Printalyzer - Darkroom enlarging timer & exposure meter

[dkonigs]

Derek,
If you made a device that didn't switch mains (the enlarger), would you be able to ship it without having to go through all the regulations? This is such a great project and I'm thinking if the device only sent pwm signal or serial data (usb) to another simple small device that would actually do the switching...an arduino keeping it open source and easy to build. So we could actually buy the main device (the brain) from you and build the small one (great project for high school electronics/photography/"steam"). Then of course release, when all is clear, a fully functional device for the ones who want an all in one solution.

I'd still have to go through some regulations, but its my understanding that the process would be a lot simpler. Honestly, I really just need to find some sort of consultant to help walk me through this stuff (and I'm not entirely sure where to find one) rather than go simply off assumptions and hearsay.

The issue is that "just plugs into mains" really is the best choice for most users. The exception to that is people who want to control some sort of custom LED head, but even they likely still need a mains plug for controlling their safelights. (And yes, automated safelight control is a useful feature here. It makes it easier to not have them interfere with meter probe measurements.)

I am dead set against coming up with some sort of completely custom interface you'd need to kludge your own electronics project to use. (i.e. every single "Just use PWM" or "Just have some simple voltage output" suggestion.)

However, there is a standard for doing this that I have begun to explore. That standard is DMX512. While technically designed for "stage lighting," its a mature standard with tons of supporting hardware. You can literally just go on Amazon (or eBay, or AliExpress) and simply buy off-the-shelf relay boards and LED controllers that can speak it. You can also obviously build your own board that can handle it too, and there are reference designs out there to start from (or I could even provide one).

The main problem with DMX512, however, is that no one seems to make a simple "cheap enclosed box /w 1-2 switched outlets" as an off-the-shelf product. While you absolutely could build one by wiring together existing parts, you can't just buy one as a complete product. The closest you can find are these 4-channel beasts (bigger than the Printalyzer box itself) in the $150-200 range, or fancy 1-channel dimmers (covered with LEDs you'd need to tape over) for just as much money or more.

The secondary problem is that you'd need to provide a separate power supply for any DMX512 relay "contraption,", so there would be a tangle of power adapters behind all of this.

Of course it is tempting to consider just having a DMX512 output, and make that "simple switched outlet box" as a separate product.

 

[dkonigs]

Small 3- and 4-digit 7-segment LED displays for clocks are cheap.

Yes, small 7-segment LED displays are cheap. But more flexible OLED/LCD displays are not.
If I'm doing a B&W only device, I can probably manage with 3-4 digits. If I'm going a color-capable device, I'm going to need more.

It only needs one button: ZERO. The user will first put a white section of a test strip over the sensor, and press ZERO. After that, all measurements will be white-relative, which are the densities users want.

Once its setup, this is correct. Well, really, it should have two buttons. One to take a reading, and one to zero it. (Or a single button with short and long press actions.)
But to actually calibrate it, you're going to need a few more buttons. Calibration involves inputting target density values for a "high" and a "low" target, then measuring patches with those values. Its not a process you'll need to repeat often, as I expect the device to be more stable than older units that relied on tungsten bulbs and less-integrated photodiodes, but I want it to be possible for the end user to do it.

On the Printalyzer, the user will press its Enter button to enter the current reading. Or, add an Enter button to the densitometer, making it two buttons.

The "runtime" interface to the Printalyzer will be very similar to how it currently interfaces with the X-Rite and Heiland densitometers. Put it over the target, push a button to take a reading, and it sends a value out its USB/serial cable. The Printalyzer then sees this value and updates an input prompt screen with it.

Why use a color sensor? Use a mono sensor having a good green response, and illuminate the paper with green LEDs that are always on while the unit is on.
Calibrate it once in your lab. If sensors are decently consistent, that one cal will be suitable for all units. If not, you'll need a cal method in the factory, adding a few minutes to assembly-time.

