Theoretical digital sensor equivalent to 8x10?

[Andrew Clearfield]

I am curious to calculate how large an array of sensors would be needed to be able to record as much information as there is on a well-made and properly processed 8x10 negative.

We all know that a digital sensor may have an image-capturing capability greater than that of a traditional, chemically-based negative of the same area; or accordingly, may be (and usually is) smaller than the traditional film-based format of a comparably-sized camera. The question is, what size, and pixel-capturing capability would be necessary to approximate the amount of information one gets from an 8x10 camera?

Here are my assumptions, to use as a basis for a comparison:

Obviously, such a calculation would be a theoretical limit, based upon an ideal contrasty subject in bright light, with the finest grain/developer combination customarily used in ordinary pictorial photography; since the standard for most digital sensors is equivalent to ISO 100, let us say that the analog-based benchmark would be a sheet of FP4 Plus processed in a metol-hydroquinone developer such as ID-11 (D-76). I prefer PMK Pyro, but the numbers used to be reported for more conventional developers. If memory serves me rightly, such a combination would have a theoretical maximum resolution of about 125 lp/mm, well beyond any lens.

The best general-purpose lenses in use for large-format photography are supposed to have a maximum resolving power equivalent to, say 40 lp/mm in the central area, and perhaps 20 out in the periphery. (I'm using Schneider's figures for their old Super-Symmar series, but I'm assuming my Commercial Ektar and Fujinon-W lenses are not that far off from this. Yes, there are many other factors which go into the design of a lens, but I'm trying to calculate maximum data-collecting capability here.)

An 8x10 negative or transparency has an area of slightly more than 50,000 sq. mm (51,562 to be more precise, but there are the unexposed borders of the film, and anyway, these calculations are going to have to be very approximate, anyway.) Even at a very high 40 lp/mm, this implies the film would have registered a little over 4,000,000 data points (intersections) from a perfect grid inscribed with all lines at the minimum distance resolvable, again under these ideal, high-contrast conditions. More likely, there would be between 1 and 2,000,000 such points recorded, as the resolving power falls off away from the central axis and under more normal light conditions. (Of course, if the image produced has this capability, the number of useful data points at the classical photographer's disposal would be the same in any ordinary photograph, it simply wouldn't be as easy to measure them.)

Now, assuming we are concerned with instantaneous image capture here—no scanning and re-sampling of the data, because the 8x10 camera is not itself limited to perfectly static subjects—how many pixels are necessary to approximate those 1 to 4 million usable grains in the image, and how large a sensor does one need to record the same amount of information? I assume one gets one bit of data (i.e., 1 or 0) from each cell of a sensor, and at 8 bits per byte, if (and this is a big part of my question) it takes 1 byte to contribute the equivalent information of that one exposed and developed grain of silver, (i.e., one pixel) it would seem that a sensor capable of generating 32 Mp would be capable of giving the same amount of data as a black and white photograph from an 8x10 camera, and, since digital sensors are always capable of recording color, the sensor would have to be 4x that size, or 128 Megapixels to have the same capabilities. Now, no one has yet made a sensor that large to my knowledge, but if they could produce one at a cost someone could afford, would that do the trick? Or have I omitted some factors? Equivalently, there is a great deal of software which reads and interprets the image before it is transmitted to a recording medium, even in Raw format. This should reduce the number of Mp, and therefore, of cells necessary to give the same result.

I also appreciate that the physical size of each cell makes some difference. Nonetheless, does a Leaf 80 Mp 40mm X 54mm sensor capture the same amount of information as, say, a 5"x7" film camera? Would it, if you made it physically twice (or four or six times) the size in order to have larger individual cells?

I understand that such an image would still not have the same character or look as the chemically-produced image. But is such a calculation as I have made fundamentally faulty, and if so, why?

[Leigh]

We all know that a digital sensor may have an image-capturing capability greater than that of a traditional, chemically-based negative of the same area;

Huh? Don't know where you got that. Sounds like advertising hype.

A pixel sensor is a physical structure that includes a lens, the active sensor element, and electrical connections.

An image forming element in film is a cluster of molecules.

------

I typically scan my 8x10 film at 2400 ppi, so the 8" side is 8 * 2400 = 19,200 pixels, and the 10" side is 24,000 pixels.

The image area is therefore 19,200 * 24,000 = 460,800,000 (461 Megapixels).

A good drum scan @ 9600 ppi = 8 * 9600 * 10 * 9600 = 7,373 Megapixels = 7.4 Gigapixels.

- Leigh

[welly]

This has been discussed a few times:

http://www.largeformatphotography.in...ht=8x10+sensor

http://www.aphotoeditor.com/2011/08/...-capture-back/

http://www.largeformatphotography.in...ht=8x10+sensor

http://www.largeformatphotography.in...ht=8x10+sensor

[paulr]

If you're curious about equivalence, read this white paper very carefully.

You'll see why it's actually quite easy to equal 8x10 in situations where you need a lot of depth of field, and almost impossible to do so in situations where you don't.

This theory is born out precisely in tests like the one discussed here.

[timparkin]

If you're curious about equivalence, read this white paper very carefully.

You'll see why it's actually quite easy to equal 8x10 in situations where you need a lot of depth of field, and almost impossible to do so in situations where you don't.

This theory is born out precisely in tests like the one discussed here.

