Cree CXA analysis

MrFlux

MrFlux

Well-Known Member
Hi guys,

While browsing the specs for the Cree CXA family (the new COBs) I happened by these promising looking spectra:
cree spectrum.jpg
Look at the beautiful purple curve of the high CRI part. It has extra power in the red and blue region, just where we want it.
Surely it looks much better than the red curve, which is the normal CRI counterpart. But is the purple curve really as sexy as it looks?

Well lets see. The main problem here is that the spectra are all relative; Each curve is scaled to have its top at 100%. This makes it impossible to compare the absolute power that the different parts put out. So the first step would be to create absolute spectra. After writing a Python script to do just that, this is the result:
spectral power.png
See what happened to the purple line. In the blue region it now has only a very slight edge over the normal CRI part. In the red region it gets more powerful after only 650nm, which is basically the deep and far end.

Instead of plotting the power we can also plot the number of photons. Since photosynthesis is driven by the number of photons this would tell the story a little bit better:
spectral flux.png

This is for the CXA1304 part btw, the smallest member of Cree COB family. The spectra of the other members are all identical and the absolute power will scale accordingly.

The Python script also calculates some properties that can not be found in the specs directly but are useful for comparing the different options. This is for the CXA1304, higher bin, 25C at nominal current:

Code: 
3000K, 93 CRI
Power in              : 3.8 W
Luminous flux         : 317 lumen
Efficacy              : 83 lumen/W
LER                   : 291 lumen/W
Radiometric efficiency: 28.7%
PAR efficiency        : 82.3%
Combined efficiency   : 23.6%
Radiant flux          : 1.09 W
Photon flux           : 5.36 uMol/s

3000K, 80 CRI
Power in              : 3.8 W
Luminous flux         : 423 lumen
Efficacy              : 111 lumen/W
LER                   : 344 lumen/W
Radiometric efficiency: 32.3%
PAR efficiency        : 86.5%
Combined efficiency   : 28.0%
Radiant flux          : 1.23 W
Photon flux           : 5.98 uMol/s

4000K, 70 CRI
Power in              : 3.8 W
Luminous flux         : 490 lumen
Efficacy              : 129 lumen/W
LER                   : 339 lumen/W
Radiometric efficiency: 38.1%
PAR efficiency        : 84.3%
Combined efficiency   : 32.1%
Radiant flux          : 1.45 W
Photon flux           : 6.77 uMol/s

5000K, 70 CRI
Power in              : 3.8 W
Luminous flux         : 527 lumen
Efficacy              : 139 lumen/W
LER                   : 339 lumen/W
Radiometric efficiency: 40.9%
PAR efficiency        : 84.1%
Combined efficiency   : 34.4%
Radiant flux          : 1.55 W
Photon flux           : 7.16 uMol/s
This is all a bit technical but the most useful parts are the various efficiencies. The radiometric efficiency tells what fraction of electricity is turned into light energy. The PAR efficiency is a property of just the spectrum, namely how well it can be utilized for photosynthesis. The high CRI part turns out to have the lowest efficiencies, both radiometric and PAR. So it would not be the best choice for an efficient grow light. It could be used for some added deep and far red though.

Also take a look at the photon flux. It turns out a lot of photons get lost in the phosphor...

It is a bit of a bummer that the 2700K spectrum is not in the specs, if you know where to find it let me know.
 
MrFlux

MrFlux

Well-Known Member
PSUAGRO. said:
Great job...........that's a BIG diff in efficiency % ^^..........+rep for your work
Click to expand...
Thanks for the compliment and the rep, glad you like it.

Abiqua, the images should be fixed now. The algorithm is basically an inverse lumens calculation. It uses the CIE luminosity function and the relative power spectrum to first calculate the luminous efficacy of radiation (LER). The LER is purely a property of the spectrum. From the LER together with the efficacy, the radiometric efficiency (= efficacy/LER) is found. From that, with the relative spectra, the absolute spectra can be reconstructed.

For the PAR calculations the McCree1972 relative quantum efficiency data is used.

