# Estimate Lux & Lumens with Light Meter? LumenToob

#### Bolster

##### Flashlight Enthusiast
EDIT: This thread has evolved into the "LumenToob" project, which is a "reverse engineered" method of making ballpark estimates of lumens using a common handheld photo meter and PVC pipe fittings.

I need to measure the output of several lights (handhelds & headlamps) so I can write better reviews! Recently my reviews have been questioned because I don't give any sort of brightness measurements.

I have a nice Gossen light meter, the Luna-Pro Digital F. It can convert from reflective to incident with a sliding dome. It gives output in EV values. How do I use this meter to measure lux, or footcandles -- and is there any possibility to somehow estimate lumens with it? (Even if it's only approximate?)

For lux, is it as simple as putting the flashlight one meter away from the light meter, taking an EV reading, and then converting to Lux or FootCandles? Is this done with the meter in reflective or incident (dome) mode? This phrase "2 - Your light meter will be set up with 1 meter distance for those measuring in lux (or 1 foot for those measuring in foot candles) between the dome of the meter sensor and the outside edge of the bezel ring" makes me think these are incident readings, with the light pointed at the dome -- not simple reflective mode. Correct?

Is there any way of using a ceiling bounce (or other method) to estimate lumens, other than an integrated sphere? A ceiling bounce or some other home-style method won't do for guesstimation, such as this "integrated tee"? At least for comparative guesstimation, if not absolute? Perhaps not, from this comment: "When you find a Lumen measuring light meter, let us know. There's probably 4,000 of us here on CPF that would like to buy one." I keep thinking sliding incident dome should somehow allow me to estimate lumens. Because the purpose of the dome is to measure light hitting it from any and all angles within half a sphere.

Certainly all these questions have been asked before, and I've searched and read many 'light meter' threads, but have not found a "how to" thread on this topic, although this one came close. Nothing in 'threads of interest' either. There must be a thread on this somewhere, I just can't find it.

lux w/o light box

Last edited:

#### Robin24k

##### Flashlight Enthusiast
Re: How to Estimate Brightness with a Light Meter?

Generally speaking, you should only treat lux readings as relative, meaning it's only good for making runtime graphs (that's what I use my Extech light meter for). Without an integrating sphere, you cannot compare different lights because reflectors and bezel size will throw off your readings. In addition, your light meter is probably not calibrated for LEDs (standard light meters aren't), so that will also introduce errors.

The reason lumens is difficult-to-grasp unit is because it's luminous flux, the energy of all light that is produced. Therefore, you need something with a known area in which light is spread uniformly, hence the need for an integrating sphere.

#### MikeAusC

##### Enlightened
Re: How to Estimate Brightness with a Light Meter?

. . . . Without an integrating sphere, you cannot compare different lights because reflectors and bezel size will throw off your readings. . . . . .

That's why it's important to take readings many metres away from the light. Especially for tight throwers, you can get inconsistent readings if the Light Meter is less than 10 metres from the light.

#### Robin24k

##### Flashlight Enthusiast
Re: How to Estimate Brightness with a Light Meter?

That still won't give a reliable reading, spill is ignored and values for reflector-based lights will be too low. You cannot get flux without integrating over a Gaussian surface (hence, an integrating sphere where the Gaussian surface is a sphere).

#### MikeAusC

##### Enlightened
Re: How to Estimate Brightness with a Light Meter?

That's why it's important to take readings many metres away from the light. Especially for tight throwers, you can get inconsistent readings if the Light Meter is less than 10 metres from the light.

Of course that's only for measuring Lux.

#### Robin24k

##### Flashlight Enthusiast
Re: How to Estimate Brightness with a Light Meter?

I see, yes, that would be fine then.

While peak beam intensity can be calculated without an integrating sphere, an LED-calibrated light meter is still required. With a standard light meter, I don't think lux or candela readings conform to the FL1 Standard.

