Printalyzer Densitometer - A compact budget-friendly densitometer project

[Mikolaj]

Hi @dkonigs I just got my Printalyzer Densitometer and after first test I loved it! Thank you and I'm waiting for Printalyzer Enlarging Timer!

 

[dkonigs]

To begin today's little update, here's a montage video I recently put together with clips from the process of assembling the first production batch of Printalyzer Densitometers:

(My goal with the video was sort of a retro-manufacturing-video vibe. I probably missed clips of a step here or there, and didn't keep it going through calibration and packaging, but this should show most of it. Didn't want to make it too long.)

I've also finally begun doing some initial R&D on what can be thought of as a "color densitometer" project. I'm probably going to call it a "spectrodensitometer", because the plan is to use a sufficiently capable multi-spectral light sensor. That sort of sensor will help majorly reduce hardware cost/complexity and will also allow me to simulate the response curve of pretty much any density "status" within its range. (Hopefully, unless people are measuring bandpass filters, this should be good enough to do the job of a typical color densitometer.) The sensor I'm currently evaluating for this is the AMS AS7343, and the hope is that I'll be able to share most of the same mechanical components with the current B&W model.

Now I must stress, there is absolutely no guarantee that this follow-on project will be successful. I don't even plan to make any sort of official product announcement or info page until I'm well into the testing and evaluation process with full hardware prototypes and have high confidence in the data. The biggest challenge with this sort of project is that there are far more elements that need calibration, including light spectrum response. And even then, I need to show that I can match the results from a more conventional color densitometer across a wide range of common materials.

So far, I've collected a lot of data from sensor evaluation boards and a spectrometer, and expect to collect more. The next step is building a couple of test rigs that let me evaluate a couple of different LED light choices with these instruments to see how they compare. I've found one particular choice that has a pretty good and broad emission spectrum, though I still need to decide which color temperatures to use for reflection and transmission. Keep in mind that its not a decision made in a vacuum, and what matters is both the LED's spectrum and the sensor's sensitivity across it.

In any case, one problem that keeps perplexing me is the case of yellow film. Or more specifically, the "Y-Patch" on C-41 Process Control Strips. Nearly all of the common old X-Rite densitometers seem to measure an unusually high blue density off this sample. (~2.30 for Fuji strips I've been using) Meanwhile, none of my other spectral measuring instruments show a density higher than ~1.50 in the blue wavelength. I do not see this strange behavior in the green or red ranges. I really want to get to the bottom of this, and would love to get a second opinion from something that isn't a 20-30 year-old X-Rite. A big challenge here is that different/newer color transmission densitometers are extremely hard to find (though I may attempt to get a newer Tobias).

My current theory is that something goes wonky with the blue transmission channels on those old X-Rites, but its hard to prove without more data. One thing I did notice, however, is that the difference between Kodak and Fuji Y-Patch measurements on my other tools is actually the same as the difference the X-Rite measures. (And I don't see this weirdness on the other patches of those control strips, or even on the Y-Patch of RA-4 control strips measured with reflection.)

P.S. Don't worry, I'm also working on the Enlarging Timer project as well. The current plan is to bounce back-and-forth between the two projects, now that I'm not spending as much time building the B&W densitometers. Its just that the spectrodensitometer project is a much easier "fast follow" because the hardware side is really just "take same design, change a half-dozen parts" and all the complexity is from testing/calibration/software.

 

[albada]

I would love to buy a modern color densitometer, especially if it handles both reflection and transmission. My old Tobias transmission densitometer uses a tungsten bulb, and I wonder what I'll do when it burns out.

 

[Joel_L]

@dkonigs

I just happened to be looking at used densitometers on ebay and saw this thread. Thanks for making all your work available, I think will be making one of these.

 

[dkonigs]

Hello again, everyone!

When it comes to follow-on densitometer projects, I've found myself reaching a bit of an impasse. So I'd like to present the situation, and see if anyone here has any thoughts on what the best way is to move forward.

