DSLR digitizing project

Lachlan Young · Nov 16, 2020

grat said:
Just as a follow up on the original question.

After picking up a very inexpensive spectroscope, I can say that the light from my LED is continuous. There are no gaps (unlike a CFL or a regular florescent). Below 430nm it's pretty weak, and there's a small dip around 470nm-- not quite as much as reddesert's example in post #30, but noticeable. Then it's quite strong all the way out to about 680nm.

I went looking for a published LED spectrum that matched what I'm seeing, and found this one. For bonus points, it includes a standard halogen curve as well.

View attachment 259599
Image copied from Lumicrest.com web site

Having had a (very) brief play around with the file, it would be useful to know what the film used was, how old it was & what EI it was shot at. Secondly, I don't think there are any visual spectral discrepancies in the neg from the source, but I think that filtration of the source to 3200K might help to clarify if the higher blue output of the LED is doing anything untoward - I wouldn't want to draw any firmer conclusions until I'd had the chance to compare the two against each other. Finally, there are no obvious problems with the Canon sensor running out of dynamic range in terms of getting most of the detail out of the neg's highlights & shadow values.

grat · Nov 17, 2020

Fuji 400H, exposed in May at box speed. Film expires 2021-09. Expired film is an extra complication I don't need right now.

In theory, the panel was set to 5000K. I did an earlier run @ 3200K, but my masking was poor, and I wasn't happy with the result. Will try again. I don't believe I've got an appropriate filter handy.

I think I now understand what Nick Carver meant when he complained that scanning with the DSLR produced "Canon" images-- the resultant image is very much in line with what I would expect had I taken the photo with my EOS 90D in the first place.

Helge · Nov 17, 2020

grat said:
Just as a follow up on the original question.

After picking up a very inexpensive spectroscope, I can say that the light from my LED is continuous. There are no gaps (unlike a CFL or a regular florescent). Below 430nm it's pretty weak, and there's a small dip around 470nm-- not quite as much as reddesert's example in post #30, but noticeable. Then it's quite strong all the way out to about 680nm.

I went looking for a published LED spectrum that matched what I'm seeing, and found this one. For bonus points, it includes a standard halogen curve as well.

View attachment 259599
Image copied from Lumicrest.com web site

I’m sure you are familiar with the concept of smoothing, and this curve is clearly smoothed.

The small hump around 690 for example, is almost certainly a tall thin spike that has been smoothed out, that they couldn’t leave out without being totally dishonest.

The manufacturer could probably come up with dozens of reasons why that’s ok.
But the fact of the matter is, it matters where you have several filters with their own response working together.

Almost all white LEDs work with a combination of two or three phosphors pumped by a UV or other LED underneath.
The research during the last twenty years or so has concentrated on finding (sometimes digging out from research libs) phosphors that work together to have a better spectral response to the human eye.

There is a lot of talk over the years of nano dot LEDs and other exotica. But they have yet to materialize on the market.
It’s fine doing research. But if you, for whatever reason, can’t practically put the fruit of your efforts out on the market, it doesn’t matter a whole heck of a lot.

Fuel cell research is a classic example of that.
It has been five years into the future since the fifties. With numerous “breakthroughs” through the decades.

Phosphors response depends on the spectrum that pumps them. And even with ideal excitation (which LEDs are not) their response is not ideal and probably never will be.
It’s just physics.

Something better than black body radiator filament bulbs and flash will come at some point in the future.
But it’s not here yet.

The film is a filter, the sensor is a filter and the adjustments you apply in editing is a filter. They all interact in ways that are not always obvious.

Sure you’ll be able to “get away” with it.
You’ll get quite acceptable results initially, without nothing to compare against.
And you’ll have trouble with more “pathological” photos than you should.

Question is: Is “acceptable” good enough or are you looking for as good as you can do within reason?

PS. Using a spectrometer, even a simple one, well, takes preparation and care.
You need to narrow the source to a pinpoint or very thin line.
And your eyes while handy and available, are not ideal as sensors.
They are not that sensitive and tend to compensate and “cheat”.

warden · Nov 17, 2020

Lachlan Young said:
In a word: no.

Good.

grat · Nov 17, 2020

Helge said:
I’m sure you are familiar with the concept of smoothing, and this curve is clearly smoothed.

