N.B. This is not really a question, but an answer to a year old thread.

Based on a discussion about quality difference in digital printing from 5x7in vs 6x7cm (see Digital threads) I decided, being a scientist, that I should run an experiment. The impetus was conflicting statements about whether there were discernable differences between rollfilm and 4x5 when printing to a Lightjet5000 at sizes up to 20x24 (4000x5000 pixels). Bill Nordstrom has said that there would be no significant visible difference. Joe Holmes said that a 4000x5000 scan from 4x5 would be absolutely sharper than the same resolution scan from 6x7 (or 6x9). Since I have great faith in both of these individuals, I decided I had to run a test.

So, I shot the same scene in 4x5 and 6x9. Both cameras used Velvia, and I selected a lens for the 4x5 that was 50% longer than for the roll film. The lenses were a 180 Apo-Symmar for 4x5 and a 120 Apo-Symmar for rollfilm. The 4x5 was exposed at f/22, the 6x9 at f/16 to compensate for both differing diffraction limits and depths of field. The 4x5 was scanned at 1400 dpi, the 6x9 at 2100 dpi. Assuming (as everyone seems to say) that the optimum input to the Lightjet is 203 dpi, those resolutions correspond to a print with a maximum dimension of about 30 inches.

I have compared the resulting scans in Photoshop. They are incredibly similar. There are no specific details in the scan from 4x5 that are not visible in the scan from 6x9. However, the edge contrast is slightly better in the scan from 4x5 giving the impression of just a hair sharper image. The difference is very subtle but it does exist. I doubt that it would be as noticible in a print as on a monitor.

I conclude that Bill and Joe are both right. The 4x5 does generate a slightly "sharper" scan, but a moderate size print, say up to 20x30, from 6x9 will be virtually indistinguishable from one made from the larger sheet film.

I would love comments or feedback, and am willing to email snippets of the scans to anyone that wants to see what the difference in format means in this digital age.

-- Glenn C. Kroeger (gkroeger@trinity.edu), December 01, 1999

Answers

Glenn,

thank you for your experiment, however, I cannot see that it proves a thing. you are bandwidth limited, in a sense, constrained by your fixed size and printer resolution. you've stacked the cards against 4x5 by sampling much lower. might as well have scanned an 8x10 at even less of a sample rate. if my math doesn't fail me, you'd see the same in 35mm scanned at 5760dpi. the geometries of the scanning hardware should favour better conrast/tone in the larger prints, but I would think it would be marginal in this testcase.

if you were to reverse the process, and reduce the image size, would you be surprised to see little difference in the three, 1" images?

-- Daniel Taylor (aviator@agalis.net), December 01, 1999.

Daniel:

I don't disagree with your statements, but I was not trying to prove that 6x9 is equal to 4x5 in any sense, only that the additional information on the 4x5 is not manifested in the specific size prints I was investigating. No question that I was limiting the 4x5 with the sampling at 1400 dpi. But that is the "real world" limitation with the Lightjet at these print sizes. On the other hand, the results suggest that with real world lenses and film backs, 2100 dpi is approaching the limit of usable information from roll film (its approaching 50 lp/mm). The 20x24 or 24x30 size is simply the trade- off point of these limitations. For those of us that never print larger than 20x24 inch prints, the combination of high-res drum scanning and the Lightjet have effectively removed the gap in final print quality between roll-film sizes and 4x5.

-- Glenn C. Kroeger (gkroeger@trinity.edu), December 01, 1999.

Have you actually compared enlarged prints? I think this would provide the real answer. Keep in mind that a monitor can only display a resolution of 72 dpi (some 90 dpi) I just feel that viewing these in Photoshop is not the right place to compare. Thanks.

-- Jim (jimblps@earthlink.net), December 01, 1999.

