How many pixels in an 8x10 neg?

[bill_1041]

A coffee break discussion prompted the following question:

How many pixels would a file contain if an 8x10 digital back were available (col or or B/W)?

Seems there was a National Public Radio story that a consumer camera available in 2001 will have a 10 megapixel chip as standard equipment, rendering images eq ual to any film-based effort. I was arguing that in a 1/15th of a second, an 8x1 0 piece of film can collect huge amounts of "pixel" information, probably dwarfi ng 10 megapixels. Anybody know?

Thanks in advance - Bill

[Huw Evans]

Well, here's my rough answer: in round figure an 8x10 negative is about 200x250 mm. If we say that a good B&W film is capable of resolving well over 100lpm (again, just round figures), then you need at least 200 pixels per mm. LF lenses won't match that resolution, but you'll still need it to replicate the tonal gradation.

That gives 200x200x250x200 = 2 billion pixels. That is 200 times the size of that 10Mpixel chip, and that is almost certainly an underestimate!

Not only that, but I think I'm right in saying that each pixel uses several bytes for 24-bit colour rendition, and it's clear that for an uncompressed image you will need several GigaBytes of storage. I hate to think what the power supply would be like.

So, if my arithmetic is right, and you're planning that week-long photographic trip into the backwoods with current digital technology, you'd better hire half a dozen pack mules and a portable generator!

Huw Evans.

[sheldon hambrick]

"rendering images equal to any film-based effort. "

Yeah, but I'm sure they weren't considering LF...as most people don't.

[Chris Werner]

Huw,

You're right about color rendition - that requires a minimum of 3 pixels (R, G, B). 24 bit color likely requires more, but this is beyond my specific knowledge.

I think I also recall that a 35mm negative contains roughly 25-30 megapixels of data, but don't quote me on that. If right, though, digital is still a long ways away from a mano a mano quality comparison with film at any format. I wonder when the dramatic improvements we have seen in digital will run out of gas in the face of practical limitations on miniturization, and whether this will be before or after digital exceeds chemistry in data storage capacity. It's easy to say now that digital will ultimately win out, but I question whether it's really that simple (see the human brain).

Wow, I fell into an unanticipated philosophical twist. ;-)

[james norman]

we can also look at this from the other end. it requires about 300dpi to achieve truly visually-satisfying reproduction (print) quality (for example, the reproductions in a high-quality fine art book). 300 x 10" = 3000 pixels across, 300 x 8" = 2400 pixels high - that translates to 7.2 million pixels. at 24-bit resolution, that is about 173MB file size. at 600dpi (to allow for 16x20 enlargements for exhibition purposes), you would be looking at file sizes of about 690MB. this level of image quality is within the capability of existing technology (well, i admit you need a lot of RAM), and i would guess variations of it will be common in the near future. on the downside, if kodak wont even make a b/w film in 4x5 readyload anymore because the market is too small, who in the world do you think will manufacture an 8x10 digital back? (i wonder how long it will be before the LOC accepts any form of digital imaging - digital files, digital prints, and/or digitally printed "negatives" or some other form of hard-copy "original", as part of its archival collections...)

[Nic Benton]

I have done some experiments digitising film to see what you can get out of it and does appear that the limiting factor is lens technology.

You can see the grain in Provia F at about 6000 dpi but above 3-4000 dpi the lenses I have used have not provided any more information.

At 3000 dpi this produces a 720 MegaPixel image (2 GBytes (24 bit)).

If you are interested the results of scans at 1600, 3000, 6000 and 12,000 dpi they are at:- http://193.113.131.213/pg/dpi/?lf2

Viewing the images requires your browser to run a Java applet. I will have a none java system by the end of the month.

Regards

[Pete Andrews]

Juggling numbers doesn't actually tell us very much about image quality.If you consider the fact that every grain in a film can only be 'on' or 'off', i.e. developed on not, then film itself is very much a digital medium.Now consider the nature of a pixel: Sure, it covers a much larger area than a film grain, but it's also capable of showing a whole gamut of brightness levels, from absolute black to pure white. Which is really closer to showing a truly analogue tone scale; film grain, or pixels?