I'm using a color sensor because it makes it easier to calibrate to the specific wavelength I care about. It also means I don't depend on specific/unique/rare products that may be hard to come by or require "big company" manufacturing relationships to be able to source. My goal is to be able to match the output of a proper densitometer as closely as I can, likely using the X-Rite 810 units I have as a "standard" for this.
Chances are that I'm always going to need some sort of calibration process as part of building the thing. I won't know how much device-to-device variation there is until I build a bunch of them and do comparison testing. However, the device will have onboard EEPROM where it can store that calibration, and the end-user should also be able to re-do the process off any standard reference.

An ATMega or PIC is cheaper than the STM32, and is well-suited for this light-duty chore.

While I know there are a variety of microcontrollers I could use in this project, its not really a major factor in overall cost right now. Also, STM32 is a big product line with lots of choices. The one I'm using here (STM32L052K6) does crystal-less USB and has some on-board EEPROM, so it helps cut down on external components. Its also cheap enough ($2-3 depending on quantity). Picking something else may only shave off a dollar or so, at the cost of commonality and familiarity. The areas where I'm more concerned about cost adding up are the "fluff" you might not think about. Things like the enclosure, cable assembly (if included), buttons, connectors, etc.

 

[DMJ]

I'd still have to go through some regulations, but its my understanding that the process would be a lot simpler. Honestly, I really just need to find some sort of consultant to help walk me through this stuff (and I'm not entirely sure where to find one) rather than go simply off assumptions and hearsay.

The issue is that "just plugs into mains" really is the best choice for most users. The exception to that is people who want to control some sort of custom LED head, but even they likely still need a mains plug for controlling their safelights. (And yes, automated safelight control is a useful feature here. It makes it easier to not have them interfere with meter probe measurements.)

I am dead set against coming up with some sort of completely custom interface you'd need to kludge your own electronics project to use. (i.e. every single "Just use PWM" or "Just have some simple voltage output" suggestion.)

However, there is a standard for doing this that I have begun to explore. That standard is DMX512. While technically designed for "stage lighting," its a mature standard with tons of supporting hardware. You can literally just go on Amazon (or eBay, or AliExpress) and simply buy off-the-shelf relay boards and LED controllers that can speak it. You can also obviously build your own board that can handle it too, and there are reference designs out there to start from (or I could even provide one).

The main problem with DMX512, however, is that no one seems to make a simple "cheap enclosed box /w 1-2 switched outlets" as an off-the-shelf product. While you absolutely could build one by wiring together existing parts, you can't just buy one as a complete product. The closest you can find are these 4-channel beasts (bigger than the Printalyzer box itself) in the $150-200 range, or fancy 1-channel dimmers (covered with LEDs you'd need to tape over) for just as much money or more.

The secondary problem is that you'd need to provide a separate power supply for any DMX512 relay "contraption,", so there would be a tangle of power adapters behind all of this.

Of course it is tempting to consider just having a DMX512 output, and make that "simple switched outlet box" as a separate product.

I'm familiar with DMX512 connectors because I did sound and lighting for concerts. For the things I build I use MIDI, even for non-music devices, it's a simple protocol, works over bluetooth and there is lots of hardware. I think there are devices that use MIDI with relays too.

Have you seen this:
https://www.adafruit.com/product/2935?gclid=EAIaIQobChMIt6u8-qGN8AIVPAytBh3ZJgs_EAQYAyABEgK6afD_BwE

I'm not familiar with the PoE protocol and it seems that this device uses a molex connector. Also USB?

 

[dkonigs]

Have you seen this:
https://www.adafruit.com/product/2935?gclid=EAIaIQobChMIt6u8-qGN8AIVPAytBh3ZJgs_EAQYAyABEgK6afD_BwE

Yes, I've seen that. It comes up every single time I try to search for solutions.
No, I'm not going to use it as my solution. Its big and clunky (for a single outlet), I would need two of them (there's only a single relay), and its worthless for anyone not using US outlets.

Plus, once I go through the effort to provide a bunch of robust switched-DC outputs on the back of my device, I might as well just do this right and use a proper protocol that can support all the possible use cases.