So's the halogen curve. A true representation of halogen's spectral output is pretty spiky as well.

Almost all white LEDs work with a combination of two or three phosphors pumped by a UV or other LED underneath.
The research during the last twenty years or so has concentrated on finding (sometimes digging out from research libs) phosphors that work together to have a better spectral response to the human eye.

Yes-- Philip's L-prize LED uses, if I recall, red and blue LED's to excite the phospor. But you're repeating yourself.

Fuel cell research is a classic example of that.
It has been five years into the future since the fifties. With numerous “breakthroughs” through the decades.

In GM's fuel-cell "ElectroVan" in 1966, the fuel cell took up most of the cargo room. Then they shrank it enough to fit on the Apollo modules. By 2000, it could fit in the underbody of the Precept, had 3 times the range, and had efficiency down to 1kW / kg. Now they're reasonably practical, if you can find a hydrogen fueling station. Along the way, they've reduced the need for platinum from "massive" to "negligible".

The progress has been, for the last two decades, more "evolution" than "revolution", although work is being done on non-platinum based catalysts, and that would be a major breakthrough. As recently as a few months ago, promising research has been published.

What that has to do with illuminating negatives... *shrug*

Sure you’ll be able to “get away” with it.
You’ll get quite acceptable results initially, without nothing to compare against.
And you’ll have trouble with more “pathological” photos than you should.

Hey! Sounds like the history of color printing to me.

Question is: Is “acceptable” good enough or are you looking for as good as you can do within reason?

So-- care to express an opinion on the sample photos, or would you merely like to continue down the hypothetical path of massless strings and frictionless pulleys?

How can the DSLR photo be improved? What, specifically, is wrong with it?

PS. Using a spectrometer, even a simple one, well, takes preparation and care.
You need to narrow the source to a pinpoint or very thin line.
And your eyes while handy and available, are not ideal as sensors.
They are not that sensitive and tend to compensate and “cheat”.

Now you're just being tiresome-- not to mention being disingenuous, since even a cheap spectrometer already has the aforementioned slit to narrow the light source. Otherwise, it's just a prism or diffraction grating.

Further, while the human eye may not be precisely calibrated, it's certainly sensitive enough for this purpose.

I'm sorry you disagree with my methodology, my testing, and me in particular, for all I know... But I'm not the first person to go down this path, and I won't be the last, and in spite of your assertion that it's impossible to get acceptable results, many people are in fact, getting acceptable results. You remind me of the people in the early 20th century who proved scientifically that if you drove over 30 mph in an open car, you wouldn't be able to breathe, and would suffocate.

Personally, my goal is to get comparable image quality to my Epson scanner with higher detail. I would say right now, I'm within a few percentage points of the Epson in terms of quality, and massively ahead in detail-- and the image quality is close enough that it's almost certainly down to personal taste.

Helge · Nov 17, 2020

grat said:
So's the halogen curve. A true representation of halogen's spectral output is pretty spiky as well.

Yes-- Philip's L-prize LED uses, if I recall, red and blue LED's to excite the phospor. But you're repeating yourself.

In GM's fuel-cell "ElectroVan" in 1966, the fuel cell took up most of the cargo room. Then they shrank it enough to fit on the Apollo modules. By 2000, it could fit in the underbody of the Precept, had 3 times the range, and had efficiency down to 1kW / kg. Now they're reasonably practical, if you can find a hydrogen fueling station. Along the way, they've reduced the need for platinum from "massive" to "negligible".

The progress has been, for the last two decades, more "evolution" than "revolution", although work is being done on non-platinum based catalysts, and that would be a major breakthrough. As recently as a few months ago, promising research has been published.

What that has to do with illuminating negatives... *shrug*

Hey! Sounds like the history of color printing to me.

So-- care to express an opinion on the sample photos, or would you merely like to continue down the hypothetical path of massless strings and frictionless pulleys?

How can the DSLR photo be improved? What, specifically, is wrong with it?

Now you're just being tiresome-- not to mention being disingenuous, since even a cheap spectrometer already has the aforementioned slit to narrow the light source. Otherwise, it's just a prism or diffraction grating.

Further, while the human eye may not be precisely calibrated, it's certainly sensitive enough for this purpose.