First, as I've mentioned on occasion, I'm attempting to develop a color transmission/reflection densitometer with a similar design as the B&W one currently being sold. The key difference is that I'm using a multi-spectral sensor (AS7343), from which I should theoretically be able to extract the color channels.

This brings us to the biggest major stumbling block of the project... Actually translating the raw sensor data into spectral data, and then ultimately into Status A/M equivalent readings, is far easiest said than done. I have made some decent progress at getting "in the ballpark," as it were, but you really need to get a lot closer to get readings comparable to existing filtered-photodiode based products (e.g. the old X-Rite units).

There is a different multi-spectral sensor option that has many more channels (AS7265x), but it has lower light sensitivity and is actually a 3-sensor chipset. So it would need extra diffusion, which would reduce incoming light even more. In other words, it would be a losing battle to illuminate the film enough to actually read higher densities.

Alternatively, other companies make mini-spectrometer products (e.g. C12666MA). The possibilities of these are quite tempting to explore, but the cost makes them completely impractical. We're talking about $200-300 for just the sensor, so imagine what the overall device would have to cost.

Finally, you could suggest the good 'ole filtered photodiode approach. But this brings in its own complexities. Mainly that the necessary filters really aren't low-cost off-the-shelf products, it would be hard to use a single sensor ring for both reflection and transmission, and the electronics get more complicated. So its probably the hardest to actually implement as "the little guy."

Meanwhile, I have had a few people reach out to me and ask about UV densitometry. For anyone working with alternative processes, this is something that would be useful. There does appear to be a decent off-the-shelf sensor that would work (AS7331), and there wouldn't be a major spectral calibration problem to overcome. The biggest issue with this sort of project is actually figuring out what spectrum "UV" really means here, and selecting the appropriate light source for it. (Might have to acquire and characterize an X-Rite 361T to find out.)

So anyways, what I'm trying to figure out is what the best next step here is:

1. Continue to plug away at the spectral calibration problem for a traditional Status A/M color densitometer
2. Produce a color densitometer that approximates "red", "green", and "blue" based on the closest sensor channels, but does not actually follow any sort of Status standard for that that means.
3. Do something with a much fancier sensor, knowing that the end product would cost a fortune and be out of reach for many people
4. Switch gears, and develop a UV densitometer instead

Part of the challenge with color, is that the potential customer base is both individual users and labs, where labs likely have different requirements and might be less tolerant of an "approximate" product.

 

[calebarchie]

Perhaps now it seems like a good time to finish the timer, then return to this?

Just my 2c, good luck!

 

[dkonigs]

Perhaps now it seems like a good time to finish the timer, then return to this?

Just my 2c, good luck!

I'm actually working on both simultaneously right now. (Going to post an update in the timer thread soon.) That's part of why I'm kinda resisting any designs that require a lot of new and difficult engineering work for a follow-on project. I'd like to reuse as much of the existing hardware and designs as possible.

The ideal second model would use the same hardware, and only minor changes to the electronics. The challenge is picking an approach that doesn't also cause the software and calibration part of the project to explode in complexity.

One thing that makes the UV densitometer idea attractive is that very little would actually need to change. It would just be a new sensor, a new transmission LED, and some relatively minor firmware updates.

Of course the color densitometer idea really isn't much more difficult than that from the hardware side. Its just massively more difficult from the firmware/calibration side.

 

[albada]

Part of the challenge with color, is that the potential customer base is both individual users and labs, where labs likely have different requirements and might be less tolerant of an "approximate" product.

Another alternative is to tell labs that color is approximate. If that's good enough for users measuring colors on C-41 film and RA-4 prints, then I suspect that's good enough for most buyers.

Here's a simple approach: Use a monochrome sensor covering the visual range, and illuminate it with RGB LEDs having appropriate wavelengths. You would be measuring at only three points, instead of over three ranges, but that might be fine for individuals. Also, does the ISO standard require a D65 illuminant? That could be costly, unless you use filtered tungsten.

A similar approach that might work well: Cree offers LEDs in colors PC-blue and PC-red having wider spectra than typical LEDs due to phosphor conversion (PC). And PC-lime with a filter could serve as green. These might give you good enough ranges with a monochrome sensor.

