DIY spectrometry for calculating PAR of cheap LEDs?


LED lights are cheap and plentiful, but they’re often rated in lumens and color temperature for rooms rather than PAR for plants. PAR meters are expensive, and their calibration can depend on knowing the spectrum of the lights. I’ve been thinking for a while, “Wouldn’t it be nice if there were an inexpensive DIY method to measure the spectrum and intensity of whatever cheap LEDs are readily available?”

Today I found these:

I haven’t built any of their spectrometer designs yet, but here’s a page where people have uploaded spectrum images they measured from various LEDs:

/cc @Webb.Peter

More links (edited):

Spectroradiometer that can measure PPFD from LEDs
Use of RGB LED strips for MVP

I experimented with a DIY spectrometer tonight, and it seems easy to visually distinguish “different” spectra. I’m envisioning that maybe we could get good lab tests done on representative lights and then use crude cell-phone spectrometry to match an unknown light type to the closest lab tested light we’ve got data for.

The picture below is from what I collected tonight with a piece of DVD-R taped to the back of a phone. I held the phone in one hand, put a black knit cap over my other hand, and held my phone against the cap with a V-shaped slot for light to hit the diffraction grating at a steep angle.

Description of lights & spectra:

  1. Warm white CFL with characteristic mercury vapor lines
  2. Cool white (more blue-ish) CFL with characteristic mercury vapor lines
  3. Another cool white CFL
  4. 2x Ikea 3000K 600 lumen LED bulbs. I think these use some kind of phosphor coating.
  5. Incandescent torch lamp–maybe Xenon–smooth spectrum but with lots of red and less violet than the Ikea LEDs
  6. Bright white CFL (maybe “daylight”?) that looks very blue in the room. Notice the mercury lines, but there’s also a bunch more blue-green compared to the other CFLs.
  7. Another warm white CFL.

The two pictures in the bottom right show the DVD-R shard taped to the back of a phone that I’m holding against my hat.

And here’s a shot of splitting the DVD-R, and preparing a piece for taping to the phone. I washed the purple stuff off the slice using regular hand soap:


@webbhm made one of these years ago and gave it to me. We found it to be very difficult to get working consistently, perhaps he can remember where he based his design on. I can post pictures of it later.

I’m interested to get your thoughts on how we could incorporate one of these into a system or make a step by step process by which someone could evaluate their own system. I’m honestly impressed and extremely curious as to what this really means and how it could be analyzed consistently. Have you dug into what exactly a Quantum PAR meter does to calculate PAR?

Is #6 blurry because of poor picture quality, or is it that what the CFL is actually putting out?

In your research have you found any correlation between lighting temperature ratings and par/lumen ratios? For example, if we know a CFL is 1800 lumen and 3500K, could we roughly calculate the PAR? I’m sure there are many more factors at play here such as how different wavelengths of light travel through space, it may be a different answer for 6" as opposed to 24" distance. @ferguman and I are trying to source replacements for the V2 lights and think this would be a very beneficial tool.


@Webb.Peter There are a few things people might do…

  1. Participate in the Public Lab Spectral Workbench community. They’ve got what looks like a good system for qualitative analysis of spectra with CFL based calibration. Their tools would probably be adequate as-is for measuring the peak wavelength of an unknown LED or identifying the color temperature of a CFL bulb. If you combined that with a table of lux to PAR conversion factors–I don’t know where you’d get that, but Dr. Kubota implied such things exist–you could probably use a cheap light meter to estimate PAR.
  1. Somebody with a PAR meter, a lux meter, and a spectrometer could measure different kinds of lights and publish lux to PAR conversion factors. Somebody else with just a lux meter and a spectrometer could look at the spectrum of an unknown light, match the spectrum chart to the closest lux to PAR conversion factor, and use that with their lux meter to estimate PAR.

  2. Somebody could try building a combo lux meter and automatic stepper-motor driven monochromator. If you used the right sensors and got the math right, it might be reasonable and affordable to measure photon flux vs wavelength and calculate PAR for real.

I’m not familiar with the specifics of any particular quantum PAR meters. But, according to my admittedly hazy understanding of the physics behind it, they would have to assume a spectral distribution of the light hitting the sensor. Also, I think Dr Kubota made some comments about how PAR meters intended for Mercury Vapor lights can be way off for LEDs–presumably because of the vastly different spectral energy distribution (incandescent=smooth vs. LED=spikes).

#6 is a really blue looking, higher color temperature, bright white fluorescent, so the blue-green-indigo-violet stuff is present in the light. It looks different from the other CFL spectra because there’s other stuff in the tube to give off the additional wavelengths.