 

[DMJ]

Yes, I've seen that. It comes up every single time I try to search for solutions.
No, I'm not going to use it as my solution. Its big and clunky (for a single outlet), I would need two of them (there's only a single relay), and its worthless for anyone not using US outlets.

Plus, once I go through the effort to provide a bunch of robust switched-DC outputs on the back of my device, I might as well just do this right and use a proper protocol that can support all the possible use cases.

I hope you find a solution soon , we want the printalyzer !

 

[dkonigs]

So its probably time for another update on this little "densitometer side project."

First, I've now built my second set of prototypes around the AS7341 spectral sensor:

And I'm now running my 3D printer a lot and accumulating a lot of varied prototype assemblages across my desk:

Keep in mind that the purpose of these prototypes is to prove out the "core hardware" (sensor head shell, light source, sensor, microcontroller, data processing, USB interface) and not anything to do with the actual user interface of the final device.

Thus far my results have been promising, but I'm far from done with this stage. The two sensors also lead to two very different approaches at how I'm going about solving the problem of getting correct density readings. So I figure I might as well attempt to briefly explain each one:

TCS3472 (RGBC Light Sensor)

• Use some form of a "lux equation" to transform the sensor's readings into a scalar value that represents the amount of reflected light from the target.
• Take the logarithm of this "lux" value.
• Use measurements of a known calibration target (e.g. X-Rite Reflection Standard calibration plaque) to find the relationship between these readings and density values
• Calculate density values (its a simple 2-point line equation at this point)

This works fairly well for "visual" (B&W) density from what you'd consider "paper base" all the way to around D=1.70. Beyond that, it diverges a bit and struggles to give higher density readings. However, this divergence is predictable and compensating for it may simply be a matter of characterizing the response of the sensor across that portion of its range.
Another challenge is figuring out how to tune the "lux equation" to center its response on the actual center of "visual" density (570nm) rather than the center of "lux" (555nm). I have some crazy ideas for measuring the response of the photodiodes in my X-Rite 810 as a comparison source for doing this, as I obviously cannot use a normal lux meter for the purpose.
Finally, while this approach will work for normal B&W paper, it probably won't work so well if the paper has any strong color tints.

AS7341 (Spectral Light Sensor)

• Go through a process to normalize the sensor channel's readings across its range
  
• Use one of several methods to get per-wavelength response from the target (either some form of interpolation, or one of several large data matrices provided with the sensor's evaluation board software packages), across some large portion of the 350-770nm range.
  
• At each wavelength, find the ratio of the reading from an ideal "perfect transmitting diffuser" (or as close to one as I can get) and the reading from the target I'm actually measuring.
• Do some math between these values, some data tables provided with ISO 5-3, and take the logarithm of the result.
• At this point, I actually do have the density value (or something close to it). However, I'll adjust it based on readings from a calibration target to hopefully compensate for any errors.

This isn't giving results quite as good as the TCS3472 for my initial experiments, yet, but with more work on calibration and data processing I should be able to get there. However, its far less confused by colored targets, and should be able to handle "Status A" (color density) in addition to "Visual" (B&W). The data I've gotten so far is very encouraging, though obviously there's a lot more processing work involved. My favorite part about this method is that it actually allows me to follow the process laid out in ISO 5-3:2009 and "do it right".

In both cases, I think I'm now at the point where I have the "fundamentals" of the process working, and its mostly a calibration problem.

 

[albada]

TCS3472 (RGBC Light Sensor)

• Use some form of a "lux equation" to transform the sensor's readings into a scalar value that represents the amount of reflected light from the target.
• Take the logarithm of this "lux" value.
• Use measurements of a known calibration target (e.g. X-Rite Reflection Standard calibration plaque) to find the relationship between these readings and density values
• Calculate density values (its a simple 2-point line equation at this point)

Here's an idea which I think would make the above procedure more tolerant of color tints: (1) Do all of the above steps on each channel, yielding four densities; (2) Select the highest density relative to the corresponding paper-white.
This per-channel method would cause you to measure the wavelength-range that is reflecting the least light relative to that channel's paper-white.

