A virtual emulsion?

[rmazzullo]

One idea I would like to share, as I hope it bears some fruit in the near future. I like to 'think outside the box' sometimes to see what might actually be accomplished. (Walk with me on this)....it appears that there are some basic emulsion making steps that are common to almost all emulsions that we are currently aware of - and using - on our level. It might be possible to create a 'template' in Excel that would provide a variety of 'controls' to adjust (or select) component type, amount, strength, addition rates, temperature, vAg, grain shape, motor speed, digestion time, ripening time, etc, etc that can be used to create a virtual emulsion, with the settings of each control providing a value of each component (and process step) to create a top down view of how the emulsion "should" turn out, with logging of each moment of each step. Now here is the fun part...it might be possible to use the excel spreadsheet to control actual servo motors, sensors, etc to actually get close to the emulsion you just 'designed' in excel....with actual feedback controlling the mix and recording everything - every misstep, success or failure, much like an aircraft 'black box'. If this is doable (and the sketch up I made to visualize this is physically well within reach) then we can design and create many variations of simple emulsions as we'd like. The drawbacks are that not everyone's hardware will be the same, neither are motors, house power, water quality, etc, etc. I am still trying to figure out how to enable users to create the same emulsion by sharing a 'make file', irrespective of locale. (If two people have the same recipe for a chocolate layer cake, with exactly the same ingredients and made the same way, there will still be two different chocolate layer cakes in the end.)

Now I know I am missing a whole bunch of pieces here, but my basic premise is that it might be possible to electromechanically control the emulsion making process - and have a log of each step (and whatever variations may occur) on a very small scale. From everything I have been reading and digesting (no pun intended) so far, this idea is very likely decades old and something somewhat similar had to be done by the big guys for process control for just as long. I haven't even included / approached coating issues with this as yet - this was just an idea concerning emulsion making. I know it is way cheaper to do all this by hand and just take really good lab notes, but....

And...what if AI could be incorporated to make it's own emulsion suggestions based on what it extrapolates from the data sets? Could be interesting.

These are just some random thoughts intended to start a discussion...I am not anywhere near the finished excel spreadsheet (yet). It may turn out that I am trying to re-invent the wheel...again.

I wouldn't be surprised if the Photo Powers That Be wished we didn't go where this could go. Maybe when I make and test a few dozen emulsions I will wish I didn't go there either.

Sorry for the blather. I hope PE isn't up there shaking his head...

Bob

 

[koraks]

I wouldn't be surprised if the Photo Powers That Be wished we didn't go where this could go.

I wouldn't worry about that. What you came up with doesn't change much about how the industry works and to an extend it's if course how the major manufacturers do this.

Your idea is certainly feasible. It's relatively straightforward to use a microcontroller (e.g Arduino) to control a heater, stirrer, some peristaltic pumps etc. People do this all the time for various applications. If you do a lot of emulsion making (eg for testing) it would be a sensible approach, especially if you're handy with electronics and embedded systems.

Btw, Excel would be OK although technically not the 'best' approach. But it'll work "in a pinch". Here's something I did about a year ago: https://tinker.koraks.nl/photography/getting-organizized-a-database-for-carbon-transfer-tissues/
It doesn't cover the automated mixing/weighing etc, so it's only about the first half of your idea really.

If you automate the actual emulsion making, you'll end up doing some things differently from a manual process. E.g. you'll probably forego the "noodling" and washing associated with manually made emulsions and instead rely on proper ingredient ratios that render this part of the process obsolete/unnecessary. (Fuji et al do no noodling, LOL!)

 

[Monomer]

I had a similar idea, and last week I built a mathematical model to predict the morphology, size, and structure of crystals by predicting their growth behavior. The only variable parameters are temperature and pAg. Written in Python. It's in its early stages and there are still many problems.

 

[Peter Schrager]

Following

 

[Monomer]

I had a similar idea, and last week I built a mathematical model to predict the morphology, size, and structure of crystals by predicting their growth behavior. The only variable parameters are temperature and pAg. Written in Python. It's in its early stages and there are still many problems.

Here's the good news. This is a crystal growth structure trained in ten hours by accessing a deep-learning crystal database.

 

[koraks]

I like the path you're exploring here @Monomer. Very interesting. It would be great if this could be augmented with some empirical data in the future, too, to confirm the validity of the models.

 

[Lachlan Young]

I wouldn't be surprised if the Photo Powers That Be wished we didn't go where this could go.

Given that modern highly computerised (data-driven and simulatable) emulsion techniques are covered in both Mowrey's and Shanebrooks' books, no. As a matter of fact, Ron's book covers some details about how to go about using computer controlled peristaltic pumps to make double-jet emulsions (essential if you want to form highly controlled crystal structures) with appropriate feedback controls (salt bridges and the like) - alongside the data that defines the formation of specific grain forms.

It's just that you'll never have access to the same volume of data or the same quality of analysis of emulsion data (e.g. are you able to arrange x-ray microdensitometry to weed out nonsense from fact in terms of real-world sharpness performance? - there are ways around this, but they people who know what they are looking for as a comparator - and systems that eliminate potential errors/ consistent misreading) that the big manufacturers have access to on a routine basis. Modern emulsion techniques are able to create essentially all necessary crystal forms without the use of certain environmentally/ personally unpleasant heavy metal salts (which seem to have been mainly responsible for specific crystal forms at precipitation - elsewhere, custom organics can replace them) - but it took several decades for manufacturing process control to catch up with what was research knew was feasible - and they already had a massive pool of data 40+ years ago to draw on that has informed products since. Today, you have as much computing power as it took to make high-aspect-ratio crystals reliably in your phone. But you don't necessarily have access to all the critical data. Which isn't to discourage, just to encourage a degree of realism between what you may be able to achieve with weeks/ months of effort and what a researcher at Harman (for example) may be able to achieve in a few hours.

 

[Monomer]

That makes a lot of sense. I’ve been struggling a lot with integrating quantum problems with deep learning—it’s been a huge challenge.

Oh well, guess that’s to be expected since my background is in biochemistry anyway!