Regardless of the arithmetic, silver process film and print making makes different images. You like it or you don't. You see it or you don't.
Regardless of the arithmetic, silver process film and print making makes different images. You like it or you don't. You see it or you don't.
[i]I evolved from 35mm and MF film to high-end digital [...]
Evolution often takes wrong turns.
But who says 300dpi is required? Who says you can't digitally upsize and get "good enough" resulting quality.
It is called interpolation, something like homeopathic imaging, magic thinking.
"Good Enough"? I see, if at first you don't succeed, then lower your standards, right? If your huge prints look like crap, just stand as far away as possible. No farther than that.
A couple of months ago I purchased a D2X, having already matriculated into digital (starting in 1999) via the Canon G3, Nikon D100 and Kodak DCS Pro SLRn. Needless to say the D2x blows all of the previous camera away in every respect, although I do miss the DCS Pro's "Custom Looks"). The D2x was a natural choice because of all the Pro Nikon glass I own. Anxious to verify how good the D2X is (compared to film) I compared it to 6x9 using a Horseman VH, Velvia 100F and a few top LF lenses (47XL, 80XL, 135 Sironar S, 180 Sironar S and Nikon 270T-ED). The comparison product for the test was prints, as that's always MY end product. For the D2X I used both primes and Pro zooms (14ED, 20/2.8, 50/1.4, 85/1.4 - 12-24ED, 17-35ED, and 55-70ED). The resulting film was scanned on a Polaroid SS120 @ 3200 ppi, then resized (using bicubic sharper in CS2) to match the D2X output in size. Initial comparisons on my monitor clearly showed the 6x9 film images contained FAR more detail (and grain/noise) when viewed at 100% magnification in CS2. I then printed two test series... the first were A3's at 300 dpi on an Epson 2200. The D2X images were rezzed up from the 240 capture to 300 dpi, while the 6x9's were scaled down to the same dimensions at 300 dpi. I used Noise Ninja on all of the images and did modest sharpening in CS2. For the second set of prints made on an Epson 7600, all images were sized and rezzed to 22x33 inches. Both sets of images were printed on Pictorico PGHF. The results... the A3 sized prints exhibited virtually identical detail levels and other than the digital "feel" of the D2X images, were in all ways comperable to my eyes. On the other hand, the 22x33 inch prints showed an easily seen difference in quality. The film images contained far more detail, especially highlight detail. Almost all of the D2X images exhibited CA to one degree or another, which was quite pronounced in some cases towards the edges of the prints. I hate CA! After 55 years on the planet, I have yet to observe purple/blue edges on anything in nature.
This test, along with additional shooting, has changed my equipment selection. I find myself using the D2X quite a lot nowadays, but not if I know the output will be in excess of A3 sized prints. If movements are not needed I am even using the D2X for catalog type product work, as here the very (too) clean digital files and workflow perform the best. It's also my choice for portraiture, assuming A3 or smaller print sizes. When it comes to landscape work (or anything requiring larger than A3 sized outout) I still ALWAYS choose film... the bigger the better, as I almost never print landscapes smaller than the maximium width my 7600 allows. My most frequently used format for landscape work is 6x17... printed to 22x66 inches. I wouldn't want to even guess on how much effort it would take to duplicate 6x17 film using digital (D2X) capture methods.
With the current level of DSLR performance, in the end it's all about output requirements. As digital capture technology improves these benchmarks will also, but FOR ME - FOR NOW A3 is my digital (D2X) capture benchmark. One additional thought... between the G3, D100, DCS PRO and now the D2X I've spent over $11,000 (minus their vastly diminished resale values) to work my way into digital usage. My Horseman VH (and numerous other film camera) are still in my inventory of tools, performing exceedingly well as they have since their initial purchase. No more chasing the digital resolution conundrum for me... I'm sticking with the D2X until the next mega-jump in technology. For me that will mean an affordable ($5000) digital solution that raises my benchmark of acceptability to 24x36 inch prints. Even then, I'll still need my 6x17 and larger film cameras.