I'm sorry you disagree with my methodology, my testing, and me in particular, for all I know... But I'm not the first person to go down this path, and I won't be the last, and in spite of your assertion that it's impossible to get acceptable results, many people are in fact, getting acceptable results. You remind me of the people in the early 20th century who proved scientifically that if you drove over 30 mph in an open car, you wouldn't be able to breathe, and would suffocate.

Personally, my goal is to get comparable image quality to my Epson scanner with higher detail. I would say right now, I'm within a few percentage points of the Epson in terms of quality, and massively ahead in detail-- and the image quality is close enough that it's almost certainly down to personal taste.

The halogen curve is spiky like the suns. But not as spiky as phosphor LEDs.
It’s the relativist and continuity fallacy over again.

If I’m repeating anything it’s because you’re making me. ;-)

Fuel cells has to do with this topic because it’s a perfect example of your and billions of other peoples enamored relationship with the fiction of research and sciences steady progress and its inevitable results in a steady stream of better products.

Sure there is better ways to make fuel cells. But if producing them for the consumer isn’t practical, because it’s too costly, because there is the problem of a more well developed older product that’s shading the new potentially better one out of nourishment and market, or simply heaven forbid, patents in the way.
Then, the product will never see the light of day.

I can almost guarantee you that we won’t see fuel cells used in cars in the next decade either. And the problem is not pumping hydrogen. That problem was solved long ago.
Fuel cells don’t need to use liquid or pure hydrogen anymore.

LEDs have some fundamental problems. But sadly they have been declared “good enough” by half-schooled lobbyists and industry insiders eager to sell a product, and for political reasons. So it’s going to be a looong while before we see some real improvements. Or even alternative technologies.

I have no opinion on your samples because I have no reference, and I have no idea how you processed the image.
They look fine to me. But some questions you might ask yourself is, were they easy to get to where they are, and are you satisfied with them?

I’m being tiresome‽
“Toy” spectrometers come in whole variety of configurations. I have no idea how yours looks.
And the eye is not a particularly good sensor in this regard.
There is also a whole lot to say for training and experience of the scientific eye.
The eye very often see’s what it expects to see.
Vis-à-vis: https://zenodo.org/record/1429443#.X7QI5hbqHDs

grat · Nov 17, 2020

Helge said:
It’s the relativist and continuity fallacy over again.
If I’m repeating anything it’s because you’re making me. ;-)

Speaking of fallacies....

Fuel cells has to do with this topic because it’s a perfect example of your and billions of other peoples enamored relationship with the fiction of research and sciences steady progress and its inevitable results in a steady stream of better products.
...
I can almost guarantee you that we won’t see fuel cells used in cars in the next decade either. And the problem is not pumping hydrogen. That problem was solved long ago.
Fuel cells don’t need to use liquid or pure hydrogen anymore.

Full cell vehicles that have been sold / leased to the general public in the past two decades:

Yes, they're largely R&D projects. But they exist. And one of the problems is not PUMPING hydrogen, it's the actual infrastructure-- if you can't find a refueling point, you can't refill your car.

Same issue electric cars have been facing, but electrons are easier to move than H2, so the electric charging infrastructure is growing faster, at least in the United States.

Hey, you know what country has the highest density of H2 stations in Europe? Some country called "Denmark". Perhaps you've heard of it. Copenhagen apparently, as of last year, is operating 20 or more fuel cell powered taxis.

All of this of course ignores the real problem with fuel cells: you're going to an awful lot of trouble to use H2 as a battery. Skip the losses on conversion, and just charge the batteries. Yes, it's slower to recharge.

LEDs have some fundamental problems.

ALL light sources are compromises. But you appear to have a fundamental problem with LED's. We get it. You don't like them.

The eye very often see’s what it expects to see.

Indeed.

Helge · Nov 17, 2020

Yeah hydrogen powered cars have been with us a long time and the problem is as always wide spread infrastructure, but that could be solved.

That’s not what I talked about though.
I talked about fuel cell cars. And fuels cells in general.
If they ever become viable it will probably be in the direct methanol version.
Methanol while a little less efficient per liter is a lot easier and safer to distribute.

It’s easy to imagine a better lightsource than LEDs. The problem is not so much the phosphor as such, it’s more the exciter source.
Electrons has been suggested as a possibility. It would be possible to make a very efficient electron emitter with a cold graphene cathode.