 

[dkonigs]

So all of my research on this "Printalyzer Densitometer" project (as I guess I should call it) has been going quite well. I'm currently at a point between testing various bits of hardware, but all of my experimentation has yielded extremely promising results. While I'm not yet finished, I'm now quite confident that I should be able to pull off a B&W reflection densitometer that narrows in on the correct spectrum. (I'll get more to the details near the bottom of the post, if you're curious.)

But this has made me start to wonder... What should I fully develop into a "sell-able/distributable product" first? Should it be the main Printalyzer Enlarging Timer/Meter, or should it be the Densitometer? Right now the main Printalyzer is a fully functional prototype, actually sufficient for me to use in my own darkroom. But bringing it to market adds a huge amount of additional work. The densitometer is far earlier in its development, but is far simpler with fewer components.

Changing course and doing the densitometer first has a few advantages:

• Its a far less complex device to assemble, once all the details are worked out
• Certification issues will be far less of a burden, so I can confidently distribute prototypes to people (all low voltage, powered via a USB cable)
  
• Component shortages are less likely to affect my ability to build it
• It'll be a much cheaper device, which may increase the number of people willing to actually buy one (even if only to play around with it a bit).
• For all the above reasons, its a much easier/cheaper way to "get myself started" in doing all of this before I jump head-first into a larger and more expensive device.

Now as I think about how to do this, there are basically 3 different device designs I could take:

1. Simple reflection-only "puck," similar to what you see in the above photo (but with a screen and more buttons). Would be as compact as possible.
2. Reflection/Transmission device ("stapler design", similar to the X-Rite handheld units in shape), but still compact. Probably would only be able to fit around 4x5" (or maybe 5x7") film at the largest.
3. Full-bore Reflection/Transmission device large enough to wrap around 8x10" samples, either by simply being larger, or by being some sort of add-on apparatus that holds option 1 in a bracket and provides a transmission light source below it.

I'm currently leaning towards option 2, for a couple of reasons:

• People probably would rather have reflection and transmission, and I'm not sure how wide the appeal of a reflection-only device would be (please correct me if I'm wrong).
• The most likely use cases for this device are film and paper testing, where the measurement area is in the form of a test strip or a step wedge contact exposure (or a large image feature on a 4x5" negative).
• The moment someone wants to measure the tonal range of a photograph on an 8x10" negative, its not so unreasonable to expect them to be at the point where they'd be okay simply buying more expensive devices that already exist.

The way I see it, I'm basically competing against two options that already exist:

1. A used X-Rite "clunker" off eBay, which sells typically for $300+ and may be in questionable condition. Might feel like overkill as a "for the heck of it" option.
2. The Heiland TRD-2, which sells for about $800+ (give or take), while being potentially less capable than those X-Rite clunkers (but new and with fewer issues). Way too expensive for most people to justify.

Not yet sure what a realistic price target for me will be, but I'd love to be able to make something in the $50-200 range. (I know that's pretty wide, but the biggest variables are the cost of the mechanical bits and the calibration materials.)

Any thoughts?

Oh, here's a picture of an light/sensor test rig I recently built, to make it easier to test more combinations with the eval boards for the sensors:

P.S. I should also mention that doing color (Status A/M) is probably going to become back-burner for now. Its going to take a lot more testing to be certain as to whether the AS7341 sensor can actually pull it off with the right amount of data processing (since its spectral channels don't line up exactly with that I need), or if I'll need another approach there.