Glenn, I shoot only 6 x 9 cm color transparencies ( I bought my Horseman VH from you and love it). Although I suspect there is virtually no difference between high quality digital prints up to 20" x 30" (drum scanned and printed on a Lightjet 5000) from 6 x 9 cm or 4" x 5" slides, I have a couple of methodological problems with your test. First, the maxmium resolution from the Lightjet is at 120 dots per cm, about 300 dpi. Although Bill Nordstrom prints everything at 80 dots per cm, many other people (including Bill Atkinson and Nancy Scans) claim you can see a slightly improved quality when printing at 120 dots per cm. Especially for your test this would be sginificant as the Lightjet increases the resolution from res 8 to res 12 when the lower res files are provided. Second, who did the scanning? What kind of a scanner was used? I would hope a Tango Drum scanner or equivalent. Did the scan oeprator sharpen in the scan at all? This may account for some difference. Moreover, if you didn't sharpen at all in the scan stage, this would not be a very practical real world test because everyone sharpens scans before printing them to the Lightjet. A more useful test would be to sharpen each print at the scan stage and in Photoshop to the point where you would want to for that given print; you could then comopare (a) differences in sharpness, and (b) differences in tonal subtlety (i.e. the 6 x 9 scan which may be sharpened a tad more . . . but maybe not because the shorter focal length gives better depth of field . . . may give up tonal subtlety for the Photoshop enhanced sharpness). . . A gratuitous plug -- when you get a chance check out my new site at www.naturelandscape.com. It displays my

-- Howard Slavitt (info@naturelandscape.com), December 02, 1999.

Glenn, you started your experiment with a flawed method. If you want to capture all the information the LightJet can output, you need to scan to get a 300MB file. If you want to make a valid comparison of the formats, you should scan each file at a resolution that will yield 300MB files. After downsampling the files to the desired output size, you will see a bit more sharpness and more detail from the larger original.

Two of my friends own a business specializing in scanning slides and preparing them for LightJet output. Their clients include some of the best landscape photographers in the world. I rented time on their Tango drum scanner a few weeks ago and we discussed this very issue while we were working. After doing this for the past six months since he got the Tango, Rich has seen what different formats are capable of producing. His opinion, based on examining the final output, is that larger formats do give better prints, even though they are scanned to the same file size as medium format.

My 6x7 slides required a bot more than 4100 pixels per inch to capture 300MB data. If we assume 4x5 film has the same resolution as 120 film, and the lenses have the same resolving power, the larger area will yield more details (think of veins in a leaf). When you prepare the file in PhotoShop and apply a sharpening filter the greater detail will result in a sharper print.

And as someone already pointed out, monitors are low resolution devices. The correct way to definitively test this is to scan both slides to 300MB and output them at 120 pixels/cm (304.8 dpi) which is the LightJet's highest quality output.

All that being said, the differences among various original sizes are much more subtle using digital methods than when using traditional methods. Right now we're limited by the LightJet's output size restriction, if you want to think of 50 inches by 50 inches as a restriction.

I ordered a 4x5 camera last weekend. In a few months I'll go back and scan some 4x5 slides. I might just try this experiment myself and see what happens. I could slap a 6x9 back on the 4x5 and use my 135 and 210 Rodenstock lenses for the comparison. Which leads to another flaw in this experiment. Medium format lenses have more resolving power than large format lenses. Sigh. Does nayone have a Hassleblad or Rollei I can borrow for a few months? Current models preferred.

-- Darron Spohn (sspohn@concentric.net), December 02, 1999.

take a balloon, blow it up to a reasonable size, and with a felt-pin marker, draw a 6x9cm rectangle. mark several horizontal lines, representing a scan line across the rectangle. now, blow the balloon up till the rectangle becomes approximately 4x5" and that is what you have done. same image, bigger, less scan lines per inch, but identical to what you started with. nothing has changed.

you are constrained at one end by your printer, and the resolution of your scan hardware and film structure on the other. obviously the 4x5 will allow for bigger prints, and the increase in scan data would allow average and smoother tonality when scaled downward.

project your argument in the other direction, and you would speculate that APS would scan as well. scanned high enough, you could easily present the amount of data the printed required, though image data would be equaled by noise and grain acquisition.

give me oversampled 4x5. tonality to die for.

-- Daniel Taylor (aviator@agalis.net), December 02, 1999.

Daniel, I like the balloon analogy, but for digital work we need to modify it a bit.

