Mega Pixel Vs Spectrum of Light

Some years back a former Kodak guy, Dick Dickerson, had an interesting article in Photo Technique [sp] magazine, where he simulated "grain" using an Excel spreadsheet. (Try making two columns with a random function, =RAND(), then show results on an x-y graph.) The result was shown on a 2D graph, and had a striking resemblance to "grain" patterns. (Hit the recalc button to see different patterns.) So it seems likely, or at least plausible, that what we see as film grain can be largely explained statistically as a somewhat random distribution. Real film is not limited to a single plane, so grain patterns can be much more complicated.

A second comment relates to what we see as "grain" through a microscope. Years back, circa 1980(?) I had set up some QC procedures to verify "adequate" preset focus on a proprietary camera system. Essentially, resolution targets were photographed, and the processed film was rated via a microscope - somewhere around 50 to 100 X as I recall. I still recall initially seeing the "grain" (actually it was color neg film) and thinking that it would limit the resolved detail. Then being surprised to see that the resolved detail was much finer than what seemed possible.

I won't try to explain except to say that anyone looking at color film "grain" under a microscope without an underlying "image" may be easily fooled as to the limiting effect on recorded detail. After this, I looked at grain as not setting a limit to fine detail, but being more akin to a coarser overlay that could obscure some of the underlying fine detail.

Ps, if it matters, the film used was the Kodak professional color neg of the day, either VPSII or III.

 

"Unless the dyes react chemically, I'm not sure that blending between layers after processing would necessarily be a problem.."

- This is getting deep to esoteric now!

The 'dyes' used in colour film are really dye-couplers - that's half a molecule of dye in simplistic terms. The other half of the (cyan, yellow or magenta) dye is provided by a developer oxidation product, which is common to all 3 dyes. Therefore any mobility of dye-couplers across the 3 colour sensitive layers before development could cause colour pollution.

To prevent this, the couplers are mixed with an oily substance that reduces their mobility, and also restrains the size of individual dye clouds.

An homogenous mix of dye-coupler, AgX crystals and gelatine matrix just wouldn't work very well.

I don't pretend to know the exact details of the chemicals or processes used to anchor the dye-couplers in place. This is just information I've gleaned from various text books and articles over time. AFAIK, it's reasonably accurate.

 

Interesting, Joe. Okay, thank you. I learn another thing! (Having started with digital, one of these days I'll get all the way back to self-developing film and mastering effective aperture adjustments with focal length on a large format camera. For everyone who finds digital scary, I assure you that going the other way is much more complicated.)

 

I agree. Analog refers to electronic technology. Audio tape is analog. The signal recorded on the tape varies as an analog of the pitch and amplitude of the original sound. a CD is digital audio. The original sound is sampled at a very high rate and reduced to a number. That number is etched as a binary number onto the substrate of the blank CD.

Photographic film is a CHEMICAL process.

So, the opposite of digital photography is not analog photography, but chemical photgraphy.

 

Paul: I'm not sure I'd buy that distinction. An analogue computer is still an analogue computer if it works with hydraulics (there are a few examples). Babbage's analytical engine is a digital, mechanical computer. EDSAC and Baby weren't less digital just because they used mercury delay lines and CRT phosphors.

The opposite of (or at least, alternative to) chemical photography is electronic photography. Older TV cameras (the output of which was recorded on film or tape) were electronic, but definitely analogue. Once everything is quantised to a set of distinct numbers, then it's digital. It's natural numbers vs integers.

The distinction in this thread was that the activation of silver halide crystals is (as I understand it) digital, specifically binary: they're activated by a photon or they're not. Effectively, the silver halide forms a very fine halftone screen. The complication is that the distribution of silver halide crystals is non-uniform so you don't get a neat grid as with a digital camera. In graphics terms, it's a non-ordered dither. We don't tend to look at the individual crystals, so the resulting image appears to have continuous tone (affected by grain), but at the level of the recording medium, others are right to say it's "digital".

A photocell, in contrast, actually captures a continuous value (quantised only because of the number of electrons in it); the result is explicitly quantised by an analogue-to-digital converter during sensor read-out, after which the image is "digital" (but not usually bi-level). It then gets processed numerically, of course.

