How much data is in that picture?

[jp80874]

Clyde Butcher was interviewed in the February 2004 issue of Shutterbug. http://www.shutterbug.net/features/0204sb_clyde/index.html
"He uses three Deardorfs (sic) (a 5x7, an 8x10, and an 11x14) and a 12x20 Wisner. Clyde says digital is a long way from how he works. ‘I’m working on a gray scale of 200MB files and the biggest file you can get on a camera is 11MB.’”

I would like to use a similar comment when asked why I am still using an 8x10 for B&W, but honestly I don’t quite understand his reference. Is he referring to the 12x20 with the 200MB file comment or can he get that much data out of any of these formats?

What is a reasonable or defensible data size for an 8x10 negative? If I am not giving enough information let me know what questions need to be answered. One wonders how I sold software successfully over the last 20 years.

John Powers

 

[Dave Parker]

Depends on the scanner,

I can derive about a 50 meg file from a 35mm velvia slide at 5400 dpi, my drum scans of 35mm color slide put out about a 75 meg file, now if I use my new scanner which is 7200 x 7200 dpi, I can get about a 200 meg file, it all become relative at a certain point, but film still offers far more information that your going to find in even the best digital cameras.

Dave

 

[Frank Petronio]

Unless it's a BetterLight back... or you figure that you can enlarge a digital capture more than a film scan because of generational losses. In that case a medium format digiback with Zeiss optics or a 4x5 digi-back with digital lenses is pretty hard to beat.

Really though, I doubt most people use the greater than 4x5 formats because they want to scan even larger files. It's the medium that's the message...

 

[Helen B]

2004? That's like back in the Bronze Age or something, isn't it?

I'm not sure why you need to justify your choice to anyone else, or why the visual evidence isn't sufficient, but here's a very simplistic view of the highly contentious numbers:

An 11 Mp (megapixel) back or camera (was that the best there was in the Bronze Age?) will produce an 11 MB greyscale file at 8 bits per pixel. Nowadays it might be more advisable to work on at least 30 Mp for the best that digital can offer. The lens needs to have produced an image with a high enough resolution for the pixel value to be representative of the overall resolution.

To work out how much useful greyscale data you get from scanned film you need to know:

-the number of bits per pixel (typically 8, 10, 12, 14 or 16) that the scanner is really capable of (just because the manufacturer claims 14 bits per pixel doesn't mean that all those 14 bits are good data). The file size will be based on 8 or 16 bits per pixel, so there could be a lot of padding as well as dodgy data. There's a good argument for sticking at 8 bits per pixel for the data claim if you are going to print with an 8 bit printer;

-the true resolution of the scanner (again, not the manufacturer's claim, but the true value. A Microtek 1800f has a book resolution of 1800 ppi and a true resolution very close to that. An Epson 4990 has a claimed resolution of 4800 ppi, but the true resolution could be 1600 ppi for one particular example);

-the film image resolution (combined lens-film-focus result) needs to be equal or better than the scanner resolution - and film resolution is a slippery thing to nail down to the same extent as scanner resolution because of the effect of contrast. 1800 ppi corresponds to about 35 lp/mm;

-the scanned area; and, most importantly

-the number of angels that will fit on the head of a pin.

An example: my Nikon 9000 resolves about 3800 ppi and about 12 bits per pixel as far as I can tell. If I scan a 6x6 B&W neg in greyscale that produces a 70 Mp (or thereabouts) image with about 100 MB of good data. The file size will be larger than this - around 140 MB. If I stick to the 8 bit limit, the scan will be 70 MB.

Another example: a 5x7 B&W scanned in greyscale on a Microtek 1800f at 8 bits per pixel would be about 90 Mp and 90 MB. An 8x10 would be about 220 Mp and 220 MB. Raise the scan ppi and the file size increases accordingly.

However, if you are wet printing then all this is a load of old toot because you can't easily quantify the equivalent pixel depth (tonality or some other vague thing) etc. It's difficult to produce a valid method of comparing two things if the basis for comparison isn't clearly and completely defined. This is where, I believe, most of the digital/traditional comparisons go wrong, and why I think that the best evidence is the visual evidence, by and large.

Best,
Helen
PS An aside, but something that always puzzles me: what's behind the apparently common practice of omitting the space between the value and the unit?

 

[MattCarey]

PS An aside, but something that always puzzles me: what's behind the apparently common practice of omitting the space between the value and the unit?

You know, I never thought about this. But I would imagine it is so that the numebr and the unit don't get split up on a new line.

E.g. It reads better to say 11k

than to say 11
k.

Matt

 

[Helen B]

"You know, I never thought about this. But I would imagine it is so that the numebr and the unit don't get split up on a new line.

E.g. It reads better to say 11k

than to say 11
k."

Yes, that's a valid argument, if there wasn't such a thing as a non-breaking space. I'm not sure if its available here, but every word processor I've used has it. There are non-breaking hyphens as well.

Best,
Helen

 

[Ole]

I've had a 5x7" scanned - that gave about 113 Mp. 8 bits colour made that a 339 Mb file. For black and white I find that 8 bits is not enough to preserve detail - especially in the extreme ranges - after a little tweaking, so a 226 Mb file would be best.