So speaking of sensors... One discovery I recently made is that Color paper (Kodak Endura, as seen in the IT8.7/2 target above) and B&W paper have completely different reflectance behavior as you approach the near-infrared. This throws off measurements quite a bit when I use the TCS3472 and try to measure anything past gray patch 18 on that IT8 target. One solution I plan to implement for this is a proper IR-UV cut filter in front of the sensor. (Found a source, its just taking forever to get here from China so I haven't tested it yet.) I've also done a lot of spreadsheet fiddling, and am now planning to add the TSL2591 to the list of sensors I may use for this. Apparently, if I put an IR-UV filter in front of that sensor, and subtract its second channel from its first, the response curve I get is very close to the official curve for ISO 5-3:2009 visual density.

 

[albada]

I vote for number 2.
Compact is okay because a test-print or test-film can always be cut down with scissors.

 

[dkonigs]

I vote for number 2.
Compact is okay because a test-print or test-film can always be cut down with scissors.

Yeah, and nearly any test strip or step wedge exposure is going to fit anyways.

Just for reference, here's a mockup of the general design I have in mind:

Of course assume there will be a small display and a few buttons on top, and a USB cable connecting to the back. However, I haven't yet figured out exactly which layout/orientation would work best. Its probably going to take a few iterations of prototyping to figure this all out. (Thank goodness for 3D printers.)

 

[koraks]

How about a combination of 2&3? Make the light source and probe modular, and allow them to be clicked into a small or a large holder. You could engineer the small one first and then if the need arises make a larger holder.

 

[Bill Burk]

I think a transmission densitometer is the first thing anyone needs. Stapler design sounds good. For reading test strips, the edge within an inch of the side is all that’s really important.

For reading actual negatives then you need to allow reading to the middle of 4x5 or larger. You could make the stapler unfold and let the user provide light source like a phone or tablet.

It must not scratch the film or leave a mark. So no sharp edges and nose must be smooth or soft but not sticky. If rubber, it needs to be high quality that lasts forever not be vulnerable to ozone.

 

[dkonigs]

For reading actual negatives then you need to allow reading to the middle of 4x5 or larger. You could make the stapler unfold and let the user provide light source like a phone or tablet.

I actually considered the idea of option 1, with the ability to use something like a light table as a light source for transmission measurements. In theory, this is the best of all options. Simplest and cheapest device to make.
The problem is that a light table may simply not be bright enough to cover the full transmission density range. (I tested my Kaiser Slimlite Plano plugged in, on max brightness, with the CAL-HI patch from the X-Rite transmission calibration reference. It barely registered on the sensor.)
(And I don't trust the quality, consistency, or stability of using a phone or tablet as the source, though I expect it to have similar issues as the light table.)
The light source Negative Supply sells probably is bright enough, but by that point you might as well just buy a more expensive device that's already on the market.

It must not scratch the film or leave a mark. So no sharp edges and nose must be smooth or soft but not sticky. If rubber, it needs to be high quality that lasts forever not be vulnerable to ozone.

Its probably going to be some form of hard plastic, with slightly rounded edges, like everything else out there. However, does anyone have any specific suggestions as to what type of plastic would be best for this purpose? Ideally, I can just use a single piece of something for the entire bottom of the unit, as I'd rather not have to build the sensor head as a separate component. (Likely 3D printed for prototypes, though I have more options once I'm building many of them.)

 

[Bill Burk]

You can have the built-in be full range and unfolded limited to whatever the light source can deliver. If a phone can get you to 2.0 that’s good enough (for silver gelatin, alt processes may require higher density readings).

 

[MattKing]

Long stapler:

https://www.cleanitsupply.ca/produc...RBy3al-4neJBSwPCOPxnkfPrTpJIbEpRoCgiIQAvD_BwE
In case re-purposing might work for you.

 

[albada]

If plastics are so hard that they scratch negatives, consider molding a circular groove into the top and bottom, and put an O-ring in each.

 

[ic-racer]

If plastics are so hard that they scratch negatives, consider molding a circular groove into the top and bottom, and put an O-ring in each.

Good idea. The densitometer I use has that, others too.

 

[dkonigs]

Good idea. The densitometer I use has that, others too.

I'm not so sure of that. A rubber gasket around the stage plate may be common, but that doesn't mean it actually contacts the film.