The CXA3590 looks identical to the CXA1304, only 24 times bigger... So to get the flux etc you can just multiply the above results by 24. Btw if there is interest to analyze other LED families: I do take requests.
 
Abiqua

Abiqua

Well-Known Member
MrFlux said:
Thanks for the compliment and the rep, glad you like it.

Abiqua, the images should be fixed now. The algorithm is basically an inverse lumens calculation. It uses the CIE luminosity function and the relative power spectrum to first calculate the luminous efficacy of radiation (LER). The LER is purely a property of the spectrum. From the LER together with the efficacy, the radiometric efficiency (= efficacy/LER) is found. From that, with the relative spectra, the absolute spectra can be reconstructed.

For the PAR calculations the McCree1972 relative quantum efficiency data is used.

The CXA3590 looks identical to the CXA1304, only 24 times bigger... So to get the flux etc you can just multiply the above results by 24.
Click to expand...

Cool cool, thought it was an inverse of the lumen count, but wanted to make sure, very interesting again but how practical is this exactly and is done just at Test currents?

Just threw that big Cree up there, because they put out a news release yesterday, so it was purely coincidental. I did look for a while at the 2530 2540's etc.

Btw if there is interest to analyze other LED families: I do take requests.
Click to expand...
Vero's! 10-13-18 can be driven 2x nominal forward current and the 29 will go 1.5x [3.15a]
 
B

bbspills

Well-Known Member
CFA=bioWheemy 9758867]Forget all the math and graphs. Show me a plant underneath one! The proof is in the pudding.[/QUOTE]

Agreed. I have a plant under 2 CXA3050's and its doing great so far.

From seed to flowering using only 2700k LEDs

You can check my current grow on this site.
 
Bumping Spheda

Bumping Spheda

Well-Known Member
I like the idea of high CRI Warm White light. I don't, however, like the efficiency of Red/Deep Red phosphor conversion. I like how Philips put Red chips behind the lenses of their remote phosphor bulbs (the ones with Red chips basically have NW phosphor it looks like). High CRI, 2700k, nice R9 value, above average lumens/Watt rating; very nice bulb all around. If there was a 200W, flat panel version of the Philips L-Prize and it was reasonably priced I'd buy it in a heartbeat. It'd basically be a remote phosphor variant of the SGS-160.
 
MrFlux

MrFlux

Well-Known Member
bbspills said:
Agreed. I have a plant under 2 CXA3050's and its doing great so far.

From seed to flowering using only 2700k LEDs
Click to expand...
Sweet, the plant has that nice healthy green glow to it. There must be sufficient blue even in the 2700K. This was a bit of an open question for me, whether to use some extra 5000K or not.

Spheda, I've seen the teardown pictures of the Philips lamp. For a grow light you surely don't want to have the bright yellow phosphor covering the red LEDs?

Abiqua that Vero family looks very similar to the Cree CXA. I will do the Vero's later in the day but would expect a photo-finish there. You can place your bets now... efficiency freaks only!
 
MrFlux

MrFlux

Well-Known Member
OK, as promised it's now time for the Bridgelux Vero. The published spectrum looks very similar to the Cree:
vero spectrum.jpg

After digitizing the spectrum and having the Python script having a go at is, this is the absolute spectrum:
vero10 spectral power.png
The Decor spectrum is the purple curve. As you can see a grow room is just not the right place for it. It almost looks like the regular 3000K with a filter on it. Btw this is for the Vero 10.

So how does the Vero compare to the Cree CXA? To make it a fair the comparison the results are scaled to 1 Watt input power:
cxa vs vero.png
The solid lines are the Vero, the dashed lines the Cree. Take a look at the red curves of the warm whites. Which one would you prefer? This is all just math and graphs but I'm pretty sure everyone would pick the Vero. It has considerable more power in the red, and a bit less in the cyan. The blue region is about the same.

