertical space.

 

[Chronikool]

I still prefer the 3000K CXA3070 Z4 (more will be available).

Yeah...silly Cree engineers and their constant striking....

 

[caretak3r]

MrFlux - I know you ended up with Veros in your cab. I'm not asking a buying decision or anything because I already have some Veros, but I'm curious if they compare a bit more favorably with Cree CXAs since the Veros recently had a price reduction.

 

[MrFlux]

MrFlux - I know you ended up with Veros in your cab. I'm not asking a buying decision or anything because I already have some Veros, but I'm curious if they compare a bit more favorably with Cree CXAs since the Veros recently had a price reduction.

Last time I checked the 3070 had the most bang for the buck of all CXA's. It's $0.51/Watt. Vero's are from $0.45 to $0.51 per Watt. All at nominal current and prices from DIgikey.

Originally I planned on using the small CXA1304 but it lost out against the Vero 10 on price and performance.

 

[SupraSPL]

The Vero design has an interesting arrangement so the the radiometric efficiency of each current category is the same. So the Vero 13s first current category is .165A and it is 39.4% The Vero 18s first category is .35A and it is 39.8%. The Vero 29 first category is .7A and sure enough 39.6%. So it is arranged in such a way that the small COBs perform very well and the large COBs perform very well at high current levels. Thanks to Observe and Report for pointing this out. It was right in front of me in the spreadsheet and I didn't even notice.

 

[SupraSPL]

Supra I liked it when you had the $/watts in your spread sheet. I know watts can all contain different par watts based on efficiency, but regular $/w was a great way to quickly calc production cost.
Just a thought

To expand on my explanation, if you take a group of LEDs that are about the same color temp, match up the rows based on radiometric efficiency and then compare the $/PAR, you get a real picture of the up front cost, or how much you are paying/photon.

So using that method, if we compare the best available large COBS running medium strength at 39% efficiency, the CXA3070 wins at $2.19/PAR W. The CXA3590 get second at $3.67/ PAR W and the Vero29 gets third at $4.21/PAR W. it is worthwhile to mention, cheap drivers are available at this current level for the CXA3070.

If we crank up the amps until we drop down to 33% efficiency, the Vero 29 comes out slightly ahead at $1.51, the CXA3070 $1.55, depending on your shipping costs. The CXA3590 cost quite a bit more $2.33. Cheap drivers are not available at this range although we are on the trail of a $21 1.9A driver that might work.

If we really crank it, we drop to 29%, the Vero29 wins at $1.24/PAR W. The CXA3070 is maxed out at 2.8A and off my chart but it gets second at $1.34/ PAR W. The CXA3590 would be just over $2.00/ PAR W. Cheap drivers are not available in this range although there is a $33 3A Chinese driver out there that claims it can do 40V. Also, it may be impossible to achieve a Tj of 50C at high current levels because we are dependent on the ambient temps in the grow space. If so, there may be no cost benefit whatsoever to crank them. In summary, I see no benefit running them hard, it may actually be cheaper to run them medium.

 

[Observe & Report]

I think a fair comparison vs Veros must include holders because Veros have holes in them. You either must buy holders or use inferior epoxy. Yes, I know some ghetto growers use kapton tape It's only a couple of bucks but that's like a 5% difference on a $40 part.

Personally, I went for the Veros because I liked the datasheets better and they have holes. The cost/efficiency seemed pretty close at the currents I was looking at but the holders weren't yet available for CXA30*0. If I was doing it again, I would go for CXA3070 because they are made in the USA.

I'm no expert, but I would bet that the color of your thumb makes a way bigger difference than CXA vs. Vero or 3000K vs 4000k. Worry more about getting sufficient photons, an even spread, and keeping them cool.

 

[Greengenes707]

I totally see what you were trying to show, and like it. The only problem is the needed PAR watts per area isn't a familiar concept to most. I have a general idea...
With hps being around 36% a 1000w'er would be about 360PARwatts. Then if I take that a step further with the idea that it takes 600+w of led to compete with a 1000w...lets call it 65%...add in the slight efficiency bump(5-7%) for top bins and that is about ~250PARwatts for LED to target per 16sqft....so 15PARwatts/sqft????? How does that sound??I'm thinking we could get away with even a little less.


Based on everything I have gathered and my personal opinion...I would never run a DIY worse than an hps in the efficiency category. I figure if I can match that...the rest of the design(directionality, lenses, cooling) will allow the LED to get more of those PARwatts to the canopy and in the end allow me to do more with less watts(regular and PAR).
That is the concept I am going off when I design mine.

 

[mauricem00]

Good catch guod, did not realize the man was a nobel prize winner. Feel free to share any insight that the paper gives you.

others who reject quantum theory are albert einstein,and neils bohr.an instructor once told me that "to understand what a machine is telling you that you must first learn to think like the machine" if yo think that you are seing individual "photons" thru night vision googles than your understanding of mechanics is limted. every engineer knows that theory is a good starting point but needs to be built on and modified with real world experiments.in the real world light acts like a wave. mass was added as a quantum correction factor so that quantum equations would produce meaningful results.

 

[mauricem00]

I remounted a V-18 as the initial contact didn't feel great - I lifted it up after setting and you could see that only a quarter of the LED patch was making proper contact with the sink. Be careful and clamp well.

I ended up using gym plate weights to provide clamping force. The centre hole in my weights fitted nicely around the edge of the bezel - and flexed it flat to the sink. I stacked 8kg on each for 24hr of setting.

I don't have a temperature probe at the moment, but running ~1600ma without a fan the emitters pour a lot of heat into the aluminum - heating it to 50*C+ in a few mins. This is a big 700g heatsink, so definitely a fair bit of thermal mass.