Let's look at some actual photos of this mechanism from the two densitometers I have...

First, here's the stage plate and sensor head from the X-Rite 810:

The stage plate is a disc of opal material, surrounded by a ring of metal. The rubber you see is just to seal between the top of the assembly and the surrounding material. It doesn't contact the film. The head itself just has a flat bottom, though I'm not entirely sure what material it uses.
The design of the X-Rite 810 has the reflection sensors in the head, while the transmission sensors are below that opal disk in the stage plate. There are lamps in the middle of both sides, but I think the actual measurement process only uses the lamp in the head.

Now the newer (and popular around here) Heiland TRD-2:

The stage plate here is just a tiny (aperture-sized) circle of opal material, surrounded by a ring of black plastic, surrounded by a larger circle of material. (They make these with different aperture sizes, by simply changing the diameter of that dot in the middle.) Meanwhile, the sensor head is just a plastic shell with beveled edges on the bottom.
The Heiland design uses a single sensor for both reflection and transmission, but with different light sources for each. The light source in the head is used for reflection, and the light source under the stage plate is used for transmission.

 

[ic-racer]

The o-ring is around the sensor, to make a light-tight seal to the emulsion.

 

[Bill Burk]

Here’s the nose of a Macbeth TR524.

Amazingly the rubber shows no signs of deterioration. The generic o-rings from the hardware store rot in a month here.

 

[albada]

This thread has been quiet. Any update on the Printalyzer or the densitometer accessory?

 

[Arthurwg]

I am astonished. Simply amazing. And all this to make an analog print? Still, some folks manage to make a decent picture using test strips.

 

[dkonigs]

This thread has been quiet. Any update on the Printalyzer or the densitometer accessory?

Yes, actually. I should probably fork off a separate thread, and/or finally write a blog post about it. But basically, I've been heads-down moving from the "development rigs" (shown above) into something closer to a real product prototype.

Here's a picture of some of the first prototypes of this version sitting on my work-bench, mid-assembly:

I still expect the "enclosure" to go through a few iterations before its finalized. The current version is a more sloppily-built CAD model, but I have a much cleaner and easier to tweak CAD model I'm going to be making the next version from. I'll adjust it based on experience fiddling with this one. The biggest mechanical concerns right now are the height of the cover, the height of the buttons, how well the open-detect switch makes contact, and the strength of the torsion springs in the hinge. I'm also going to need to experiment a bit with materials, especially if I want to get a "professionally 3D printed" version. (Apparently some 3D-printable materials are transparent to IR, even if visually black, so I need to be sure to not get those.)

Functionally, this version has a few notable changes from the versions I've discussed above:

• It has both reflection and transmission light sources now
• There are four buttons for user input, plus a fifth hidden button to detect when the hinge is opened
• There's a small display on top
• I've switched to the TSL2591 light sensor, as it seems (on paper) to be the simplest and most direct way to approximate the proper spectrum for "visual" density when filtered and post-processed correctly. (I still may revisit the AS7341 in the future, or something like it, if I attempt to tackle color.)
• The sensor now has both an IR cut filter and a diffuser material in front of it.

My next step is working on writing all the software for this, with a bit more discipline than the "scratch code" I wrote for the earlier prototypes. At some point in this process, there's a good chance all the files for this project will get pushed up to GitHub.

There are still a couple of different approaches to calibration and calculation I plan to try, and implement in an automated fashion. Both in terms of getting the most "normalized" sensor data depending on LED brightness and gain settings, and in terms of how I go from that data to a density reading. I've tried a lot of different things during the previous phase of the project, and several of them do work, but I need to be automated and nitpick for the best results.

 

[albada]

Out of curiosity, what kind of display did you select? I searched Mouser a few weeks ago, and all the sophisticated displays cost too much for this accessory. The cheapest displays I saw were for clocks, but they are probably too restrictive.

 

[koraks]

Look for ssd1306 oled displays. They come in small sizes generally, and are cost effective.