Here is the printout of the numerical results, Vero 10, 25C, nominal current:

Code: 
3000K, 97 CRI
Power in              : 9.35 W
Luminous flux         : 755 lumen
Efficacy              : 81 lumen/W
LER                   : 287 lumen/W
Radiometric efficiency: 28.1%
PAR efficiency        : 83.1%
Combined efficiency   : 23.4%
Radiant flux          : 2.63 W
Photon flux           : 13.08 uMol/s


3000K, 80 CRI
Power in              : 9.35 W
Luminous flux         : 1120 lumen
Efficacy              : 120 lumen/W
LER                   : 337 lumen/W
Radiometric efficiency: 35.5%
PAR efficiency        : 87.2%
Combined efficiency   : 31.0%
Radiant flux          : 3.32 W
Photon flux           : 16.42 uMol/s


4000K, 80 CRI
Power in              : 9.35 W
Luminous flux         : 1180 lumen
Efficacy              : 126 lumen/W
LER                   : 341 lumen/W
Radiometric efficiency: 37.0%
PAR efficiency        : 85.9%
Combined efficiency   : 31.7%
Radiant flux          : 3.46 W
Photon flux           : 16.65 uMol/s


5000K, 70 CRI
Power in              : 9.35 W
Luminous flux         : 1305 lumen
Efficacy              : 140 lumen/W
LER                   : 352 lumen/W
Radiometric efficiency: 39.7%
PAR efficiency        : 84.7%
Combined efficiency   : 33.6%
Radiant flux          : 3.71 W
Photon flux           : 17.11 uMol/s
The Vero 3000K has a very respectable efficiency, notably better the the Cree. For the 5000K it is the Cree that wins. My guess would be that Cree has the better blue drive LED and Bridgelux has the better phosphor. This would mean that at 2700K the difference would be even more in favor of the Vero. It's a real shame that there are no 2700K spectra available. Was expecting this to be a Coke vs Pepsi test but that was fortunately not the case.
 

Attachments

PICOGRAV

PICOGRAV

Well-Known Member
MrFlux, thank you for this, would you mind putting up some HID results? I would love to see HPS, LSP and MH in 3000K, 4000K, maybe 5000K and 6800K, there are also the 10,000K+ lights we could look at as well?
 
PICOGRAV

PICOGRAV

Well-Known Member
MrFlux said:
Hi guys,

While browsing the specs for the Cree CXA family (the new COBs) I happened by these promising looking spectra:
View attachment 2871410
Look at the beautiful purple curve of the high CRI part. It has extra power in the red and blue region, just where we want it.
Surely it looks much better than the red curve, which is the normal CRI counterpart. But is the purple curve really as sexy as it looks?

Well lets see. The main problem here is that the spectra are all relative; Each curve is scaled to have its top at 100%. This makes it impossible to compare the absolute power that the different parts put out. So the first step would be to create absolute spectra. After writing a Python script to do just that, this is the result:
View attachment 2871409
See what happened to the purple line. In the blue region it now has only a very slight edge over the normal CRI part. In the red region it gets more powerful after only 650nm, which is basically the deep and far end.

Instead of plotting the power we can also plot the number of photons. Since photosynthesis is driven by the number of photons this would tell the story a little bit better:
View attachment 2871383

This is for the CXA1304 part btw, the smallest member of Cree COB family. The spectra of the other members are all identical and the absolute power will scale accordingly.

The Python script also calculates some properties that can not be found in the specs directly but are useful for comparing the different options. This is for the CXA1304, higher bin, 25C at nominal current:

Code: 
3000K, 93 CRI
Power in              : 3.8 W
Luminous flux         : 317 lumen
Efficacy              : 83 lumen/W
LER                   : 291 lumen/W
Radiometric efficiency: 28.7%
PAR efficiency        : 82.3%
Combined efficiency   : 23.6%
Radiant flux          : 1.09 W
Photon flux           : 5.36 uMol/s

3000K, 80 CRI
Power in              : 3.8 W
Luminous flux         : 423 lumen
Efficacy              : 111 lumen/W
LER                   : 344 lumen/W
Radiometric efficiency: 32.3%
PAR efficiency        : 86.5%
Combined efficiency   : 28.0%
Radiant flux          : 1.23 W
Photon flux           : 5.98 uMol/s