Problem is we have something that is deemed good enough by “the man”.

Same with screen technology. In some ways our screens are still inferior to CRTs (colour, contrast and scaling to resolution).
But much worse, they are without comparison worse to read from than a printed page, even though each day people read from them as much or more than from print.
Image stability has been deemed good enough by the powers in the know, so it is not something that has first priority.
E-ink is as good as dead, and viable high refresh alternatives are not on the way.

Adrian Bacon · Nov 17, 2020

grat said:
No. It isn't. First, what you're describing would produce a one-sided flash, which isn't acceptable. Lining the box with foil and/or painting the interior white, would help, but most of the flash-based setups have relied on twin flashes.

Secondly, it would require raising either the height of the camera mount, or the entire Y-axis. Or, following the logic that seems to be prevalent here, cutting a hole in the table. The first two would decrease the stability of the entire system, and the third is just daft.

This is another example of a user on this forum asking a question, and getting a totally unrelated answer. Yes, I floss.

First, LED light *CAN* be notoriously uneven. This is known. But it doesn't have to be. Welcome to the 21st century.
A/C driven florescent lights have a much worse light spread, and I hate them with a passion-- but that wasn't the question either. This isn't a room lighting solution, this is a panel specifically designed, and used, for photography fill-lighting.

Given that blue light appears to be relatively important to our circadian rhythm, and that halogen is notoriously poor at that end of the spectrum, I'm not sure it's quite the solution you think it is.

Not the question either, but getting closer.

So there's a magic property in these dyes that keeps them from being recorded properly by a Bayer filter? That sounds implausible.

Yes! And that's the question. Not "What light source should I use?", but "How do I tell if the light source I'm using is sufficient?".

The light source I'm using appears to generate identical curves in the R, G and B curves of the histogram (actually, "curves" is misleading-- I get a relatively thin vertical bar in the exact middle of the histogram that goes all the way up). This suggests to me that the light is relatively uniform, and probably continuous (otherwise R G and B would have different appearances), but I know the histogram isn't exactly a spectrum analyzer, so I came here for a second opinion.

But I'm not trying to collect light with film. I already did that. I'm trying to collect light modified by film, and I'm trying to do it with an LED panel that appears to correspond almost exactly with the digital sensor I'm using.

At the risk of repeating other people and/or being a jerk, use a flash. No it's not more expensive or complex or significantly bigger. Get a simple small manual speed light or the light that @MattKing suggested, and trigger it via your camera hot shoe. You can literally do this with $15-$20 worth of parts to hook the flash up to your hot shoe with a simple manual sync cord. Get a Logan Electric light panel (like this one: https://www.bhphotovideo.com/c/product/32011-REG/Logan_Electric_750428_8_x_10_Desk.html), rip the light and electronics out of it, cut a hole in the side and stick the flash into the hole. Done. You can put some baffling on the inside to smooth it out if you need to, but otherwise, for not a lot of money, you have an extremely good light source that gets rid of a whole lotta other problems and it fits in roughly the same space as pretty much any other light you're going to use. Need to focus? Have a phone with a flashlight function? turn the light on and set the phone down on top of the light box. It works.

Or... you can go down the path you're going down. Totally up to you.

Adrian Bacon · Nov 17, 2020

grat said:
Fuji 400H, exposed in May at box speed. Film expires 2021-09. Expired film is an extra complication I don't need right now.

In theory, the panel was set to 5000K. I did an earlier run @ 3200K, but my masking was poor, and I wasn't happy with the result. Will try again. I don't believe I've got an appropriate filter handy.

I think I now understand what Nick Carver meant when he complained that scanning with the DSLR produced "Canon" images-- the resultant image is very much in line with what I would expect had I taken the photo with my EOS 90D in the first place.

I use a 90D for scanning. The 90D native sensor white balance is 5200K. That's the Color temperature where the red and blue multipliers are almost exactly the same and nets the lowest noise levels and most dynamic range.

grat · Nov 17, 2020

Adrian Bacon said:
At the risk of repeating other people and/or being a jerk, use a flash. No it's not more expensive or complex or significantly bigger.