 

[koraks]

There is a different multi-spectral sensor option that has many more channels (AS7265x), but it has lower light sensitivity and is actually a 3-sensor chipset. So it would need extra diffusion, which would reduce incoming light even more. In other words, it would be a losing battle to illuminate the film enough to actually read higher densities.

Yeah, I feel the problem. However, you could compromise by allowing for a longer integration time. It will make continuous measurements impossible, but that might not be the end of the world. I think in most applications it will be perfectly fine if a sample takes a second or so to acquire.

Here's a simple approach: Use a monochrome sensor covering the visual range, and illuminate it with RGB LEDs having appropriate wavelengths.

Yep, that thought has crossed my mind, too. The sampling aperture can be kept small by using a light duct that illuminates the measurement area on one end, and has the LEDs shine into it from a different end. Btw, the light duct may also come in handy for the three-chip AS7265 idea. I didn't make this up myself, btw - it's how LiCi solved the same problem back in the 1980s with their ColorStar analyzers. These feature perspex light ducts that snake from the sensing aperture on the probe all the way to the (big) photodiodes in the internals of the probe.

I'm not sure if this approach of wavelength binning through LED choice actually solves the problem, though. The thing is that the AMS sensors @dkonigs has been using effectively already give relatively clean wavelength-binned output. I wonder to what extent re-inventing that particular wheel will help.

Another alternative is to tell labs that color is approximate. If that's good enough for users measuring colors on C-41 film and RA-4 prints, then I suspect that's good enough for most buyers.

The issue that would concern me is that the calibration values as provided by the manufacturer will be off. I'm not sure if that might be a dealbreaker for a lab. They could work around it a little by benchmarking the new densitometer against a known standard.

The biggest issue with this sort of project is actually figuring out what spectrum "UV" really means here, and selecting the appropriate light source for it.

Yes, that's a bit of an issue. I've been doing some work on UV light sources lately, including a brief (and as yet, incomplete) foray into light integrator territory. Long story short: wavelength matters a ton, and in several ways. Firstly, there's the non-linear response of a UV sensor to different wavelengths. The (mostly) relevant range of UV would all be concentrated in the UV-A sensor of the part you linked to, since that covers the entire range of around 350nm up to over 400nm.

What makes things extra challenging is that different alt. processes have different responses to different UV wavelengths. Moreover, I don't expect UV attenuation of 'typical' negatives to be all that linear - a reading of UV density at 400nm will likely be quite different (and non-linearly so) compared to a 350nm measurement. And then there's the matter of light sources used. LEDs are relatively straightforward and mostly break down into 400nm and 365nm wavelengths with a fairly narrow bandwidth. But many alt. process printers still use all manners of UV fluorescents (with a broader spectrum, also depending on type used) and a few probably still hang onto to their space heaters HMI's

As long as someone uses the same light source for the same printing process and the same materials and methods for making negatives, their UV densitometry readings will be pretty consistent and reliable regardless of the above. But the moment one of those factors start to move, it's going to be a mess of non-linearity and non-intuitive outcomes, I bet.

Mind you, this is an inherent problem that's also not solved by the existing UV densitometers out there. It's a fundamental problem.

 

[sasah zib]

I'm actually working on both simultaneously right now. (Going to post an update in the timer thread soon.) That's part of why I'm kinda resisting any designs that require a lot of new and difficult engineering work for a follow-on project. I'd like to reuse as much of the existing hardware and designs as possible.

The ideal second model would use the same hardware, and only minor changes to the electronics. The challenge is picking an approach that doesn't also cause the software and calibration part of the project to explode in complexity.

One thing that makes the UV densitometer idea attractive is that very little would actually need to change. It would just be a new sensor, a new transmission LED, and some relatively minor firmware updates.

Of course the color densitometer idea really isn't much more difficult than that from the hardware side. Its just massively more difficult from the firmware/calibration side.

the market for color densitometer is small and becoming smaller. There may be more of a market in "reading" strips for the small-pro labs.