The quality of the finish on my V18 contact patch wasn't great either - rougher than a typical CPU heatspreader. It'd be difficult to lap though unless you used a tiny powered polisher or something. First of you LED nerds to give it a go wins a cookie [/QUOTE]
the easiest way to determine junction temperature is to measure the voltage across the diodes when you first turn it on and then again after it warms up and temperature has stabilized. divide the difference between these two voltages by the thermal coefficient listed in the data sheet and this will give you the rise above ambient temperature for the junction

 

[MrFlux]

if yo think that you are seing individual "photons" thru night vision googles than your understanding of mechanics is limted. every engineer knows that theory is a good starting point but needs to be built on and modified with real world experiments.in the real world light acts like a wave.

When light has interacted with matter it becomes an event.

 

[robincnn]

Also the 3590 will concentrate the light into a super bright cone, maybe beyond the point of significantly diminishing returns, whereas a pair of CXA3070s would spread the light across the canopy more evenly without needing as much vertical space.

Great thread, COB at 10 inches or over is almost like a point source light. I would think 3590 will give a even coverage just as 3070.
For the same power we will need less number of heatsinks and fans with 3590 than with 3070. example: I think 2 pieces of 3590 almost equal in power to 3070.

Cree test says 3050 85degree 2250mA, that's like driving it at 150% and still it has L70 >35000hrs
35000 hrs more than enough. By 35000hrs world would have better LED's
Makes me wonder if its good idea to stick with 1000mA(50%) to 1400mA(80%) recommendation for 3070 or drive it upto 100% 1950mA ?

 

[SupraSPL]

Great questions Robin. 1925mA is the "test current" for the CXA3070 but maximum current is 2.8A. So for those running at 700mA you could say they are at 25% power. For flashlight applications or anything that is relatively short term use, running them hard is advised. But because of the long hours a grow lamp operates, we have to take efficiency very carefully into consideration. You could drive the 3070s at 1950mA and it would grow buds for sure. The main advantage doing that is to save $ up front, but the $ you save up front will dwindle to the point that I would argue a 1400mA build is actually cheaper up front than a 1950mA build. These are the main reasons:

-increased electric bill due to reduced efficiency
-additional heat due to reduced efficiency, which on a large scale can result in the need for more AC
-because of the extra heat and reduced efficiency, it requires that you buy more heatsink and more drivers to get the same job done, significantly increasing up front cost
-finding a 1950mA driver at a competitive price. We have access to cheap 700, 900, 1050 and 1400mA drivers but nothing above that (yet)
-inability to maintain a Tj of 50C will result in some lumen depreciation over time, reducing efficiency and contributing more heat and reduced yields

Another thing to consider, if you are plannning on driving at 1950 or above, you might be better served to use the Vero29 over the CXA3070 (unless you can get the top bin CXA3070s). It would be slightly cheaper and slightly more efficient.

Here are some of the numbers:
CXA3070 3000K Z2 bin @ .7A (25W) = 43% (conversion efficiency)
CXA3070 3000K Z2 bin @ 1.4A (52W) = 36.6%
CXA3070 3000K Z2 bin @ 2.1AA (83W) = 31.4%

Vero29 3000K @ .7A (24W) = 39.6%
Vero29 3000K @ 1.4A (51W) = 35.8%
Vero29 3000K @ 2.1A (80W) = 33%

CXA3070 3000K AB bin @ .7A (25W) = 49.4%
CXA3070 3000K AB bin @ 1.4A (52W) = 42.1%
CXA3070 3000K AB bin @ 2.1AA (83W) = 36.2%

 

[AirAnt]

Just curious why nobody uses red or deep red cobs, or those multicolored ones that I've seen some chinese companies producing? Looking at the charts, it would seem the 5000k spectrum white produces by far the most light. If you were also using red, or in my case with the Apollo I just ordered primarily red, 5000k would be the best choice.

I was looking for a thread just like this yesterday coincidentally, couldn't find one so just took a guess. Luckily I guessed right because I chose double 5700k whites to supplement the two 460nm blues leaving me with my 3/4 red LED ratio.

 

[SupraSPL]

It is true that the 5000K creates the most light/W, but it peaks below the ideal red range and much of its light is in the form of blue (>30%). We only need 10-15% of the light in blue during flowering, so that is the idea behind the 3000K. A large portion of its light is in the orange-red range and it has 14% blue. It doesnt suffer too badly in terms of efficiency compared to the 5000K, especially when we only have access to the lowest bin of the 5000K in the CXA3070.

We now have access to the top bin CXA3590 5000K, so if you want a huge amount of 5000K at unbelievably high efficiency levels, we have that ability now. I am testing one for vegging but I would not recommend it for flowering.

 

[Fiveleafsleft]

It is true that the 5000K creates the most light/W, but it peaks below the ideal red range and much of its light is in the form of blue (>30%). We only need 10-15% of the light in blue during flowering, so that is the idea behind the 3000K. A large portion of its light is in the orange-red range and it has 14% blue. It doesnt suffer too badly in terms of efficiency compared to the 5000K, especially when we only have access to the lowest bin of the 5000K in the CXA3070.

We now have access to the top bin CXA3590 5000K, so if you want a huge amount of 5000K at unbelievably high efficiency levels, we have that ability now. I am testing one for vegging but I would not recommend it for flowering.
View attachment 3191567

High binned 3590 5000k + a shitload of 630-660 on a separate dimmer. I think it could be a pretty awesome light!

 

[SupraSPL]

Unfortunately the 630nm 660nm chips we have are getting outdated. I still use them, but only because I had already invested in them. They are 38-40% efficient and cost $6/PAR W, but the CXA3070 3000K Z2 is 48% efficient at $6/PAR W. The AB bin is 55% efficient at $6/ PAR W. So for now, 3000K is the best bet we have available to us IMO.