Do you happen to have any insight into the processing of the film? I'm very curious as to how you'd process the plates into the specific colors needed for this to get anything close to life-like representations. I imagine these representations of the photos were taken from the raw negatives.

Judging by some of the artifacts i saw in the images i'm not even sure it was film. I think it may have been glass plates that were used for these. In a few there are areas where it looks like the substrate had flaked off of the negative when they made the positives.

Given that wikipedia says that film was starting to take over around this era I'd say that it's quite possible that it was all photographic plates since to do this well with something you'd have to have a good amount of experience with it and I wonder how much film and the processing would have reached Russia yet by then.

They were black and whites taken with color filters over the lens (R, G, B).

Project each of the three in its corresponding color and overlap them to produce a full-color photo. Overlap them in that color in Photoshop to produce the digital images you see here.

How would you project them in their corresponding colors, though? I'd imagine you would have to produce prints in the colors. Only other thing I can think of is to project the negatives with R/G/B colored light behind them and make those projections line up.

Colored lanterns is how he projected them in color, according to the article.

Thanks, completely missed that.

> Of course they look better.

Why?

Three reasons that I can think of:

1. These images are likely captured on large-format glass plates, not tiny sensors. Therefore, the amount of magnification required is much less (these may be nearly 1:1 on a standard laptop display).

2. Chemical emulsions typically have much more dynamic range than digital sensors, allowing for more highlight and shadow detail.

3. Capturing colors in three separate exposures gets rid of the need for de-mosaicing.

Because film captures details in molecules while a CCD sensor in your digital camera uses picture elements much, much, much larger than molecules.

Or, because it's analog. It's very similar to sound recordings. Whether magnetic or not (say, in vinyl) the resolution of the captured signal can approach the molecular. Digital has a discreet resolution limited by the capturing equipment; for example, CDs reproduce the volume in a sound recording only 44,100 times per second. But a magnetic tape has a resolution limited only by the number of molecules passing the recording head every second (the more [i.e. the faster the tape], the better.)

The spacial resolution of analog photographs is not at the scale of individual molecules. It is limited by the "grain" of the film [1]. Other factors include distortions from lens shape imperfections in the lens material.

Audio tapes record their signal in magnetic domains. These are areas of tape the are large enough to be written and read. Much larger than anything approaching the scale of individual molecules.

Sound recordings on vinly are similarly limited by other factors. To "read" the signal from a record groove a needle is dragged a needle through a v shaped groove. This needle jiggles up and down and left and right, hence the two channels in stereo. The time resolution of the signal (max frequency) is limited by the physics of the jiggling head. It's inertia prevents it from reading signals over a certain limiting frequency. Again, well before the molecular level.

[1] http://en.wikipedia.org/wiki/Film_grain

[edit] - Changed momentum to inertia in the last paragraph.

A couple points to add to that: with vinyl, the recording process itself (for which there are several techniques) also adds noise. The recording head (in whatever form) is also subject to physical laws. The materials used at each stage of the process are also not ideal substrates. Hell, vinyl is a polymer, which means it's a big tangle of chained molecules. Even if everything else were perfect, it would not be possible to write data at a molecular scale into such a medium.

I figured my layman explanation would suffice. While we're in the realm of correct details, it's also worth nothing that with each use of a needle dragged through a groove (or even with magnetic tape dragging across all the bits in the tape machine) the surface is degraded and the original recording along with it. Less fidelity, more noise ... hiss, pop, crack

Digital gives us error correction so that even when the physical media starts to degrade (taking the signal with it) our firmware and our software work out where the mistakes exist and attempt to correct them. Of course, once the signal has degraded beyond the point of recoverability, you "suddenly" have a catastrophic failure. (And this is why backups are important...)

We're actually approaching, if not have passed, the resolution of standard film with some higher end SLRs [1, 2]; at least in a total-usable-pixels perspective; but of course at the sub-pixel level, we're still very far off with only 14 bits per channel. There's also the question of how the total pixels are arranged, since each bit of the sensor is only absorbing one color, etc; but in time, at least as far as resolution is concerned, digital will or has surpassed analogue.

[1] http://www.mindspring.com/~lorqvonray/Sticker-5mp-Velvia.htm...

[2] http://www.clarkvision.com/imagedetail/film.vs.digital.summa...

Some ludicrous camera resolutions. Please not I'm trying to bring up cameras that are full-frame and have ludicrous sensor resolutions, so the arguments of "light doesn't work right at these densities" doesn't fly as easily, since the density is still relatively low:

http://imaging.nikon.com/lineup/dslr/d800/

http://www.usa.canon.com/cusa/consumer/products/cameras/slr_...

Ask any working professional photographer and they'll tell you we passed the practical resolution of film (35mm vs 35mm) years ago, and with the insanely low noise levels, we passed the dynamic range potential years ago too.

With the latest cameras in hand, you can capture sharper and cleaner photographs than the best 35mm film could ever manage, and you can do it with a tiny fraction of the light (1/16th the amount, or 4 stops, or 100 ISO film < 1600 ISO digital).

What you're seeing with the latest crop of pro cameras are 35mm sensors that are competitive with the best medium format film.

Maybe 35mm, but I'm still not sure have we passed the quality of medium and large film sizes. An old Hasselblad can still take pretty good pictures.

I didn't say it had passed, I said it was competitive... there may be some edge case scenarios where medium format film is still justifiable.

But when you factor in the convenience and availability of products like the D800, 5D3 and 1Dx, and combine that with the cost and quality of compatible glass, medium format starts to look like a marginal call indeed.

(Bear in mind, medium format DIGITAL is another thing altogether. That stuff kicks the pants off any of the pathetic baby toys made by Canon and Nikon.)