Paper looks interesting - must look :-)

I worked out the possible resolution for different apertures with 10x8 on Delta 100 and I got the following

f/22 = 600Mp
f/32 = 450Mp
f/45 = 300Mp
f/64 = 130Mp

However, this is resolution which is different to how the average punter assesses sharpness. Give your average punter a pictorial image blown up to 40"x50" and we reckon the following are the results (given comparitive analysis vs IQ180 on Alpa)

f/22 = 250Mp
f/32 = 180Mp
f/45 = 100Mp

The results aren't quite comparative because as you increase print size, the IQ180 overtakes the 8x10 for a short period at around 30x40 / 40x50 and then drops back again as you get up to 50x70 etc - this is because the acute bayer pixel sharpening effects peak at those print sizes whereas you can't sharpen the film quite as much without grain/noise getting in the way. Ask the viewers which they prefer though and they invariably prefer the 8x10 :-)

[Andrew Clearfield]

I typically scan my 8x10 film at 2400 ppi, so the 8" side is 8 * 2400 = 19,200 pixels, and the 10" side is 24,000 pixels.

The image area is therefore 19,200 * 24,000 = 460,800,000 (461 Megapixels).

A good drum scan @ 9600 ppi = 8 * 9600 * 10 * 9600 = 7,373 Megapixels = 7.4 Gigapixels.

- Leigh

And that means if I scan a 120-year-old 8x10 thick-emulsion, coarse-grained negative made with a soft-focus lens on your drum scanner @9600 ppi I've improved the resolution of the original camera????

In the process of scanning and photoshopping a picture, many millions or even billions of pixels can be created. Sometimes, interpolation can improve the apparent sharpness of the image. That doesn't mean any more data was captured in the original exposure. I'm talking about the amount of real-world information captured and stored on a reproducible medium (a piece of film, a chip) at the moment the exposure was made. That's very different.

[Andrew Clearfield]

@timparkin:
These numbers seem more likely. How did you arrive at them?

Sharpening effects probably shouldn't count, because you're not capturing any further information, you're just exaggerating what is already there. (Of course, if the net result is that the final print of 30"x40" looks sharper to trained eyes, and without having the flattened, artificial look of an over-sharpened print, then you've accomplished part of the objective of making a photograph in the first place. Naturally, there are a multitude of other factors than just sharpness—but I'm surprised that a 40x54mm 80 mp sensor could do this well, in an objective comparison.)

[Leigh]

And that means if I scan a 120-year-old 8x10 thick-emulsion...

I wasn't talking about scanning junk. If you wish to waste your time doing so, by all means have at it.

- Leigh

[Nathan Potter]

I'm inclined to think that the Ops comparison is meant to simply compare the image capturing ability of an 8X10 film to that of the top digital sensor available (or if it were available in 8X10 format).

There is a great difference in the physics of image capture between a digital sensor and film so even just a comparison at the film plane is difficult let alone including lens, f/stop, DOF etc.

If we look at only the film plane and consider the best digital sensor of 6 or 7 µm pitch (say about 180 pixels per mm). That's about 25 MP/ sq. in. so for 80 sq. in. that would be around 2000 MP. The caveat in this is that currently such sensors are Bayer designs where a color is reconstructed from a reading of adjacent pixels or B&W is determined from mainly the green cells. The actual size of a color pixel is really determined from the particular demosaicing algorithm used - also the case for a B&W image. Thus a color pixel is more like an 18 to 20 µm pixel, with B&W possibly less, depending on the specific algorithm used. This bayer demosaicing then reduces the effective sensor resolving capability by up to a factor of 10 for color and somewhat less for B&W. If the demosaicing algorithm is very sophisticated one can toggle every RGB pixel one by one and use the surrounding pixels to extract a color from every one of the 2000 MP but even with such a technique the color at each pixel is still a synthesis and not a full disclosure of what was at that particular pixel. This uncertainty scales with any size sensor one would choose.

Film on the other hand is a fundamentally different form of image capture as Leigh has suggested above, a "cluster of molecules". Fine grain film would typically have silver halide grains of about 0.8 µm diameter with a range of say 0.5 to maybe even 1.5µm. When irradiated by photons say as few as 1 or as many as 100s' the latent image increases in area by several times and in some extreme cases a darkening may be visible (printing out so called). Under low light conditions the latent image may still be close to the silver halide dimensions - obviously very high resolution and yielding what many traditionalists call pleasing microcontrast after development with sometimes marvelous shadow detail in a final print. At the other end of the density range, heavy illumination from the scene highlights, the larger latent image grows even larger upon development forming silver clumps up to say 20 µm in diameter. So the resolving power of film is a sliding scale depending on the amount of original exposure and modulated by the development sequence.

These are just the variables when considering the sensor/film part of the photo process. When we add the variables of lens resolution, f/stop, DOF, etc. we add a myriad of additional complications to the resolution performance of the system.

I dare say that is why many here would prefer to study the qualitative aspects of an actual negative or print as a basis of comparison.

Nate Potter, Austin TX.

[rdenney]

It should be noted that there are other effects of the format change than just resolution. For one thing, depth of field is considerably different. Another is the required capabilities of the lens, particularly at both wide and the smallest apertures. Issues of resolution ignored, it's very difficult to create an 8x10 image using a much smaller digital sensor, because the lenses are not fast enough to provide the same depth of field effect, or sharp enough at those wide apertures to perform well enough. Thus, many of the images possible with the 8x10 are really challenging with something much smaller (and vice versa).

Rick "especially if fast lenses are used for extreme selective focus on the 8x10" Denney