Of course, it's all quantum if you look closely enough.

 

We live in a quantum world, where quantum mechanics applies, not the continuum mechanics that Newton believed.

Note, though, that the characteristic curves for film curve down, such that the response decreases with increased spatial frequency, not the sharp cutoff that Nyquist indicates for digital.

 

Nevertheless . . .

 

"Note, though, that the characteristic curves for film curve down, such that the response decreases with increased spatial frequency, not the sharp cutoff that Nyquist indicates for digital."

- By 'characteristic curve' I take it that MTF response is meant, not the H&D curve, obviously.

Very few digital camera/lens combinations reach the theoretical Nyquist spatial frequency, with the lens being the main limiting factor in the case of sensors with a =< 5 micron photosite spacing. Yet they still convey a far greater impression of sharpness than any film/lens combo.

Those sensors having a low-pass AA filter are naturally restricted to a spatial frequency below the Nyquist limit.

The resolution limit for film is set by the average 'grain' size and IME very rarely exceeds 100 lppmm with a very good lens and in the centre of the image circle for slow/medium speed films. Anything with a finer grain structure is almost useless for general pictorial use, and limited to tripod use with static and (usually boring) subject matter and using unreliable processing techniques.

I still fail to see why the inaccurate use of the word 'analog(ue)' has controversially been introduced, and is being so vehemently defended. It's obviously simply a device to 'hipsterise' the use of film. Such easily swayed individuals rarely have a real commitment to whatever currently 'cool' entertainment catches their interest.

So what's wrong with the simple, uncontroversial and unambiguous phrase - film photography? Especially when the end output is often scanned, and then definitely enters the digital realm. Thus totally losing any dubious claim to being called analogue.

 

Of course, especially in this day and age, and especially on the Internet, a lot of things people claim to be obvious should be taken to be very suspicious. Giving such motives to those who most likely much more innocuously use the term “analog” is probably more projection than obvious.

 

- Well, someone, or some organisation invented, promoted and propagated the stupid term 'analog(ue) photography'. It's not the sort of phrase that pops into use unbidden and without an agenda.

I doubt that, at first hearing, the average person-in-the-street even connects the phrase with film photography at all.

I stand by the hipsterisation theory.

 

Conspiracy theory much? Pretty natural sounding term to me.

No doubt.

 

"Pretty natural sounding term to me."

- Oh yes, I well remember everyone prior to a few years ago walking around with the phrase 'analogue photography' hanging off their lips..... not!

This inaccurate description only came into use after digital cameras became widespread. Presumably invented to imply some superior smoothness of tone or other refinement not possible in the digital domain. A just plain wrong assertion.

Before this Luddite-promoted schism, nobody called photography anything other than photography. And there really isn't any need to distinguish the media used now. I don't see history recording a war of words between wet and dry-plate users, or between proponents of film and glass bases, ortho versus pan, nitrate versus acetate, etc.

It's just ridiculous to attempt to put into public use an inaccurate, cumbersome and misleading phrase, when a simple and unambiguous alternative already exists - and that's 'film photography'. Easily identifying the medium from, say, wet-plate photography, or the use of a digital sensor. A phrase that's simple, easy to say, carrying no agenda, but probably not terribly cool-sounding.

Conspiracy? A conspiracy of idiocy only.

 

Well, yes, that's exactly what I was thinking about your conspiracy theory but I didn't want to go there. Thank you for saying it, though.

Of course it's new. Duh! Before new technology replaces existing, there's not much need to use adjectives which will distinguish the two. Since there was no other kind of photography except for film at the time, we didn't refer to it as film photography either. We simply called it photography. Of course! It would be fine with me if we still did that, regardless of what one used to make their photos. But people do like and sometimes need to make distinctions.

"Analog" is a word that's being used contemporarily. Some people say "vinyl" when referring to my old records. Others say "analog recordings." I don't assume any of them have a particular agenda when using either. You're reading WAY too much into this, in order to distress yourself over some ill-conceived conspiracy that someone is trying to control our minds.

 

It may well be, by the way, that "analog" is an inaccurate term. Popular culture uses plenty of inaccurate terms. And yet, people know what each other mean and generally don't get all suspicious when someone uses one of them.

