Thinking of something small and portable, I came up with this integrating coffee tin: 1) a hole in the bottom for the luxmeter sensor, 2) the inside sanded and then whitened with two layers of latex wall paint, 3) the lid replaced by a piece of milky white 30% transmission perspex.
It uses two “diffusing” steps, first the perspex, then the light bouncing around in the mostly white interior.
The Texas Ace tube uses these two steps as well so I thought I’d get away with it as well, and this is smaller/cheaper.
As you can see, it integrates well, as long as the light source is flat on the perspex (so the light does not reflect away to the side) you can move the light around without altering the reading too much, 5.5% difference between edge and middle. This indicates that measurements of all light sizes and beam shapes will be a good representation of total flux.
Every diffusion step introduces spectral errors: blue and in lesser part green is removed from the spectrum with respect to red. The relevance of this is that cool lights (large blue peak that suffers much from diffusion) measure low compared to warm lights (already low blue peak that will not suffer much). A measure for this spectral error is the colour temperature difference between before and after diffusion. My large integrating sphere (#II) for example (1 “diffusing” step) lowers the CCT of a 5400K Yuji led to 4550K, so a 850K reduction. This coffee tin (2 “diffusing” steps) reduces it to 4300K, so 1100K reduction.
Still if the coffee tin is calibrated against my integrating sphere with a 3000K 92CRI flashlight, and then that calibration is used to measure a 5000K 75CRI light with the coffee tin, the error against my integrating sphere measurement of that same 5000K light is only 5%. (A factor that is also of influence on the spectral error is my sphere and the coffee tin use different luxmeters).
At about 4000 lumen, the luxmeter is out of range, so if I want to measure higher output, a neutral density filter is needed before the sensor.
My conclusion: with an integration error of only a few percent and a spectrum shift induced error of also a just few percent between very different light colours, I think this is a small, simple and very usable device for ballpark measuring of light flux, not worse than other contraptions around and better than some. But it does not reach the amount of integration and the more limited spectrum shift of a proper integrating sphere.