There's a good test here, D2X vs Mamiya 7
www.xs4all.nl/~diax/pages/start_mamiya_nikon_uk.html
Paul droluk,
Thanks. Great post.
Emre,
I was referring to the multi shot feature on the 22MP back. In single shot at 22MP, the 4x5 film still wins out. In multi-shot the sensor is moved while 16 shots are taken. This gives a non-bayer non- interpolated 88MP shot. Pixel dimensions are approx 10,600 wide. This exceeds 4x5 capture. Really though, you'll find the Betterlights at 48MP to be a decent match for 4x5 film. But like we agree.....the cost is prohibitive.
The one thing you didn't mention is output size. For a 16x24, there is pretty much no difference between 4x5 and the 17MP Canon. I didn't believe it until I viewed sample prints between it and 4x5 Provia......and yes.....from 18"........not from a mile as some sarcastically imply ;-)
Now if only I could afford one of those backs.......
Regards,
> the data you are espousing is probably gleaned from the "older" generation of DSLRs
Jack, not really..... I will explain more about this below. And trust me, no flame throwin here, I appreciate everyones contributions. I think this thread is very interesting, even though, it has taken on its own life already :-) But lets face it, this subject is of interest to many of us.
The fact these posts disagree with each other so much, demonstrates the anarchy in the digital vs. film "state of affairs". First, some general comments....
To compare digital vs. film, there needs to be parameters defined of the test. Then the results would be categorized by the parameter set. For example, shooting a test target with the best 4x5 gear at f8, digital lenses, excellent film alignment (see recent thread for discussion of such) no shutter vibrations, Velvia film, etc., vs. 35mm digital with a 8 MP digicam, and noisy sensor. Well in this test, we may see the scanned 4x5 film resolving in the digital file, maybe 15x the resolution of the 35mm digital file.
Now, turn the tables, we use the sharpest Canon lenses on say the 5d, (the least noise of the current line up) and compare this to a sloppy 4x5 camera, poor alignment of film, flatness issues, older vintage LF lenses, shutter vibrations, f16, negative film used, low end consumer flat bed scanner used, etc.... now, with this test, it's possible the 5d digital file can match or possibly exceed the resolution in the scanned 4x5 file.
Although this is two extremes , it would be true to quote both of these statements....
4x5 scanned film still exceeds 35mm digital by 15x.
and
35mm has surpassed 4x5 scanned film in resolution.
This is just one example of the many variables when testing film vs. digital. Of course, the next huge variable is actually testing the two using different DOF requirements for each format, whereas the LF film will suffer radically from diffraction wheras the 35mm will not. This variable is rarely discussed when making the comparisons, and its very important if one shoots scenes with lots of DOF. Next is, what color was the target, as Bayer sensors are very vulnerable to red and blue subjects, something film does not suffer. Pure red, and pure blue subjects can dismiss up to 75% of the pixels, since each pixel only aborbs one of the RGB colors, and every set of four pixels is RGGB. So my point is, when making such comparisons, unless all the variables are known, the results are hard for others to interpret.
Now, I am leaving out digital scanning backs, as they are far superior to Bayer and CCD sensors, as each pixel absorbs all RGB colors and the pixel counts are huge. However, as many know, these benefits come at a huge price.... very long exposure times, i.e. minutes vs. fractions of seconds. So this limits the applications of scanning backs, whereas film vs. digital one shot sensor are a fair comparison as they can be used for equal applications as exposure times are the same, or very similar.
The basic 1/R formula has become somewhat of a reliable formula to anticipate resolving capacity of a camera system. The lens aerial resolution and the film / sensors MTF needs to be known. Without turning this post into a math exercise, in general, digital sensors will record on the low end, about 35% of their pixel density, and on the higher end, cleaner, more expensive pixels up to 75% of their pixel density. Film, well we all know what it can record. Now I do agree with the posters above regarding alignment issues, lens sharpness issues between formats. But when all that is factored in real world tests, LF film still will do very well, until diffraction starts degrading its benefits. It's this methodology and field testing that yields the approx 10 MP per drum scanned 35mm film I quoted above.