4000K, 70 CRI
Power in              : 3.8 W
Luminous flux         : 490 lumen
Efficacy              : 129 lumen/W
LER                   : 339 lumen/W
Radiometric efficiency: 38.1%
PAR efficiency        : 84.3%
Combined efficiency   : 32.1%
Radiant flux          : 1.45 W
Photon flux           : 6.77 uMol/s

5000K, 70 CRI
Power in              : 3.8 W
Luminous flux         : 527 lumen
Efficacy              : 139 lumen/W
LER                   : 339 lumen/W
Radiometric efficiency: 40.9%
PAR efficiency        : 84.1%
Combined efficiency   : 34.4%
Radiant flux          : 1.55 W
Photon flux           : 7.16 uMol/s
This is all a bit technical but the most useful parts are the various efficiencies. The radiometric efficiency tells what fraction of electricity is turned into light energy. The PAR efficiency is a property of just the spectrum, namely how well it can be utilized for photosynthesis. The high CRI part turns out to have the lowest efficiencies, both radiometric and PAR. So it would not be the best choice for an efficient grow light. It could be used for some added deep and far red though.

Also take a look at the photon flux. It turns out a lot of photons get lost in the phosphor...

It is a bit of a bummer that the 2700K spectrum is not in the specs, if you know where to find it let me know.
Click to expand...
I would like to talk to you about the Python script you wrote, I'm not sure you are taking in enough variables in to the calculations. How did you combine the Radiometric efficiency and PAR efficiency? Then the final calculation to get Radiant flux, would like to see that.

Does the Power in Watts minus the Radiant flux Watts equal the predicted thermal outputs in Watts, and then you should see, I suppose a calculated amount of unusable Radiant flux Watts?

I am not trying pull apart any of your work but things really start to get quite complex when have to predict how well a plant is going to to conducting the this DC Flux, its almost the same concept people use to transmit radio waves through the air and space. You can put a radio wave in the air and conduct it better with a highly tuned antenna that couples with more octaves in your main wave.

CRI being the quality of your light wave, would sorta have an impact on the tone of your light, if these finally tunes pigments in side the plant, whether they are a part of Photosynthesis or not are more in tune with higher CRI light, then you would have to start adding these extra considerations into the PAR efficiency yes?

The main problem you will always have is that all your PAR calculations are based on Photon counts, some how you have to calculate where they all end up? When you think about light bouncing around pigments and such, how do to you calculate the changes in the wave energy of reflected light, white light leaves green pigments as green light waves, some of that wave energy was transferred into the pigment and you have you have a new point source a new light wave. The Photon theory just doesn't add up, these calculations would become endless...
 
MrFlux

MrFlux

Well-Known Member
PICOGRAV said:
MrFlux, thank you for this, would you mind putting up some HID results? I would love to see HPS, LSP and MH in 3000K, 4000K, maybe 5000K and 6800K, there are also the 10,000K+ lights we could look at as well?
Click to expand...
Buddy this is a LED forum. From what I have read HPS has a radiometric efficiency of around 40% and a PAR efficiency of 89%.
 
MrFlux

MrFlux

Well-Known Member
PICOGRAV said:
I would like to talk to you about the Python script you wrote, I'm not sure you are taking in enough variables in to the calculations. How did you combine the Radiometric efficiency and PAR efficiency? Then the final calculation to get Radiant flux, would like to see that.
Click to expand...
The combined efficiency is just the product of the other two efficiencies. For the actual calculations I have put an extract of the relevant code below.

Does the Power in Watts minus the Radiant flux Watts equal the predicted thermal outputs in Watts, and then you should see, I suppose a calculated amount of unusable Radiant flux Watts?
Click to expand...
The power going in is turned into heat plus radiant flux, nothing else, with no notion of 'unusable radiant flux'.