The way that the human eye sees things must also be taken into consideration. The eye accepts regular patterns of dots as a continuous tone, much more readily than it accepts the random scattering you get with film grain. A coarse grained photographic print has many more grains per inch than a good 133 to 200 screen magazine or book illustration, but the book illustration appears to be smoother in tone than the print. The regular matrix of pixels gives a better impression of truly continuous tone than film grain.IMHO comparing film directly with a digital image, on a purely numerical basis, is like comparing apples with bananas, but if you want some numbers, I think we ought to start with the human eye.The eye can resolve at best about 8 lppm at normal reading distance. Even if you increase this to 10 lppm or 20 pixels/mm, this only works out to 82 megapixels to cover a 20" x 16" print. The final viewing size is much more important than the negative or film size.Or, if you want the hypothetical equivalent of film: We'd need at least 255 film grains to show the same tonal range as a single pixel, and this involves a film area equivalent to at least a 16 micron square pixel. (The pixels in consumer digicams are about 5 microns square BTW). This is about 2.5 megapixels/square inch; you work out the numbers, they're far lower than the gigapixels that have been bandied about previously.

Anyway, wait until we've seen the results from the new full 35mm frame-sized CCD camera announced at Photokina. (At last the digital design boys have realised that size does matter.) It's only 6 megapixels, but this is well in excess of the 2.5 mp/sq. inch that I mentioned earlier.I think quite a few eyebrows are going to be raised by it.Leaf, LightPhase, and the rest should start worrying about their future.

[John H. Henderson]

Hmmmm....I've often estimated this myself, and usually come up with about the same number - 4000 dpi. The Kodak DCS-660, with 3Mx2M array, would need to have twice the linear resolution to confidently capture the detail in a 35mm negative (were the array 36mm x 24mm, which it is not).

The 100 l/mm number would require 200 pixels/mm minimum. (In signal processing, we call this 2x point the Nyquist sampling rate.) This already gives over 5000 dpi. In practical signal processing, we'd oversample somewhat. I could even argue that 4 pixels/line pair were required. (For those who can visualize it, imagine sampling a square wave right at the transitions - you'd get no variation.)

I usually start this mental exercise with about 50 l/mm to compare with 35mm film, so my "required" resolution is less than what I just showed.

Any way you slice it, you find that the VERY EXPENSIVE, top-of-the-line sensors today aren't close to capturing the information that film can.

You can follow the "300 dpi for print" rules and come up with smaller numbers. Of course, when you choose a resolution, you have to know how big your biggest enlargment or cropping will ever be. I'm a proponent of capturing all you can when you take the picture. It's kind of like cropping. If you take too much image (too much information) when you take the picture, you can crop. Likewise, if you take too many pixels, you can always downsample.

OK. I want to blather on more, but I'll stop now.

[Ron Shaw]

Check this site for a comparison of various digital cameras to each other, and to 35mm Kodak E100 slide film. It seems that the 6MP imager used in the Kodak 660 and the Phase One med. format back (also using the 6MP imager) captures more detail than 35mm film. Very interesting! http://www.imagingspectrum.com/portraitcomparison.htm

[steve_782]

Don't really think you can come to that conclusion from these "tests." The type of digitizing for the film, compression for display etc. don't give a true method of comparing detail. Plus the digital pictures were "sharpened" with software giving a bias towards appearing to have more detail.

The only way you could really make a comparison is to use a lens resolution chart, enlarge the film onto paper to show the area with the most discernable line pairs per millimeter, and then perform the same test for the digital camera with output to a Lightjet or like printer with no software manipulations to the digital file. This isn't a real test as there were no controls applied to the process. It's only a sloppy presentation of some comparisons masquerading as information.