Blow up a large balloon. Draw two identical size rectangles on each. In one box draw lines at a fixed distance apart to represent the total film resolution of a 6x9 slide. Now draw more lines lines in the 4x5 box (this is a thought experiment, the actual number does not matter).

You now have two boxes of equal size, but one contains more lines than the other. These boxes represent the 300MB scans off a Tango drum scanner, which can be printed at 40x50 inches on a LightJet 5000. The lines represent the details in the original slides. Now let some air out of the balloon until the boxes are the right size for 24x30 prints. The 4x5 box still contains more lines, even though they are both the same size.

This is where you get more detail in your LightJet prints, and the subtle sharpness difference. Sharpness is really not an issue with digital prints, as we are not enlarging anything as we do in a traditional darkroom print. The difference will be in detail resolution and grain size. The sharpness difference we can see is a result of the sharpening filter having more edges to work on.

To take this a bit further, a 35mm color slide scanned at 5,000 dpi yields a 100MB file. Scan a large format negative to 100MB and print it at 16x20. There will be less grain in the print, more detail, and a bit better sharpness. The grain and detail differences will be quite evident. The sharpness difference will be very subtle.

-- Darron Spohn (dspohn@photobitstream.com), December 02, 1999.

Darron,

you hold the answers, that is if I read your postings correctly. you just Tango scanned your images, and you have a UMAX flatbed scanner. I realize you did this for archival purposes, but if we are all saying the same thing, your ink jet 8x10 prints (you select the size) should look much the same, whether you print from your Tango or UMAX. now that, would be an interesting observation. dismissing dynamic range, and other scan issues, tell us how the prints differ, if at all.

-- Daniel Taylor (aviator@agalis.net), December 02, 1999.

As soon as I got home and opened the Tango scans in Photoshop I sold my Umax PowerLook III. The 8x10 inkjet prints from the Umax scanned files and from the Tango scans are silimar in sharpness, but the tonality and dynamic range from the Tango scans are incredible. The noise floor (borrowing an audio term for lack of a better description) is much better on the Tango scans too.

After seeing the Tango scans I just couldn't see bothering with a flatbed scanner again. The JPEG files I uploaded to my web site do not do the Tango scans justice. I have some 16x16 LightJet prints in transit to my house. These were printed from files scanned from a 6x6 slide taken with a Mamiya C330f and 135 lens. They are marvelous on screen.

-- Darron Spohn (dspohn@photobitstream.com), December 02, 1999.

"silimar in sharpness? I meant "similar" That's what I get for not proofreading before hitting submit.

-- Darron Spohn (dspohn@photobitstream.com), December 02, 1999.

After trading e-mails with Glenn on this issue, I got to thinking about how critical we are being here. Glen stated that at normal viewing distances you would not be able to discern any differences. He is right, of course. But I replied that if you get closer you can see the increased detail and reduced grain from the larger original.

After sending my reply to Glenn I reflected on something that happened yesterday. It's a bit of a story, but bear with me because it has a happy ending.

I'm working a six-month contract job at a high-tech firm in Silicon Valley. They auctioned off some old computers and unused office equipment this week. I went downstairs yesterday morning to see if there was anything interesting, and saw a black and white photo leaning against the wall. Hmmm. Nice shot from a distance. Turned out to be a 16x20 archivally matted and framed fiber print of Ansel Adams' "Aspens, Northern New Mexico." High bid at the time was $10.50. This was s silent auction where people wrote their name and bid on a sheet of paper taped to the item.

Not wanting to drive up the price, I decided to wait until a few minutes before the auction ended, fully expecting to pay $200 for this masterpiece (not an Ansel print by the way, it was printed by an assistant. I checked the Adams Gallery web site just to be sure it was what I thought. It was.

Ten minutes before the auction ended I went back downstairs. The bid was up to $14. This is robbery. I bid $18 hoping to discourage anyone from incrementing by one dollar. Just as the auction was ending the lady who had bid $14 walked up and asked me if the $18 bid was mine. I told her it was.

She said she was decorating her new house, she really liked the frame, and she wanted the photo. I told her I'm decorating a new house too, I shoot photos in my spare time, and I was prepared to outbid her. I asked her how high she was willing to bid. "$50," she replied.