Also, the home occasional user will expect it to do more for them than it probably would. Consider all the questions asked about existing BW printing aids.


the UV user-base may be growing. Certainly there is the "curio" market segment -- those who will buy it to try it.
To find out more about these people, ask here: https://groups.io/g/carbon OR https://groups.io/g/altphotolist

 

[dkonigs]

I'm not sure if this approach of wavelength binning through LED choice actually solves the problem, though. The thing is that the AMS sensors @dkonigs has been using effectively already give relatively clean wavelength-binned output. I wonder to what extent re-inventing that particular wheel will help.

Yeah, it would make the transmission light a lot more complicated, and really just introduce some new problems. Especially if I can't find a line of mechanically-identical LEDs that are both bright enough and come in the exactly correct set of wavelengths.
The real problem is that the sensor channels don't exactly match Status A/M, not that I can't figure out what they do match.

As far as the illuminant, that's a part of the problem I've already put a lot of research into. The short answer is that (per the standard) my options are quite flexible), and there are white LEDs that cover my needs quite well. (The Seoul Semiconductor SunLike STWSC2PB-E0 series is good for this, and I've tested many in that lineup.)

What makes things extra challenging is that different alt. processes have different responses to different UV wavelengths. Moreover, I don't expect UV attenuation of 'typical' negatives to be all that linear - a reading of UV density at 400nm will likely be quite different (and non-linearly so) compared to a 350nm measurement.

This is actually true across the spectrum, and even for normal B&W. Here's a chart I found in a document describing how NIST does their step wedge calibration:

The dotted line is normal B&W film. (the solid line is X-Ray film)
This is why I'm highly suspicious of anyone who thinks color densitometers don't need separate calibration numbers for each channel. But once you get into the UV-A range, its obviously an even bigger issue.

The best approach for doing UV densitometry is most likely going to be to pick a sensor that covers the UV-A range, then select an LED that emits at whatever the "best" wavelength for doing this work is.

But how does one pick that wavelength? There in lies the main head scratcher of the whole project.

 

[koraks]

But how does one pick that wavelength? There in lies the main head scratcher of the whole project.

Certainly so. I'd be tempted to give an answer (365nm), but it would be a pragmatic one, and inherently compromised.

 

[dkonigs]

Certainly so. I'd be tempted to give an answer (365nm), but it would be a pragmatic one, and inherently compromised.

Yeah, that's the most obvious choice, if only because its the wavelength one of the sensors I'm considering has mentioned in its datasheet for the test values.

Another option is to get one of the existing UV densitometers off the used market and attempt to figure out what wavelength its measuring at. (they never say in the docs, but there are other ways of finding out)

I wish there was a "right" answer here, but the ISO specs for densitometry don't really cover the UV range.

 

[koraks]

I wish there was a "right" answer here

Yeah, although even if there was one along the lines of an ISO standard etc., I'd doubt its validity for the reasons I mentioned. Any choice of a single wavelength will have its drawbacks. If it's technically feasible, I'd lean towards a multiple-wavelength measurement. But it'll be complicated to implement.

 

[dkonigs]

So its been a while, and I've been spending the majority of my time these past few months on the darkroom timer project, but I do circle back to this from time to time. As such, I figure I'll provide an update on where my "future product" research has been going in the world of densitometers.

The "color" densitometer project is mostly on hold, but I'm continuing to experiment from time to time. I've pretty much come to the conclusion that a cheap multi-spectral sensor (e.g. AS7343) may have the low-light sensitivity to do the job, but really doesn't have the spectral resolution. Attempting to match ISO-standard Status A/M would require way too much calibration effort, and even then I may only get numbers "in the ballpark". This really isn't good enough.

Sensors with more spectral resolution do exist (from the AS7265x on the cheap end, to the Hamamatsu mini-spectrometers on the expensive end). But one common problem they tend to have is poor low-light sensitivity. I did some initial experiments with the C12666MA on a test rig, and while the results were promising, I very quickly hit the noise floor once trying to measure higher density targets. I'm planning to test with the C12880MA in the future, which does purport to have better low light sensitivity (but perhaps less dynamic range), and am curious what the results will be.