When extrapolating up to larger formats, I agree Dave, some degredation needs to be assumed, so for argument sake, lets say 1/3. There is 12.5x more area of 4x5 film vs. 35mm film, reduced by 1/3, so about 8.3x more. Now, you one can accept somewhere between 8 - 11 MP equals 35mm scanned film (based on average set of variables from above), then a digital sensor would need 66 MP on the low side and 91 MP on the high side to match 4x5. This demonstrates how close digital one shots are getting to film as the 39 MP MF backs are just hitting the market.
But the real problem here is diffraction..... I feel these numbers above are relatively fair and safe at reasonable apertures, say not more then f11 for 4x5. But when a scene requires f8 on 35mm, it then requires f16 on MF, and f32 on 4x5 to acheive the same DOF. Now, the gap narrows even greater, as the 4x5 just lost even more advantage to diffraction, just winging a guestimate, maybe 40 - 65MP of clean pixels to match 4x5 (when 4x5 uses f32). And the worst part is, in scenes with large amounts of DOF, the diffraction (coupled with defocus principle at the near/far) effect dominates the scene, as the point of exact focus now only holds about 10% of the film area, whereas in a target test, the point of exact focus represents 100% of the recorded image.
So what is my dream (somewhat realistic) digital camera for landscapes? About a 50 MP sensor, on an area about 6x4.5, (so we benefit from shorter fl lenses, faster ss's, sharper lenses, less shake), hopefully the sensor will record close to the exposure lattitude as negative film, and record tonal ranges in a non linear fashion, oh, and one more request, the ability for each pixel to record all 3 RGB colors to reduce interpolation and increase the files up rez friendliness. This may be waaaay off into the future, so I think LF gear and a GOOD scanner is still a sensible investment vs. digital gear today. Of course, this assumes you want large prints 40"+, and shoot scenes with large uncontrollable exposure lattitude.
There is some very interesting shortcomings of digital as I mentioned above, so digital is NOT a cure-all for us landscape shooters. But, I gotta admit, I love digital capture! Whether market forces will ever drive force the R&D into chips bigger then the current 40MP remains to be seen. It's possible, the volume will be so low, the price at this end of the market will not drop too much. It's my guess, it would take Canon or Nikon to drive up their sensors to 24 - 28 MP and sell for <$10k pricing, which would force the MF digital companies down in price to compete. Or the other alternative, drive em out of business!
Emre, the 88 MP shot is not a fair comparison in my opinion, assuming it is the Sinar you are referring to. This method of shifting a few microns in each direction between shots is only useful for still subjects, is somewhat handicapped by the lenses shaprness vs. 4 mossaic shots taken stitched (which would be a more true 88 MP sensor), and the subject matter would have to warrant such resolution to notice its increased resolution. The problem with comparing the BL back is the fact it records more then 3x the data per pixel and more exposure ranges. So the closest comaprison we have to look in the future is the sensors that exist now and the new ones being released and extrapolate upwards, with the caveats that Dave mentioned.
Paul, what happened to Fotoman cameras? It seems no one responds to email or phone calls anymore?
Hey, Dave's the one that mentioned the 88MP shot, not me
Also, regarding 24 - 28MP Canon/Nikon sensors...do you think that it's actually possible? It seems that 35mm DSRL manufacturers are slowly approaching the limit where cramming more pixels into a 36 x 24mm sensor is becoming problematic. The more pixels, the more noise right? And they can't make the sensor area larger, because that would create a crop factor for the lenses.
When 12-16Mb DSLR's are under 1K and their technology doesn't go obsolete in a year, I'll get one.
When a 50Mb back for 4x5" is in the market at around 5K, I'll get one.
Meanwhile, I shoot film.
Posting Permissions
Reply With Quote
Bookmarks