Let's just say at this stage of the project, it adds unnecessary complication to what I'm working with. For reasons of simplicity (heh), I'm working with a platform that moves in the Y axis, and a camera that moves in the X axis (And the Z-axis, but that shouldn't be an issue as long as I'm focused, and the system is rigid enough the parts stay at a constant distance). The LED panel has no wires, is low-profile, requires no synchronization, and is in itself a light box with very even light diffusion.

And even with the low-profile, I'm intending to design a custom frame for the panel to sit in to reduce the overall system height by an inch or two.

Adrian Bacon said:
I use a 90D for scanning. The 90D native sensor white balance is 5200K. That's the Color temperature where the red and blue multipliers are almost exactly the same and nets the lowest noise levels and most dynamic range.

Now, that's good to know. First that someone else is using a 90D, and secondly that 5200K seems to be the magic number. Thanks!

Helge · Nov 18, 2020

If it was it wasn’t for likelihood that you’d call me tiresome and condescending, I’d say you sound like you should consider if you have some darlings to kill.
Unless the bearings on your rig is exceptional, you’ll need to refocus every time you move anyway.

Adrian Bacon · Nov 18, 2020

grat said:
Let's just say at this stage of the project, it adds unnecessary complication to what I'm working with. For reasons of simplicity (heh), I'm working with a platform that moves in the Y axis, and a camera that moves in the X axis (And the Z-axis, but that shouldn't be an issue as long as I'm focused, and the system is rigid enough the parts stay at a constant distance). The LED panel has no wires, is low-profile, requires no synchronization, and is in itself a light box with very even light diffusion.

And even with the low-profile, I'm intending to design a custom frame for the panel to sit in to reduce the overall system height by an inch or two.

Now, that's good to know. First that someone else is using a 90D, and secondly that 5200K seems to be the magic number. Thanks!

well, I use a 90D and Sigma 70mm ART Macro lens with a strobe for the light, and I digitize *a lot* of film with it, so if you have any questions, I’m happy to dispense information about what I do. The 90D with that lens is excellent, even if capturing the whole frame in one shot. It looks like you’re going for a pan and scan method. That can be made to work, but I think you’ll make the discovery that for most frames, the added time and complexity of having multiple shots per frame just isn’t worth it. I’m not saying there’s no value in doing that, it’s just that at 32MP per shot, the 90D with one shot per frame can already deliver quite a lot more resolution than what most outputs realistically need.

grat · Nov 18, 2020

Helge said:
Unless the bearings on your rig is exceptional, you’ll need to refocus every time you move anyway.

It's designed for CNC milling while holding a spindle running at 3000 RPM, so it's pretty solid-- each axis is on a pair of relatively short (200mm or so) 10mm steel rods, with four linear bearings supporting each carriage, driven by T8 leadscrews. The frame is 2020 t-slot aluminum, built using right-angle connectors, and the whole thing was squared up using a decent size quality square. The X and the Y axis should remain a constant distance from each other, so really, the issues are movement-induced vibration, and film flatness. The movement from accel/decel can be compensated for by a delay on the exposure, but that's part of the reason why I want to avoid unnecessary height-- more height, more wobble. And while there's no gyroscopic forces in the DSLR, it's considerably heavier than a spindle.

grat · Nov 18, 2020

Adrian Bacon said:
well, I use a 90D and Sigma 70mm ART Macro lens with a strobe for the light, and I digitize *a lot* of film with it, so if you have any questions, I’m happy to dispense information about what I do. The 90D with that lens is excellent, even if capturing the whole frame in one shot. It looks like you’re going for a pan and scan method. That can be made to work, but I think you’ll make the discovery that for most frames, the added time and complexity of having multiple shots per frame just isn’t worth it. I’m not saying there’s no value in doing that, it’s just that at 32MP per shot, the 90D with one shot per frame can already deliver quite a lot more resolution than what most outputs realistically need.

I already have an Epson v800. I can get 2400 PPI easily from it, and with a bit of post-sharpening (frequency separation is fun!), produce some very nice images.

The question is, can I do better with the DSLR for less? The answer is "No", because the DSLR + the Canon EF 100mm f/2.8 USM Macro (say that three times fast) I'm using is twice the price of the Epson, and that doesn't include hardware. In spite of it's age, the 100mm still delivers some fantastic images.

Doing some math, if I do 35mm negatives in one shot, I'm getting much better resolution than the scanner. Medium format negatives, I'm at around 2100 PPI. And then there's 4x5. Since I don't currently shoot much 35mm, without stitching, I'm doing worse than the Epson in every case.