Color temperature is weird. This is stretching my limited physics knowledge to the max, but I think color temperature ratings for lights are essentially a way of saying, “If you look at this light, your eyes will think it looks like the radiation coming from a black body heated to the rated color temperature”. The comparison gets really weird because of different spectral distributions between how our eyes respond to light, and how LEDS, CFLs, and incandescent sources emit light. See for more about black body radiation and a chart of color temperatures and spctra.

To convert PAR from lux–according to Dr. Kubota’s lecture–what you want is a lux to PAR conversion factor for that specific type of light source prepared by somebody who knows what they’re doing. I think she might have rattled off a couple conversion factors during one of the lectures.

[edit: I mentioned lux instead of lumens because lumens measure the amount of luminous flux coming out of the light at all angles. What you need is the flux passing through an area at leaf level. So, getting from lumens to PAR would require mathematical gymnastics maybe kinda like ray-tracing plus some spectral conversions. I think that developing a software tool for doing this might be fairly reasonable. I’ve wondered if you might even be able to do it directly with Blender by learning to configure light sources properly. That said, buying a cheap lux sensor or meter seems like a much more practical option.]

Sensor Data Modeling

These is a spectrum of my DIY LED lamp from 3 x "full spectrum" LED (blue LED with red fhosphor) & 1 white 5000K LED. Measured with DIY spectrometer with profesional monochrome USB camera.
Amplitude is not calibrated yet, only relative linearization with help of H4 car hallogen lamp at 14V.
According to these study these is optimal spectrum.

Spectrum of “Full Spectrum” LED

Does anyone have CSV table of PAR curves. I have both lumens curve (day (on the graph in shaded green) and night) and relative curve of 60W incandescent bulb

COB LED Lighting

Hey @wsnook,

I think you might really appreciate this (just the part at 53:35-55). Dr Kubota uses some sort of small film sensor (disposable strip) to measure PAR. I don’t know any details - but if you could find this it may really simplify data collection. One concern I have is that this might not work for LEDs and are only for sunlight?


It sounded like she said they put the test strips in a spectroradiometer.

I’ve kinda stopped worrying about this sort of thing because I’ve concluded that, to develop new light arrays, you really need access to test equipment that’s way beyond my hobby budget. Maybe I’ll get the chance to play with stuff like this at a properly equipped lab some day.


@wsnook I’ve been digging this thread since it started but haven’t had much to add until today. I went to Apogee’s site and noticed a ‘Field Spectroradiometer’ that’s half the cost of their lab grade stuff. For me, it comes down to consistent and repeatable measurements. I’m thinking the greatest value of a spectrometer in the context of this forum is simply comparing commercially available light sources in real-world applications. I’m all for tinkering with custom light sources but ideally we’d find a way to compare say a set of GE Bright Stiks against a commercially available grow light. Do you feel we could generate confidence of measurement with a DIY build? I’d love to think so, but sadly I’ve resigned to plunking down $3k when the time comes on a meter others will trust over time.


I’m not sure. I feel more confident about saying that the path to useful DIY measurements would–somewhere along the line–need to pass through a lab with a real spectroradiometer because you’d need to calibrate the DIY gear. Seems to me that it’s not too hard to get spectral distribution (e.g. fingerprint type of light source with a phone camera & diffraction grating) if you don’t care about the intensity at each frequency too much. The Spectral Workbench stuff uses predictable spikes on a CFL fluorescent spectrum plot for calibration of frequency, but, as far as I know, phone cameras aren’t a reasonable way to measure the intensity.

It’s also not too hard to get relative intensity with a lux sensor if you don’t care too much about differences between your target spectrum and an ideal blackbody radiator. The trick is that you have to care about those differences if you want to measure LEDs effectively. That’s why it comes back to lab grade spectroradiometers–you need calibrated readings for spectrum and intensity. But, if you came up with a DIY instrument design that took repeatable measurements, presumably you could calibrate it.


I’m not sure if this will help, but on page 4 they talk about converting foot candles into PAR.
They use a conversion factor for the sun and HPS bulbs. If you could measure the intensity of your LEDs vs a HPS bulb, you could estimate a conversion factor and at least come close to finding your PAR


I have also found a java calculator that may help you estimate your PAR. Note that for white light sources a cri of 100 is used so the numbers may be a bit off for white LEDs. The color LED calculator appears to be correct.

(As a disclaimer I did not create that page or calculator)


@webbhm @wsnook Do you think this is feasible?

These are great links! Thanks for sharing.

I’d love to add a feature to the MVP that allowed for a lux sensor to measure output based on known “approved” light sources for which we have testing data. How deep do you want to dive? :joy: