Sizing up the ColorHug open source colorimeter [LWN.net]

Sizing up the ColorHug open source colorimeter

Posted Jun 3, 2012 3:41 UTC (Sun) by rahvin (subscriber, #16953)
In reply to: Sizing up the ColorHug open source colorimeter by rgmoore
Parent article: Sizing up the ColorHug open source colorimeter

Getting a light source that always provides the same spectrum is extremely difficult. The engineering alone is difficult and the quality control on the parts and assembly is also expensive. I think if you think about how hard it would be to get a light source that regardless of medium provides accurate color from reflected sources would be very hard. Precise white point is hard, its even harder when you are dealing with large temperature ranges, humidity levels and varying background lighting, source materials and every other natural thing that can get in the way of accurate light levels.

Commercial parts that provide what they are talking about are in the $500 range. Without commercial volumes the price could be 10x that much. Normal consumers don't generally buy these but they are essential in the printing and art reproduction business and doing the hardware right is very hard and expensive.

Great article, as someone who's done enough amateur photography that I was forced to buy monitor color calibration hardware and go through color calibration routines to ensure I got good reproduction of what I was looking at on screen. When I purchased though I wasn't aware that the commercial providers expect you to buy new hardware every few years because they abandon all support and driver updates. I can't afford to buy every couple years which left me having to either keep older hardware and software around to do photography on or to simply stop calibrating. I wasn't aware of the OSS options. Argyll made my older but still perfectly functional hardware usable not only in Linux but newer versions of Windows without being forced into another multi hundred hardware purchase. So thank you. I'll add my voice as someone that would be very interested in a OSS reflected light spectrophotometer that I could use for color calibration. Even if it was more expensive than the commercial options it would be better for no other reason than the hardware and software being OSS would be far less likely to be abandoned.

(Log in to post comments)

Sizing up the ColorHug open source colorimeter

Posted Jun 3, 2012 23:46 UTC (Sun) by rgmoore (✭ supporter ✭, #75) [Link]

Presumably you can avoid the issue of background lighting by (as Klavs pointed out) making the colorimeter thoroughly light proof and placing the print you're reading on a black backing so external light can't shine through and you don't get any stray reflections from anything behind the print. I'd think that a well designed LED light source would be pretty stable, though it might not be as stable as you need. The only cheap approach I can think of would be to provide something like a Gretag Macbeth Color Checker with the colorimeter and use it to calibrate the device before each use. That would have to be a lot cheaper than trying to provide a nearly perfect light source.

And I'll say, I've been very happy with Argyll. It not only lets you keep using your device after the manufacturer has stopped supporting it, it also lets you use it better than the manufacturer intended. In a lot of cases, the only difference between different models is the software. Argyll lets you use its most sophisticated software even on the cheapest hardware, which is a big bonus.

Sizing up the ColorHug open source colorimeter

Posted Jun 4, 2012 2:52 UTC (Mon) by rahvin (subscriber, #16953) [Link]

Yes, I wasn't aware of Argyll and now that I've used it with my Spyder2 I have to say I'm stoked that I found it.

Light tight source exposure is pretty much a requirement for these systems. The problems you run into is that you have to have very expensive light source. Most light sources change during warm-up and even as they age. You see this most with florescent lighting but even LED's have variations. Then you have to regulate power tightly because as with most light sources if the input power varies the light sources will change spectrum. Even a tiny shift in spectrum will significantly alter the results. Think about it, you have to have absolutely precise knowledge of the spectrum being produced by the light or you can't measure anything with accuracy. Given how light sources behave and their limitations it shouldn't be hard to see why this would be so hard to do right, and in this case the difficulty directly affects the cost. As I said, at commercial volumes these devices typically go for $500 for low end models. That includes some software costs but by far the hardware is the most expensive piece.

Sizing up the ColorHug open source colorimeter

Posted Jun 7, 2012 12:52 UTC (Thu) by slashdot (guest, #22014) [Link]

Can't you just measure using material of known reflective properties, then measure with the printed document, and divide the values?

That should cancel out the light contribution, and would seem to work with any light as long as it emits at decent power at all wavelengths you are measuring, and is constant over time as you switch the measured material.

Of course, the reference reflective material also needs to reflect at all wavelengths in question, and the spectrometer needs to be very high resolution, so that light and reflectivity over the atomic intervals of measurement are minimal.

It could probably work as some sort of tray supplied with the spectrometer where you put samples to measure on after initial calibration, with the spectrometer being a camera-like device on a fixed tripod pointed at it.

This would also be very user-friendly since you could just lay down the sample on the tray, and wouldn't need to touch the spectrometer at all.

I'm not really knowledgeable about this kind of hardware though, so maybe there's some issue with this design.

Sizing up the ColorHug open source colorimeter

Posted Jun 7, 2012 13:16 UTC (Thu) by slashdot (guest, #22014) [Link]

Hmm, there's a catch with this: if the shape of the BRDF of the reference and sample are different, this won't work, unless the light is the same from all directions.

Which probably means there needs to be an enclosure, and some way to have an "diffuse" light source, or a rotating one-directional light source taking with multiple exposures taken.

But in general, don't people want to know the full BRDF if they are bothering with spectral measurement/

So maybe something like a rotating laser in an enclosure or dark room or similar is the way to go.

Sizing up the ColorHug open source colorimeter

Posted Jun 7, 2012 19:47 UTC (Thu) by rgmoore (✭ supporter ✭, #75) [Link]

Can't you just measure using material of known reflective properties, then measure with the printed document, and divide the values?

This is what I meant when I mentioned calibrating the device. You have it measure a reference standard, ideally one with several different patches that are expected to behave differently so you can measure its performance better, and use that to figure out what your light source is actually doing.

and the spectrometer needs to be very high resolution, so that light and reflectivity over the atomic intervals of measurement are minimal.

I think that's exactly backward. You want to have an average reading for a decent sized patch that's supposed to be the same color. If you take your readings at too fine a resolution, you risk having your calibration thrown off by different behavior of the material over the scale you're looking at. For example, when you're measuring the output of your printer, you need to make sure you're looking at a scale substantially larger than individual ink droplets so you correctly measure the effect of its dithering pattern.