#### Bolster

##### Flashlight Enthusiast
Re: How to Estimate Brightness with a Light Meter?

OK, can this egg be cracked with a different approach...perhaps by the application of some statistics...again, I'm not looking for Truth, I'm looking for comparative estimations, so I can relax the parameters a bit and leave the precise lumen counts to the labs.

Imagine you shone a flashlight into one side of a 2" white PVC 90-degree elbow bend. Imagine you measured the reflected light from the other end of the elbow bend. This would ensure that only reflected light was measured, nothing direct, and it would bounce around in a reasonably small reflective space.

The light meter uses an incident reading, so it is also diffusing the indirect light it's seeing in the elbow. You should be getting a reasonable mix of reflectance of light from multiple (mostly round) surfaces at this point, regardless if the measured light is spot or flood. Again, this reflectivity gets diffused through the meter's dome, so there's a "second mixing/diffusion" point.

You record various flashlights of known lumens (in my case, I'm using manufacturer claimed lumens - again, I'm looking to compare manufacturer claimed values, not looking for absolute values.)

Then you build a regression formula, to predict (in my case, manufacturer's claimed) lumens by EV values shown on the meter. You could also enter degrees of beam spread as a predictor--who knows, it might go significant. You could even enter manufacturer as a variable (although that would complicate things with a categorical variable). The computer would give you a regression formula whose result would predict what the average manufacturer would claim a light to be. (And by "average manufacturer," I mean the average of the manufacturers' advertised lumen counts you used to build your formula.)

Basically you leave the hard work of precise calibration (including the tuning of the light meter to LEDs) to the regression formula. The result would not tell you the lumens. It would tell you what your average manufacturer would claim for lumens. Which is actually what I'm looking for...I would like to place an unknown light into a lineup of other lights.

Last edited:

#### Bolster

##### Flashlight Enthusiast
Re: How to Estimate Brightness with a Light Meter?

By the way, never got a confirmation of how to do the lux readings. We jumped over that and into the lumens discussion.

In post 1, do I have the lux reading procedure correct?

#### Robin24k

##### Flashlight Enthusiast
Re: How to Estimate Brightness with a Light Meter?

I don't know about EV, but you would use incident mode. Your meter is designed for photography, right?

#### Echo63

##### Flashlight Enthusiast
Re: How to Estimate Brightness with a Light Meter?

I used my photographic light meter (a minolta flash/ambient meter) to measure lux/convert to CP the other night to measure my Maxabeam (it hit 7.2million cp)

I used mine in EV mode (at 100 iso - some meters change the numbers at different iso, some don't)
Then used a table I found on the Internet to convert EV to lux
Lux x distance from source to meter x distance from source to meter = Candlepower (or lux at 1m) although for smaller lights you could measure it at 1m.

I know one of the review sites had a fairly close way of estimating lumens using a milk carton and a luxmeter, it would require a few lights through it to calibrate it thiugh

#### Swedpat

##### Flashlight Enthusiast
Re: How to Estimate Brightness with a Light Meter?

A light meter is good for measuring the lux values of hotspot and spill from a certain distance and make a runtime graph.
Also this is discussed; but I am personally convinced that you can get a very good estimation of the total output by comparing different light with careful ceiling bounce test.

Last edited:

#### Bolster

##### Flashlight Enthusiast
Re: How to Estimate Brightness with a Light Meter?

Yes, Robin, it's a photography light meter.

I think I am basically doing a "ceiling bounce" using a 2" PVC elbow. I'm testing the concept now, and with only 15 datapoints from 4 lights, printed out on an Excel scatterplot, it looks like I'm so far able to predict the lumens (not lux) within about a 10-15 lumen error from a range of about 20 lumens to 200 lumens. All this is eyeballing it, with not much data, so things could change. I need to get on my work computer to run a regression formula. Then we'll have an r-squared that will tell us whether this approach is worthy or not.

Last edited:

#### Bolster

##### Flashlight Enthusiast
Re: How to Estimate Brightness with a Light Meter?

Progress on the LumenToob...