 

Yes MTF. I was thinking about characteristic (H&D) from some other post.

Note that analog watches move the hands in discrete steps, from either a balance wheel or divided down crystal oscillator.
The display is still analog, based on theoretical continuous positon of the hands.

While film has discrete grains that are either developed or not, the readout is optical analog, limited only by the quantum nature of light.
Grains contain a large number of atoms, such that the size is close to continuous on a human scale. The position of grains is also pretty much continuous.

Compare to digital, where predefined pixels, samples in Nyquist sampling, are defined, and also the sample values are quantized.
Neither predefined pixel positions nor predefined quantization levels exist for analog film photography.

 

The Seiko spring drive does continuously move the second hand. It essentially lets the mainspring continuously unwind in a controlled fashion via electronic braking(as opposed to being in discrete steps as happens with an escapement/balance wheel/hairspring system).

You also have synchronous motor AC clocks, which move continuously.

 

I had thought about synchronous motors when I wrote that, but didn't know about Seiko.

But it now occurs to me, that is besides the point.

A system is analog if one physical quantity is represented by (is analogous to) another quantity.

Film photography is unusual in that the two quantities are related.

Consider analog television, where a voltage (in a cable), or the amplitude of a radio signal, is analogous to light intensity.

Or, to make it interesting, the optical sound track on movies, where light intensity is analogous to audio signal voltage, analogous to the pressure component of a sound wave.

In the case of a clock or watch, the time is represented by, is analogous to, the angle of the hands with respect to vertical. That is true if they move continuously, or in discrete steps.

Also, note that it is not sampling that makes digital signals digital. One can sample in an analog system. As an example, there are audio storage systems, I believe used in recordable talking greeting cards, that store analog voltage samples on a series of capacitors. The signal is sampled, but quantized only by the fact that only whole electrons charge the capacitor.

In the case of film, the optical transmittance at any (diffraction limited) point is pretty much a continuous function. No analog or digital system has infinite bandwidth.

And in this week of the beginning of baseball season, how many different velocities can a baseball pitcher pitch? Consider the stadium a quantum well, and consider the QM wave nature of a baseball.

 

I managed to scrape by when I took quantum mechanics, and that's a part of my graduate education I'm happy to leave well behind me! I do remember the baseball pitcher problem, though.

 

Not to fuzzy things up in this good discussion, but I've become a staunch believer and member of a small but growing group that is critical of current RGB sensor technology being based on 3 colors.

Be it CCD, CMOS or other there are hard limits to the optical filters used on optical sensors. The filter itself has a particular bandwidth for each color and isn't entirely discrete. A good deal of color interpolation is used to keep thing level, and the end results tend to be increasingly non linear as luminance and saturation levels increase. Note the universal complaints about dSLRs having trouble with dense reds and oranges. That's because the red filter on the camera has to sort 625nm vs 660nm light when the filter doesn't distinguish like our eyes do. If you make the filters denser to improve accuracy then you sacrifice sensitivity. Increasing to 4 or 5 sensor colors at acquisition level would vastly improve noise and color accuracy with digital sensors.

 

I completely agree there Scott. The RGB filters in nearly every Bayer array are too closely cut. You can see this if you try to capture a continuous spectrum with a digital camera. All you get are three red, green and blue bands. Also, the excess of green sensors in a Bayer array is extremely light-inefficient.

I've previously suggested substituting cyan and yellow filters for the two greens. This would naturally fill in any spectral gaps. The green channel is then synthesised by subtracting the red level from yellow, and the blue level from cyan. This gives two green levels that can in turn be cross subtracted from the cyan and yellow signals to render additional red and blue information. Voila! You effectively get twice the RGB information from the same 4 photosites, with no spectral gaps and 2x the sensitivity.

Alternatively, an array could be built using true triads of RGB. The geometry is a challenge, but at least it would get rid of Bayer's stupid surplus of green.

There's also a question mark over the validity of tri-colour theory altogether. The rather un-scientific fudge applied to the CIE 'horseshoe' in 1931(!) should have rung some alarm bells. Together with the fact that no real-world primaries can encompass said horseshoe.