I am not trying pull apart any of your work but things really start to get quite complex when have to predict how well a plant is going to to conducting the this DC Flux, its almost the same concept people use to transmit radio waves through the air and space. You can put a radio wave in the air and conduct it better with a highly tuned antenna that couples with more octaves in your main wave.
Click to expand...
Not sure that I follow your view on how light interacts with matter here. It would be more useful to think of light as discrete energy packets, i.e. photons.

CRI being the quality of your light wave, would sorta have an impact on the tone of your light, if these finally tunes pigments in side the plant, whether they are a part of Photosynthesis or not are more in tune with higher CRI light, then you would have to start adding these extra considerations into the PAR efficiency yes?
Click to expand...
PAR by its very definition is only concerned with photosynthesis. For pigment activation or plant morphology you can set additional constraints on the spectrum like the red blue ratio.

The main problem you will always have is that all your PAR calculations are based on Photon counts, some how you have to calculate where they all end up? When you think about light bouncing around pigments and such, how do to you calculate the changes in the wave energy of reflected light, white light leaves green pigments as green light waves, some of that wave energy was transferred into the pigment and you have you have a new point source a new light wave. The Photon theory just doesn't add up, these calculations would become endless...
Click to expand...
Could it be that you are overthinking this a bit?

Anyway this is the code. The variables wavelength, spectrum, cie and rqe are all arrays and will use array multiplication.
Code: 
def calc(wavelength, spectrum, cie, rqe, powerIn, lumen):
  joule2umol = 1e-3/(const.h*const.c*const.N_A)  # for wavelength in nm
  s = sum(spectrum)
  ler = 683.0*sum(cie*spectrum)/s  # luminous efficacy of radiation
  eff = lumen/powerIn/ler
  powerOut = eff*powerIn
  parEff = sum(spectrum*rqe)/s
  avLambda = sum(spectrum*wavelength)/s
  flux = powerOut*joule2umol*avLambda
  powerSpectrum = 1000*powerOut*spectrum/s  # in mW/nm
  photonSpectrum = powerOut*spectrum*wavelength*joule2umol/s  # in umol/nm
  red = sum(spectrum[(wavelength >= 620) & (wavelength < 750)])/s
 
PICOGRAV

PICOGRAV

Well-Known Member
MrFlux said:
The combined efficiency is just the product of the other two efficiencies. For the actual calculations I have put an extract of the relevant code below.



The power going in is turned into heat plus radiant flux, nothing else, with no notion of 'unusable radiant flux'.



Not sure that I follow your view on how light interacts with matter here. It would be more useful to think of light as discrete energy packets, i.e. photons.



PAR by its very definition is only concerned with photosynthesis. For pigment activation or plant morphology you can set additional constraints on the spectrum like the red blue ratio.



Could it be that you are overthinking this a bit?

Anyway this is the code. The variables wavelength, spectrum, cie and rqe are all arrays and will use array multiplication.
Code: 
def calc(wavelength, spectrum, cie, rqe, powerIn, lumen):
  joule2umol = 1e-3/(const.h*const.c*const.N_A)  # for wavelength in nm
  s = sum(spectrum)
  ler = 683.0*sum(cie*spectrum)/s  # luminous efficacy of radiation
  eff = lumen/powerIn/ler
  powerOut = eff*powerIn
  parEff = sum(spectrum*rqe)/s
  avLambda = sum(spectrum*wavelength)/s
  flux = powerOut*joule2umol*avLambda
  powerSpectrum = 1000*powerOut*spectrum/s  # in mW/nm
  photonSpectrum = powerOut*spectrum*wavelength*joule2umol/s  # in umol/nm
  red = sum(spectrum[(wavelength >= 620) & (wavelength < 750)])/s
Click to expand...
"Even today, near the end of the twentieth
century, most people are still confused even about the
meaning of non-relativistic quantum mechanics, or the
multitude of sub-nucleonic phenomena, and the latest
theory of everything is still far from being the final solu-
tion of anything."

http://www-3.unipv.it/fis/tamq/Anti-photon.pdf
 
You must log in or register to reply here.
Top