\$7 for PVC elbows, tees, and caps. Opted not to use PVC pipe to put it together, the masking tape is sturdy enough. The meter cap is milled out to just mate with the meter, which fits in during use. On the right, aluminum foil keeps some of the reflected light in the tube. When you view through the meter slot you see a nice even spread of light, regardless of the beam shape of the light. Now collecting EV values to build the regression formula.

Beam placement on the right is not very critical. You can move the light around a fair amount and not see any difference in the EV values. I take that as a good sign.

With lumens on the X axis and EV values on the Y axis, I'm getting a curve that looks something like an asymptote, just as you'd expect. Ie, a few more lumens make a big difference in brightness at the low end of the curve, and lots more lumens make a little more brightness on the high end.

Last edited:

#### Bolster

##### Flashlight Enthusiast
Re: How to Estimate Brightness with a Light Meter?

What? No criticism? Feedback...?

Here's the data going in... X is EV values from the light meter, Y is lumens claimed by mfgr...unfortunately I only own 7 flashlights for which I have the manufacturer's stated lumen count....wish I had more.

Still, you can see the curvilinear shape appearing...the fact they're not in a perfect curve represents error--from differing lumen estimation methods of the different manufacturers, from imprecise repeatability of my light meter, from different beam patterns registering differently in my "lumen toob," and so on.

Next we let the computer sort it out...draw the line through the points that misses them all by the least possible amount. Two independent variables: EV (shown below) and angle of beam (not shown, but including it in case it explains lower readings for full flood lights and perhaps help compensate for my home-made "lumen toob" system).

Last edited:

#### Robin24k

##### Flashlight Enthusiast
Re: How to Estimate Brightness with a Light Meter?

That's an interesting idea, but as your data shows, the error is pretty significant and non-linear (parabolic or exponential). I'm still going to say the same thing...you should only treat light meter readings as relative.

As a reviewer, any data you present should be factual, and this type of estimation is unprofessional and negligent. That being said, you do have a fairly controlled environment for making runtime graphs.

What? No criticism? Feedback...?
CPF wasn't working consistently for the past couple days, and it seems like whenever I checked my email to get topic replies, it's always down (or very slow)...

#### Bolster

##### Flashlight Enthusiast
Re: How to Estimate Brightness with a Light Meter?

That's an interesting idea, but as your data shows, the error is pretty significant and non-linear (parabolic or exponential). I'm still going to say the same thing...you should only treat light meter readings as relative. As a reviewer, any data you present should be factual, and this type of estimation is unprofessional and negligent. That being said, you do have a fairly controlled environment for making runtime graphs. .

LOL, I like you Robin! Good one. And thanks for the reply, I was beginning to think nobody cared.

Error: Yep, it's there. My guess is, the biggest source comes from the different ways the mfgrs measure their own products, an issue which has been widely commented upon. Another source of error could obviously be my LumenToob...it may not do as well with some beams as with others. In fact, I'd guess that's the case, so am entering a variable (beam spread) to possibly counteract. That said, even with the current data, you could draw out a curve by hand that would give you reasonable estimates. Once the formula comes in, I'll know how "off" it actually is. You can predict your own data and then observe the error. For example, that 25 lumen datapoint would appear to be a gross overestimate by the manufacturer. If it really is 25 it should be putting out 6-7 EV. I need to go back and verify that datapoint, maybe it's me who made the error. Even a low battery could do that.

FWIW, deviance in the direction of the upper left portion of the graph is "bragadocious," and deviance to the lower right is "humble."

Non-linear: That's not a problem, because I'm introducing nonlinear terms into the equation. I'll be fitting a curve, not a straight line. We have the technology! (More accurately, I will be fitting a number of curves, if the beam spread becomes a significant variable.)

Data factual: I don't work at the University of East Anglia, so my measurements are based on data...I'm not making up the EV values. (If I were making them up I'd have made them up in a more perfect curve! Like the University of East Anglia!) I can't speak for the manufacturers' lumen values, but remember...I'm trying to predict what a typical manufacturer would SAY the lumen output is, not what the ACTUAL lumen output is. The purpose of which, is to be able to place an unknown light in a relative lineup of other lights in the marketplace (for which we almost universally use the manufacturer's own published numbers, for better or worse).