Do I *need* 6000 PPI? Of course not! Will I ever take a photo so breathtaking, so stunning, so overwhelming that I need to print it 6 feet tall by 12 feet wide? Oh, I hope so. But I think I'd better get used to disappointment.

It's a fun project, though. And it's either that, or renovate my kitchen, and frankly, this is far cheaper.

Adrian Bacon · Nov 18, 2020

grat said:
It's a fun project, though. And it's either that, or renovate my kitchen, and frankly, this is far cheaper.

lol... I totally get that....

I reached the stage quite some time ago that I need fast and productive scanning with high enough quality results to cover most output needs. Yeah, the Epson can deliver more total resolution with larger negatives, but it still completely loses in the speed department. In all honesty, people have been making pretty big prints with digital cameras that have a lot less resolution than 32MP for quite some time, and I’d actually be hard pressed to complain about many of those prints in terms of fine detail or overall sharpness.

as a group, we tend to get stuck down in the weeds of “better, sharper, more resolution” and tend to lose sight of the fact that the minimum you need to look good on a print or via a display is significantly less than what many people realize, and that digital imaging systems surpassed that minimum threshold a long, long time ago. All that being said, I won’t say no to more resolution, as long as I’m not jumping through hoops to get it.

Helge · Nov 19, 2020

Adrian Bacon said:
lol... I totally get that....

I reached the stage quite some time ago that I need fast and productive scanning with high enough quality results to cover most output needs. Yeah, the Epson can deliver more total resolution with larger negatives, but it still completely loses in the speed department. In all honesty, people have been making pretty big prints with digital cameras that have a lot less resolution than 32MP for quite some time, and I’d actually be hard pressed to complain about many of those prints in terms of fine detail or overall sharpness.

as a group, we tend to get stuck down in the weeds of “better, sharper, more resolution” and tend to lose sight of the fact that the minimum you need to look good on a print or via a display is significantly less than what many people realize, and that digital imaging systems surpassed that minimum threshold a long, long time ago. All that being said, I won’t say no to more resolution, as long as I’m not jumping through hoops to get it.

Resolution has a way of manifesting itself even in reproductions of a size where it really shouldn’t be visible.

Without hopefully going into borderline mysticism or magical thinking, there is a certain feel of ease and smoothness to a high resolution image, even on the lowest of lows, Instagram.

24MP really isn’t that much. Neither is 32MP.
You only need to crop a bit to begin to see problems.
It certainly is adequate for a lot of things though.

Adrian Bacon · Nov 19, 2020

Helge said:
Without hopefully going into borderline mysticism or magical thinking, there is a certain feel of ease and smoothness to a high resolution image, even on the lowest of lows, Instagram

uh,huh... feel free to label me a skeptic...

that being said, yes, there is value to having a higher spatial sampling rate than your output, if you require 100% contrast response all the way up to the maximum sampling rate of the output. In the age of pixel peepers, that’s very popular. I’m personally of a mind that there are uses for more resolution, but there are lots of places where it just aesthetically is less than ideal. Have you actually seen a ~50MP portrait of a person taken with a high resolution lens? There is such a thing as too much resolution.

I’m also intimately familiar with the knowledge that how you handle the original sample data has a really big impact on the output. A way bigger impact than many people realize.

Helge · Nov 19, 2020

Adrian Bacon said:
uh,huh... feel free to label me a skeptic...

that being said, yes, there is value to having a higher spatial sampling rate than your output, if you require 100% contrast response all the way up to the maximum sampling rate of the output. In the age of pixel peepers, that’s very popular. I’m personally of a mind that there are uses for more resolution, but there are lots of places where it just aesthetically is less than ideal. Have you actually seen a ~50MP portrait of a person taken with a high resolution lens? There is such a thing as too much resolution.

I’m also intimately familiar with the knowledge that how you handle the original sample data has a really big impact on the output. A way bigger impact than many people realize.

All I know is I can make a reasonable guess whether something is a daylight 12MP iPhone shot, or a photo from a higher resolution camera with an equivalent focal length lens at a high f-stop, on Instagram where the resolution is around one megapixel.
The jury is out on how much oversampling is too much oversampling AFAIK.