There's still more detail in the film than in the scan, though...

 

[doughowk]

Michael Mutamsky, in his article "Considerations in scanning the large format film" (View Camera, Sept/Oct 2005), states:

There is one fundamental difference between silver printing and scanning that must be highlighted. Many silver printers do not intend to print 'straight' prints when they select their contrast grades.... This approach (dodging, burning etc) is not possible with the scanner, unless two separate scans are made, and complex combinations of the two scans are produced...

B&W film can have up to 12-15 zones of information (source, Bruce Barnbaum) whereas a scan emulates paper tonal range of 5 zones. It would seem, therefore, that file size calculations should be doubled in order to represent the possible information in B&W film.

 

[Helen B]

"B&W film can have up to 12-15 zones of information (source, Bruce Barnbaum) whereas a scan emulates paper tonal range of 5 zones. It would seem, therefore, that file size calculations should be doubled in order to represent the possible information in B&W film."

Doug,

I'm not disagreeing with the general message from you or Michael, but I am puzzled by your statement.

What does 'emulates paper tonal range of 5 zones' mean? I'm unclear on that, and exactly how it relates to the dynamic range of film (12 to 15 stops, as quoted here). From my experience, scanning (as a process) can capture the full density range of a negative rather more easily than paper can. That's one of the advantages of scanning. I'm in complete agreement about scanning at '16 bits', even if they aren't all good bits, or doing multiple scans if necessary. As I said above "if you are wet printing ... you can't easily quantify the equivalent pixel depth..."

What is the reasoning behind your '5 zone' statement? I can easily demonstrate that an 8-bit per pixel greyscale file will produce smooth, indistinguishable tones over a range of around 8 stops - that limit is set by the limit of my demo, because the print I use only has a density range of 2.45 (which isn't too shabby). 16 bit scans are common - routine for some of us - but they do, of course, produce file sizes twice that of 8 bit.

I'm still not sure what the real point is, or terms of reference are, for these general attempts at quantifying the quality difference between film and digital. If you want to do it rigorously you need a much more in-depth analysis directed at a specific application. Without a clear frame of reference you can argue for ever.

Best,
Helen
PS For clarity - do we need to be careful about Mb and MB?

 

[morkolv]

With respect to the original question, I feel that file size is a very confusing term to use as this only says something about the scanners settings with respect to colordepth and requested file resolution. And not any thing about real resolved resolution. This has however in the past given me huge impressive overweigth files which has scared the hell out of my harddrive, RAM and processor etc.. for no good reason..

Over at the "other" LF site, I seem to recall a discussion concluding that a 4x5 25 Mpix filmback was close but not really as good as a good drum scan of a 4x5 slide. See also the Canon 1Ds Mark II and 4x5 over at luminous-landscape.com, which shows nearly the same.

Using this as a conservative figure one could state that a 8x10 is in the order of a 100 Mega pix camera with respect to resolution, TODAY!! But might become even better as the scanners become better (see Ole's comment that there is "more" details "left" which the scanner did not get) Try that on a digicam!

Not to mention the astronomical cost of a 100 Mpix cam!!! :rolleyes: If it existed!!

)

 

[jd callow]

This is mostly an apples and oranges issue.

If your scanner can record what is on the film and if the colour space is equal to or greater than that of the film you could figure out a reasonable comparison.

If the final product is to be printed to something that could use all the information there might be some value in knowing.

FWIW, film at its best can (I am told) differentiate between 160 line pairs per mm this would equate to 320 pixels per mm. A very good lens will be able to resolve about 100 line pairs per mm. This is not to say that film or sensor need only be able to resolve 100 line pairs per mm (ok lpmm I'm sick of typing). It is very beneficial for the film or sensor to 'over sample', how much oversampling is a question I can't answer.

So 160 lpmm is roughly equivalent to 4000 lines per inch. Each line would require two pixels or 8,000 pixels per inch (or something like a 42 mega pixel 35 mm sensor). This is all useless because we would be talking about Kodalith or 1bit image. Once you add tonal range the equation loses all form.

 

[L Gebhardt]

See also the Canon 1Ds Mark II and 4x5 over at luminous-landscape.com, which shows nearly the same.

I have looked at this in the past and there is no contest here. To say that the Canon is acceptable at 30x40 seems to be a stretch.

With a decent drum scanner I would wager that 6x7 slide film still renders more detail than the current crop of medium format backs. 6x4.5 is probably close. Of course these digital backs cost as much as my car, so they don't make sense for me since I am not a pro.

 

[jp80874]

Thank you all for your thoughts on this. I have a small solo show of 16x20 B&W prints from 8x10 negatives coming up next week at Akron University. This is where I have been taking photo courses. In the next gallery there will be a considerably larger show titled “Electronics in Art”. I thought the answer to my question might address the very common query of why I am shooting film rather than digis. I think I will simply point to the detail and say, “because the school has let me learn how to make film do that.” I’ll give show details next week if any Apugers are in the area. Both shows begin set up tomorrow, so I should soon have a better idea of the other show’s scope.