You have my promise, however, that I won't represent my findings as equivalent to an integrated sphere. I'm reverse engineering this process with \$7 of PVC, a photo meter, and a regression program. Perhaps I shall call my estimates: "Bolster's negligent and unprofessional estimates." It has a good ring to it.

If you are interested in a "blind" test of this, when done, send me a light for which you know the lumen output, and I'll predict it. Then you'd have an n=1 test of how much error is in the system. Of course, the calculation will do that too, it will give us an r-squared. I don't know what it will be but if it's not better than .7 I'll be disappointed. That would say I'm only able to predict 70% of the variance in the DV, by the IVs.

Yeah, CPF has been achy and breaky for me, too.

Last edited:

#### Robin24k

##### Flashlight Enthusiast
Re: How to Estimate Brightness with a Light Meter?

FL1 Standard light ouput ratings require an integrating sphere, so manufacturers should be measuring their products in the same way. However, due to manufacturing variation, some manufacturers understate the specifications to guarantee at least the stated spec. If you see nice round numbers (80 lumens, 200 lumens, 160 lumens, etc.), it's probably understated and your calculated result may be higher than expected.

The fact that it's non-linear is what's alarming. Your results will be skewed on the lower and upper ends, but since you're looking for ballpark values, I guess that's not an issue. If you do get significantly different values, keep in mind that it's only one sample and not a statistically significant indication.

It's nothing personal, but the reason I'm nit-picking here is because estimating lumens is dangerous business (plus, there's no correction for the fact that the light meter isn't calibrated for LEDs).

#### Bolster

##### Flashlight Enthusiast
Re: How to Estimate Brightness with a Light Meter?

Hey Robin, I want you to nitpick. I appreciate the feedback, and I don't have my ego invested in \$7 of plastic so it's OK to pick on it. But I don't think it's ready for the trash yet. Here's why:

The eye perceives brightness along a logarithmic scale. That's the curve you're seeing, that's bothering you. Graph it logarithmically on the Y axis and you get your straight line (with some error, of course). Trust me when I tell you that the regression program will handle the curvilinear relationship just fine. If there were a straight line without a logarithmic scale...then you should feel alarm, because someone would have either altered the laws of physics, or the laws of biology.

So, how big a ballpark? With the data I've got, the largest offender so far is that 25 lumen rating that should be about a 10 lumen rating. So my worst datapoint so far is off by 15 lumen. That one point is so bad, it could probably be legitimately labeled an outlier and deleted. That 120 lumen datapoint is off, in the opposite direction, it should be more like 140 lumen.

The fit of the curve gets more accurate (better able to predict) as you give it more data. If I had 50 flashlights graphed I'd be feeling pretty smug. As is, only 7. I could do a lot better than that.

Same problem as always...I NEED MORE LIGHTS!

PS: About the light meter not being calibrated. That might or might not be a problem. If the light meter is consistently off, let's say it has a harder time reading LED light, so it's always low...then it's not a problem, because the method I'm using adjusts for it. However if my readings are off more for some lights than others...then that's a problem.

If I have an intermittent problem, it would be indistinguishable from any other error, say, different mfgrs measuring lumens differently (at the emitter, out the front, whatever). But it would drag the r-squared down until the regression program told me, "your attempt to predict is crap." For this project, I'd consider an r-squared of .5 to be crap (really big ballpark), and I'd be disappointed with anything under .7 (regular size ballpark). If the r-squared gets to .8 or over, I'd be happy to use it as a guesstimation device.

Last edited:

#### Robin24k

##### Flashlight Enthusiast
Re: How to Estimate Brightness with a Light Meter?

The tube is actually a very good idea for making runtime graphs (might be taking an idea or two for my runtime test setup...). However, I was expecting your calculated lumens to be off by a constant since flux (lumen) depends on area. It shouldn't have anything to do with how the eye preceives brightness (lumens has nothing to do with "brightness").