I probably mostly hear from purists and those people that don't want to look at the facts because "records are better than CD's" and similar arguments. I didn't know they were competing with medium format film now. That's .. saying something. o.o

Which is so funny because when music is produced digitally (99% of the recordings in the past decade) a CD can be a bit-perfect copy of the original mastering.

A vinyl record is a stamp of a stamp of a stamp of a scraping of the original master. Better than a CD? Evidence of complete mental breakdown, more like.

This always makes me wonder whether there isn't something about the imperfections from these analog copies (not to mention various analog playback mechanisms) that 'adds' to the richness/perceived value of the music... Perhaps on some level some percentage of us prefer to know that we have a copy of a copy, or some level of imperfection, in our version that we play, vs a straight up 100% clone

That's completely fine.

What I object to is the unqualified use of 'better'. You can't claim that vinyl is somehow 'better' when what you mean is that vinyl is perhaps more 'satisfying' on a personal level.

I personally disagree; vinyl's imperfections irritate me just as much as a low bitrate MP3 does.

To me vinyl's upsides are the large cover artwork, and the lack of an easy 'skip' button. Less exiting is the knowledge that every subsequent play will sound worse than the last, even if by a tiny bit.

i agree with you about unqualified use of better - I think it applies to lots of areas of technology, and even areas of culture, etc... It would be nice if people were more understanding that a person's opinion on something is in fact influenced by all of their various value judgements and an enormous level of subjectivity; of course given that people in similar networks will have similar experiences and likely hold similar value judgements this leads to the tendency to 'circle jerk' on an issue, and then something becomes 'better'.. The first time they meet someone who doesn't share these judgements a flame war errupts

This is nonsense, analog also has a resolution. The stated resolution for the best audio cassettes i seen were 70us, correspondent (according to Niquist formula) to 28kHz digitalization, 1.5 worse than CDs. Most were worse, and it was just the resolution of the tape itself, telling nothing about ability of the recorder head to write them or read.

Same goes to photo film: image consists not of molecules, but of silver granules on film which are about a micron size, and either black or white. Actual resolution (which depends on contrast) is indicated on a film roll (for quality films), as a chart lp/mm vs contrast. While high resolution (and low-sensitive) film beats CCDs in this regard, reaching and somewhat exceeding 100 lp/mm which is equivalent to 2.5 microns per pixel, remember about Bayer pattern on chips to make color photos, vs about 5 microns per pixel in CCDs on DSLRs, this only comes with low sensitivity and other issues. General purpose photo films in most common shooting conditions are normally about same resolution as a good DSLR (full-frame because otherwise you have to apply a crop factor), and even than is an overshoot because most lenses does not provide that resolution, see MTF chart for most of them, the 'resolution' line is typically drawn measured at 30 lp/mm equivalent!

I am not even talking about such a fundamental effect as diffraction. It will block any further increase in image quality, whatever the lens and film is, long before you get down to 'molecules' (in fact, diffraction is in fact a limiting factor for resolution in many shooting conditions).

How important is the diffraction? Let's take practical best and worst case.

Best case: Full frame 20mpix DSLR, 6.5u pixels. Canon EF 50mm F/1.2L fixed focal length lens, resulting in about 41.6mm aperture. Resolution at peak sensitivity of 0.55 microns for this lens means that diffraction disk diameter will be 1.22 * 0.55/41666=1.61E-05 rad, which will be seen from 50mm distance (focal length, i.e. distance to film) as having 0.8u diameter - 8 times less than the pixel. So diffraction has no meaning here.

But, it is easy to see that there is a direct relationship between focal ratio and Airy disk diameter, so when diaphragm closes to 1.2*6.5/0.8 = ~f/9.7, Airy disk becomes as big as the pixel - making smaller pixel useless. This is not an extreme value - doing outdoor shooting in all-auto mode normally results in images done at f/8 or f/11 diaphragm. So, making pixels much smaller than they are now is all but useless even in a perfect optical system (which does not exist). Of course, same equally applies to film.

You say "only 44,100 times per second" as if it were somehow deficient. It's not, assuming the purpose of the sampling is to be heard by a human ear.

In fact, you'd be hard-pressed to make a legitimate, evidence-based argument justifying a significantly higher rate, for consumer music distribution purposes.

http://drewdaniels.com/audible.pdf

If film captures details at a molecular level, why are my old 400 ASA APS snaps so much grainier and lower-detailed than my cell phone snaps?

Exactly

> CDs reproduce the volume in a sound recording only 44,100 times per second.

Because humans can't hear sounds higher than 22 kHz, so with a 44.1 kHz sampling rate you're guaranteed to capture everything humans can hear.

Besides, tape has wow and flutter and records invariably have scratches and dust, not to mention limits in dynamic range.

http://rationalwiki.org/wiki/Audio_woo

The analogous concept here is the product of sampling rate and bit-depth, the number of bits per sample i.e the precision of the recording

Part of the reason is because the glass plates the images were photographed on are relatively large. The glass plate was 3inches by 9inches, meaning each image was approximately 3x3. This is somewhat larger (about 80% more area 9 sq in vs 5 sq in) than medium format film and about 8 times the size of 35mm film. While it's generally accepted that 35mm is "dead", it's still being debated whether digital has surpassed medium format.

Take a look at shorpy http://www.shorpy.com/ sometime. They have medium-res scans of full glass plates (8x10, 80 sq inches) and you can see the level of detail that a large negative captures. A full scan of the plate, capturing every transition between dark and light (line pair) would surely be in the hundreds of megapixels.

In fact you can see that many of these images are limited not by the resolution of the plate, but the lens technology (the corners of the images are often blurred, and lens imperfection called coma)