What's perhaps more interesting, is that I've decided to go full-steam-ahead with prototyping a UV-capable densitometer. For now, my working name for this project is "UV/VIS Densitometer". It'll have "visual" (Photopic) sensitivity for reflection and transmission, and UV sensitivity for transmission only (unless there's a UV reflection use case I'm unaware of).

The motivation for forging ahead on this project is the discovery of the new AMS TSL2585 sensor, which I'm also using for baseboard metering in my other project. In addition I'm also using different LEDs (which I may also run at higher power), and am also including (at least on the first prototype) a focusing lens and a temperature sensor.

I'm not yet sure if this project will end up being a second offering along-side my current B&W densitometer, or if it'll end up being "version 2" of that product with more capabilities. That really depends on side-by-side testing, given all of the changes.

For UV wavelength, that's still a bit of an open question. I've been doing a bit of research, and found some writings claiming that the X-Rite 361T measures UV in the 350-420nm range. Then I went ahead and actually measured a real 361T, and found its peak to be around 390nm.

Based on various conversations, and actual availability of UV LEDs, I think my real choices here are going to end up being 385nm or 365nm.
For what its worth, I'm probably going to build prototypes with both and see how they compare. (The 385nm LED can be driven with more current than the 365nm LED, but I really have no idea how much I'm actually going to need until I do some testing.)

Regardless, I'll probably be building the first prototype units in a week or two.

The main issue with any of this is really one of calibration. You can still get calibration strips for the X-Rite 361T, but that's it. So to really calibrate this, especially if I end up doing 365nm, I'm going to need another reference. To that end, I've custom ordered a set of optical filters with spectral data, which I'll have in a few months. Hopefully having that will finally give me a good "master" for all sorts of calibration tasks.

 

[Bill Burk]

UV reflection use case: Confirming the quality of / integrating exposures with UV exposing units.

For example the bumps on the cell of this meter allow sliding back and forth under a fluorescent bulb to check if it’s still burning consistently across its length.

 

[dkonigs]

UV reflection use case: Confirming the quality of / integrating exposures with UV exposing units.

For example the bumps on the cell of this meter allow sliding back and forth under a fluorescent bulb to check if it’s still burning consistently across its length.

Unless I'm missing something, that looks more like a use case for a UV-capable incident/baseboard meter, than a use case for a UV reflection densitometer.

 

[Bill Burk]

Unless I'm missing something, that looks more like a use case for a UV-capable incident/baseboard meter, than a use case for a UV reflection densitometer.

Right.

I don’t know what you could do with UV reflection densitometer. Determine fluorescence? That would take a UV emitter and regular light receptor. I guess it could help you know if a print paper has brighteners. Then it might not look as brilliant in different settings like a museum.

That Blak Ray meter is sensitive to a wide spectrum so if you are using it in daylight, you have to take a reading without yellowish UV filter and a reading with it, the UV is the difference.

 

[koraks]

I think my real choices here are going to end up being 385nm or 365nm.

Seems plausible. I'd lean towards 385nm because you'd still get a meaningful response with still very popular 390-400nm LEDs. If you pick 365nm instead, sensitivity at the longer wavelength may be relatively low, with detrimental effect on measurement resolution. 385nm is at this moment a relatively safe 'middle of the road' choice for a densitometer.

unless there's a UV reflection use case I'm unaware of

I couldn't think of one.

 

[dkonigs]

Seems plausible. I'd lean towards 385nm because you'd still get a meaningful response with still very popular 390-400nm LEDs. If you pick 365nm instead, sensitivity at the longer wavelength may be relatively low, with detrimental effect on measurement resolution. 385nm is at this moment a relatively safe 'middle of the road' choice for a densitometer.

Which LEDs are you referring to? The ones someone might use in their alt-process printing rig?

Though transmission of B&W film does sharply drop off as you approach 350nm, so I could see that as a motivating factor.

 

[koraks]

The ones someone might use in their alt-process printing rig?

Yeah, those.