Thanks,

John Powers

 

[avandesande]

Personally I hate to think of applying such a nasty word to my negatives.

 

[Troy Ammons]

I really dug into this a while back since I have a drum scanner and started doing all sorts of test shots.

I will compare film to digital MP and not MB since MB can be sort of misleading, ie jpeg, compressed ??, 24 bit tiff, 48 bit tiff, etc, etc.

Shooting a super sharp lens in one of my film cameras I can generally achieve almost an identical "look" to a digital photo (ie similar edge sharpness and smooth) with E100G drum scanned at 2000-2500 dpi.

I have done 1 test scan of 8x10 Efke 25 at 3000 dpi just to see if I could produce a mondo panorama or print (10-16x) and it came out really sharp and really huge. Too big to work on really. I have a friend with a 6' wide inkjet.

In the end of all that malarky I figured E100G with the sharpest MF or LF lens was worth about 5-6 mp per square inch. 800 film is probably worth 3 mp per square inch. Something like Provia or especially UC 400 is probably worth 4-5.

Efke 25 and microfilms are worth more, but I am not sure how much. My scanner can go up to 4000 dpi which should be good for 60lp/mm but I have rarely seen a case where going over 3000 dpi did much to resolve more detail except for a mamiya 7 but then I was only resolving slightly more detail but a lot more grain.

If you figure the best case at 6 mp per sq in that would give you around 460 mp for 8x10 film but realistically its probably more like 300-400 mp. I usually refer to 4x5 as an 80mp equiv and 8x10 as 360mp. that seems about right to me.

Oh a 8x10 2000 dpi 48 bit tiff will end up at around 1.3 GB if I remember correctly.

Another way to look at it is, if your scans were single pixel sharp (similar to a SD9) and each row of dots was a line, and you were resolving 40 lp/mm then you would have roughly a 20,000 x 16,000 digital file.

You can always say, hey I can scan 6x7 at 8000 dpi and that gives me a 380 mp, but there is a point that you go past that resolves no more detail and the extra rez does nothing but resolve grain.

I would agree that digital has a long ways to go. I even have a hard time picking up a 645 film camera after using LF much less digital.

 

[Bruce Watson]

I'm a drum scanner operator. I scan my own 5x4 Tri-X to higher resolutions than Clyde - I work mostly with files in the 500+MB range, 16 bit grayscale, or 1.5MB 16 bit RGB if I'm using, say, 5x4 160PortraVC.

There are really two questions here. One is how much data is there, and one is how much information is there. Data and information are not the same thing, and are not interchangable.

What makes this hard to understand is the stochastic (think random) distribution of film grains of stochastic size in the film (this is the data), fs. the deterministic, fixed grid and fixed grid size of the scanner.

Basically, you can't image the film grain. It's just not physically possible with today's scanners. What you get instead is what I think off as digital grain. The scanner looks through the hole in it's grid and takes a sample - it measures the average value it finds there and assigns that value to this new pixel. This pixel doesn't look like film grain in any way - it's perfectly uniform and perfectly square.

Since you can't image the film grain, what are you really doing anyway? You are trying to get as much image information from the film as you can. Basically, the fine detail formed by tonal transitions. Film records this information using lots of film grain (the data), which is why scanning works at all. Ious, there's lots of data underlying the information - there's not a one-to-one correlation.

Given that background (simplified such as it is) there are three main schools of thought on scanning. First is to scan at about half the average grain size (say, 2000 ppi) which allows you to get the image information below the grain size. Second is to scan at about the average grain size (say, 4000ppi) to make sure you capture substantially all of the image detail. Third is you scan at approximately twice the average grain size (say, 8000ppi) and downsample so that you absolutely do get all the image detail you can.

For large format scanning, much of this discussion becomes a moot point as computers, OSes, and image editor programs are still largely 32 bit systems, and poorly implemented at that (remember when Bill Gates said that we'd never see a program as big as 640KA? Well, he didn't get any better at predicting ;-). The practical limit for a program like Photoshop today is about 1.5GB, and every operation on a file this size is terribly slow even on the fastest hardware available. I'm talking go-walk-the-dog slow.

This lets out option three right off - can't handle the file size. You can scan under option two, as long as you stick with 5x4 film. Anything any larger than 5x4 and you scan under option one. Just the way it is - it's not a choice we get to make.

My testing has convinced me that you get just about everything there is to get by scanning near average film grain size. So I'm thinking Clyde is being mighty conservative in his statements. We are a very long way from being able to replicate with digital capture what he can do with 10x8. And just forget about what he can do with a ULF camera. My prediction is that no one will even attempt that with digital capture. Not enough market size to warrant it.

 

[BradS]

....

I would like to use a similar comment when asked why I am still using an 8x10 for B&W, but honestly I don’t quite understand his reference. Is he referring to the 12x20 with the 200MB file comment or can he get that much data out of any of these formats?

John Powers

Geez, wouldn't it be enough to simply show a contact print or 8x10 color transparency?