Adrian Bacon · Nov 19, 2020

Helge said:
All I know is I can make a reasonable guess whether something is a daylight 12MP iPhone shot, or a photo from a higher resolution camera with an equivalent focal length lens at a high f-stop, on Instagram where the resolution is around one megapixel.
The jury is out on how much oversampling is too much oversampling AFAIK.

people zoom in and crop like crazy and apply all kinds of filters on top of the image all the time on Instagram. I’d be pretty leery of guessing the camera. You also might be seeing a difference between using the jpeg from an iPhone vs. a much higher quality jpeg from a raw file that’s been processed in LR then output as a jpeg for upload to Instagram. Again, how the original source data is handled makes a really big difference. Then there’s also the sample bit depth and noise levels to take into consideration, along with the lens used. There’s so much more to it than resolution that’s going to effect what you see.

any downsampling algorithm that is worth anything removes all spatial information that is a higher frequency than the sampling rate of the destination resolution before down sampling. Failure to do so introduces a whole host of image artifacts that are really visually unpleasant. As long as your source resolution exceeds what would be 100% contrast response at the smaller destination resolution, everything else being equal, you won’t see a difference. If it doesn’t look the same, then everything else is not equal.

Helge · Nov 28, 2020

Old Gregg said:
@grat One consideration that hasn't been mentioned here is that modern camera sensors "see" a much wider light spectrum than RA4 paper. CN emulsions were designed to hit the well-known spectral response of each paper layer, with little consideration for colors a paper can't see.

Using a high-quality "continuous" light source is a double-edge sword in that regard, as your CMOS sensor will be picking colors that lie outside of the original design envelope. This is particularly noticeable in the red channel, especially with "consumer" grade emulsions like Ultramax. Scanners get around this by using light sources and tuning their CCDs for those frequencies [1], but for cameras you'll have to do this with post-processing. Search the archives, Adrian posted his approach by shooting 18% grey cards and building contrast curves for each channel. That's probably the most comprehensive way to go about it, but I suspect he's losing emulsion-specific "charm" in the process.

[1] And this is why LED-based approach has more potential than flash. Not only it's more compact and ergonomic, but you can actually design your light source to be similar to a scanner's and limit the amount of "junk" light hitting the sensor.

Tungsten is the default.
That much is given.
That’s why halogen is not a bad choice.

You can however compensate when using flash with the right filters.
That’s why I’d recommend some kind of initial physical filtering no matter how you’re going to set it up.

Flash however offers the advantage of a tonne of light in a very small slice of time.
That allows you to stop down, diffuse and filter almost as much as you need/want, and takes vibration in the setup out as a factor.

Adrian Bacon · Nov 28, 2020

Old Gregg said:
Adrian posted his approach by shooting 18% grey cards and building contrast curves for each channel. That's probably the most comprehensive way to go about it, but I suspect he's losing emulsion-specific "charm" in the process.

Yes and no. My process has undergone a fair amount of revision over time. While in principle, I still do it the same way, I don't do it uniquely for each emulsion any more. I use the same per color channel tone curves and the same Hue, Saturation, and Value operations for every emulsion (unless I have good reason not to). In essence, I have one "digital paper" that gets applied to all C-41 emulsions. The only thing unique to each emulsion is the white balance. I have gone through each emulsion and generated a baseline white balance value, but also have controls so that I can fine tune it per roll to account for roll to roll variances. Doing this means that the emulsion to emulsion differences do come through quite a bit, if they differ from the baseline. I find that it's more true to what you'd get if in all analog land and only printed on one RA-4 paper.

Currently, the tone curves for each channel are calculated from Kodak's reference control strip, and the baseline white balance is generated from exposing a whibal card (approximately 70% reflectivity, roughly the same as skin) at 5500K and centering it to neutral so the skin tones look correct. The HSV operations I generated are based on Portra 400. From there, the contrast and color falls where it may relative to Portra 400.

fdonadio · Dec 7, 2020

MattKing said:
A halogen light dichroic colour head for an enlarger can make a great light source for digitizing.

I once thought about machining an adapter that would screw into the filter thread of a Micro-Nikkor 55mm and connect to my Durst 605 where the lens board goes. That way I could use a digital camera to scan film.

But then I realized that seems like overkill.

DSLR digitizing project

Privacy & Transparency

Privacy & Transparency