Stunned, I told her I would gladly pay $200 for the photo. She looked at the photo, looked at me, and said "It isn't that good a photo. I could do as well up in the Santa Cruz mountains myself." At that I was determined to buy the photo at any price. I had visions of her throwing away the photo and putting a cheesy shopping mall picture in the frame. She broke my train of thought when she said. "No. I'd never pay that much for that photograph. I paid $200 for a 3D photo last weekend. This one isn't worth more than $50."

I hung the Adams photo on my living room wall last night.

The point of my ramble here is that what we're arguing, while interesting, is academic. We're quibbling on fine points along the lines of which Zeiss lens is sharper, a Rollei or a Hassleblad. Ooops. Wrong forum. Make that which LF lens is sharper, a Rodenstock or a Schneider.

The vast majority of people neither care nor could see the difference between a 6x9 printed to 24x30 or a 4x5 printed to that size using traditional methods, much less drum scanned and printed on a LightJet 5000. But I do find all this discussion fascinating.

So who's going to let me borrow a Rollei 2.8GX?

-- Darron Spohn (dspohn@photobitstream.com), December 02, 1999.

Darron,

actually, it is a very interesting dialogue and I don't view it as a Canon/Nikon debate at all. it does have to do with identifying a print quality that meets your criterium, and working backwards, to establish the minimum format you could use and be pleased with the results. I just bought this Linhof Technikardan, mostly out of the 'zen' of working with a precision machine, and slowing my pace. I am sure there will be many a time, that I will agonize over that fact that 6x6 would yield the same results and be a lot easier to tote around. and then again, my neighbour, Christopher Burkett, would argue that even 8x10 is too small at times!

-- Daniel Taylor (aviator@agalis.net), December 03, 1999.

Hi Darron;

"Noise floor" is a commonly accepted term for what you were trying to describe in the image capture world.

High-end drum scanners use photo multiplier tubes as their capture elements. PMTs have inherently superior noise performance compared to CCDs for any given level of sensitivity (scientific CCDs do have very good noise performance, but their sensitivity is low, so you really have to blast them with light to attain full saturation with a reasonable line period).

That having been said, there are CCD-based scanners on the market which offer extremely good noise performance. I've had a few 4x5's scanned on a CCD-based Scitex flatbed recently, and the results were good enough that I frankly can't see any reason to drum-scan them. Of course, the Scitex in question is a $50K scanner, so it's really not any cheaper than a drum.

-- Patrick Chase (patrick@sdd.hp.com), December 03, 1999.

All of this discussion has been wonderful, and of course, if simply carrying an 8x10 guaranteed that my images would all be like Christopher Burkett's, I'd already have one.

But Patrick's post about his experience with a "low-cost" scanner, and the balloon analogies, raise a question that I am wondering if someone can answer, and that concerns the independence of sampling area on adjacent pixels.

I work with Landsat images which are processesed to yield 28m pixels (ground resolution). But we always know that what we take to represent a 28x28 meter patch of ground is really a number that represents light entering the instrument from a circular gaussian distribution that overlaps adjacent pixels.

Given the optics and mechanics of the scanner and light scatter in the film base, how "independent" are pixels scanned at say 5000 dpi on a Tango (or for that matter at 11,000 dpi which I guess the Tango quotes as an upper limit), or what kind of footprint does the PMT actually see?

Afterall, what we are all discussing can be viewed in a different context. Given a print size and input resolution to the LightJet (res80 or res120), the input file will have the same number of pixels regardless of the original size of film scanned. What makes the three (RGB) 8-bit integers in those pixels "better" if the scan comes from a larger original? Clearly noise, film grain and pixel spread are some of the factors.

-- Glenn C. Kroeger (gkroeger@trinity.edu), December 03, 1999.

This discussion really leads back to the fact that most printing paper has a resolution that is lower than film. Once you have chosen a print size, then whatever the reproduction method you eventually reach a point where the extra detail on a larger negative simply cannot be expressed on the paper. Exactly where that point occurs is a subjective issue, but it has to occur somewhere.

Basic information theory would imply that finding the point of no extra print quality should be a simple mathematical excercise, but it turns out that the 'point' is approached gradually, so there is always room for personal choice in exactly how redundant you like your image data to be. In analogue imaging grain complicates the issue further.

This is true even in the digital world. Glenn mentioned remote sensing data, in which the pixel density is higher than the optical point spread function (roughly speaking, the resolution). The same is true in scanners, and in fact the optical path in many scanners is deliberately designed to blur the projected image so that adjacent sampling cells overlap slighly. Doing so automatically prevents aliasing (it's exactly the same as band-limiting an audio signal before digitisation) and is one reason why scanners which allow adjustment of the illumination level do so by changing the brightness of the light source and not with an iris diaphragm which would upset the amount of blur.

Drum scanners don't use imaging optics as such, but instead scan a spot of light across the film and measure how much passes through at each pixel position. Their resolution limit is set by how well you can focus the spot (which is ultimately determined by collimation of the light source and diffraction in the focussing optic) and such things as control of scattered and stray light. There are specialised techniques (SNOM: surface near-field optical microscopy) which allow you to go beyond the diffraction limit and measure optical absorbtion on the atomic scale, but nobody does that with film - at least, not when scanning to make a print.

Provided a scanner is 'good enough', it's true spatial resolution is not actually very relevant because beyond about 4-5000 dpi normal pictorial films are all grain, and increasing the scan resolution merely reproduces the grain structure at higher and higher magnification. Photographers value the result because prints with corner-to-corner sharp grain are regarded as good, but the amount of image information isn't changed much.

-- Struan Gray (struan.gray@sljus.lu.se), December 03, 1999.

Taking an offshoot from Darron's latest post, I think a more interesting comparison to 4in x 5in vs. 6cm x 9 cm drum scans/lightjet prints, is 4" x 5" traditional prints (such as cibachrome) vs. 6 cm x 9cm drum scans/lightjet prints. Has anyone done any empirical comparisons along these lines? I'd be very interested in the results. My guess is that the 6 cm x 9 cm drum scans and lightjet print will be superior to the 4" x 5" traditional, cibachrome print. (Yes, that's right, the smaller original should, IMHO, yield a superior final image because the printing process is so much better.) Howard http://www.naturelandscape.com

-- Howard Slavitt (info@naturelandscape.com), December 03, 1999.

Let me add my voice to Howard's for feedback on the 6x9 (or 6x7 or whatever MF you want to choose) digital vs. 4x5 "analog" print comparison. I've been curious about this one for a little while, and have spoken to a few people about it. My next step is an actual test (I know, the only way for me to really know). But in the meantime...I too would appreciate any information from folks who actually have made that comparison already.

By the way, in my talks with Bill Nordstrom and the folks at Calypso and Evercolor, they agree with Howard's thesis: they all say that a digital print of a certain size from 6x7 film will beat a conventional print of the same size from 4x5 film. Does anybody with experience agree or disagree?

-- Greg Lawhon (glawhon@unicom.net), December 03, 1999.

for the record, I took my first image with the Technikardan on Polaroid 55pn, cut up the negative, and scanned it in as a panoramic 35mm negative on my PhotoSmart scanner at 2400dpi. the tonality is much smoother than the corresponding 35mm negative scan, and grain appears to be an order of magnitude less, as it required less unsharp masking and less sizing. smoothness of tone transition is the overwhelming observation, and I am very pleased with the results printed on my Epson 1200 in 1440dpi mode.

with the new Quadtone inks, I'll be ecstatic!

-- Daniel Taylor (aviator@agalis.net), December 04, 1999.

This may have no relevance to the original question of scanning 67 vs45, but here goes. I had my 1st roll of 35mm hi-res scanned "Kodak picture CD".I manipulated them to the extent the software on the disk allowed-sharpness/framing/contrast/brightness. I printed the results on my very low tech HP 820 ink jet printer on el-cheapo plain old paper-8 1/2X11. The results were "pointilist paintings". They really looked fantastic.I mounted 2 of them and they catch my eye every time I walk by. Has anyone out there had this result? Re: digital-this months Shutterbug had many articles on same. George Nedleman

-- George Nedleman (gnln@thegrid.net), February 22, 2000.