Top-end digital concerns

[bglick]

Nice response Bruce....

I updated my graphic for final print sizes for 5 formats at different DOF scenarios, in a new thread...

http://www.largeformatphotography.in...ad.php?t=45186

[Frank Doering]

I love crunching numbers. But for real world comparisons, there is occasionally a point to pixel-peeping. Here is a real life comparison from two architectural shoots, 4x5" film and Nikon D3, both off a sturdy tripod, with mirror-lockup for the Nikon. I don't remember exact sharpening routines, but for the observed differences they don't matter.

100% crops at http://doeringphoto.com/temp/testimage.jpg

[rvhalejr]

...but for the observed differences they don't matter...
100% crops at http://doeringphoto.com/temp/testimage.jpg

I'd like to read the small print on the clock. If your in the mood see if you can
resolve it in the image plane (with the help of a 40x loupe) and if so maybe try
to get it on ISO 25 film .,.

You got guts using a 24mm on the DSLR to approximate the 4x5 90mm .,.

[Frank Doering]

rvhalejr, maybe I wasn't clear: the differences are obvious: the 4x5" negative smokes the DSLR at the 24x30" print size. I didn't try to approximate, I just needed a color image for a small repro size.

Reading the fine print: the small letters on the clock face are about 0.05mm tall on the negative. Making them legible would require, I imagine, at the very least 4 cleanish line-pairs per letter height, i.e. around 100 lp/mm. Good luck at f/22 which I needed for DoF. At any rate, in order to read the letters in the 24x30" print you'd need tremendous eyesight because the lines are 1/3mm tall.)

[Kirk Gittings]

Frank, the outcome of your comparison seems quite obvious from the get go. I'm not sure what the point was really. Actually I guess, it was not set up as a test was it? Just a defacto comparison because you happened to shoot both on a job. To be fair the best test would be the current top rated DSLR, the Nikon D3X with a first rate prime lens against a drum scanned 4x5. What would be more enlightening even still would be to see how many stitched images it would take with the D3X and a primo prime lens to equal the 4x5 in resolution, 2X, 4X, 8X, 16X including a test for DR? If I had access to the Nikon equipment, I would love to try it but I don't unfortunately.

[Gordon Moat]

One thing I thought of with Frank's two images was exposure. In the case of 4x5, with a small enough aperture and long exposure, a swimmer in that should would appear as a blur, unless the swimmer was lit with a strobe. Compare that to stitching, and again there would be some trouble having a swimmer in the shot.

Ciao!

Gordon Moat Photography

[rvhalejr]

rvhalejr, maybe I wasn't clear: the differences are obvious: the 4x5" negative smokes the DSLR at the 24x30" print size. I didn't try to approximate, I just needed a color image for a small repro size.

Amen, No disrespect intended or implied.,. I simply LOVE real world comparisons and yours is great !!!

Reading the fine print: the small letters on the clock face are about 0.05mm tall on the negative. Making them legible would require, I imagine, at the very least 4 cleanish line-pairs per letter height, i.e. around 100 lp/mm. Good luck at f/22 which I needed for DoF. At any rate, in order to read the letters in the 24x30" print you'd need tremendous eyesight because the lines are 1/3mm tall.)

Absolutely Outstanding Example and well proven procedure.

I digress as the (arguably loopy) 35x-40x Loupe researcher (and a fanatical film flatness person) attempting to push the envelope in small way, maybe from 3200dpi (commonly practiced) to 4000dpi (a rarity with good reason).

Using this as a textbook example (with the final product being High IQ pixels ref
...largeformatphotography.info/forum/showthread.php?p=430929#post430929
...largeformatphotography.info/forum/showthread.php?p=431786#post431786
resulting from an arguably fairly consistent yield) I'd like to know if some bracket shots (with the following criteria) might be capable of resolving the small clock letters (lets stipulate in this example the clock OEM is interested in your piece).

I've included four images that suggest if the lens is outstanding (a given) and a 40x loupe used to dial in the peak Point-of-focus behind ideally perfect ground glass (intentionally staying away from plenums here), with a finely calibrated system (another given) and ISO 25 film flat to within .001" AND a tilt angle of not more than 1.5 degrees (I'm admittedly a newb at the very subtle art of tilting).

So the DOF calculator image seems to indicate 17ft at f22, but perhaps f16 or f11 might work if I'm interrupting the tilt spreadsheet correctly.

The defocus spreadsheet images attached seem to suggest that a very slight defocus might be acceptable and perhaps allow one (in theory) to approach 100 lp/mm resolution (under ultra precise conditions and restricted criteria).

In this model (admittedly not yet verified and poorly integrated) film flatness to .033mm (as an approximation to .001") would be critical since there would is only a tiny amount of defocus allowable (all sources summed) without diminishing the resolution (or bending of straight lines).


DISCLAIMER
I've indicated that these images are of a --DRAFT-- nature as the likely hood of a significant mis-understanding is fairly significant. I do admire your posting the film and DSLR together (I think everyone uses both) because on occasion I've felt the need to use "I'm putting on Body Armour" now or the "Here sharky, Here sharky-Sharky" to try and lighten things up a bit.


I'f anyone sees any glaring mistakes or assumptions please comment.

[Frank Doering]

Kirk, you are right, I didn't set up a test. The point of posting the images was to inject some visual evidence into a thread discussing the resolution limits of LF film as against various digital options. Eyeballing my images, I find them in good agreement with the calculated numbers bandied about here: it would take a file at least four times as large and of the same quality as the 12MP D3 file to get into the ballpark of the (down-sampled) 4x5" scan. I also wanted to offer a visual comparison to those who don't have access to both kinds of system. I suspect that there are many LF shooters on the fence who wonder which jobs can be done digitally with a system that can be amortized in a small business.

Incidentally, I sure hope Nikon primes do better along the edges than this allegedly superb zoom. Being used to LF lenses, I have found the lenses for 35-mm digital to be disappointing. Poor edge performance aside, they all distort significantly, which means you lose detail in the correction. The posted D3 file shows this loss. This is another reason for doing real-world comparisons: if you care about straight lines, your final 35-mm digital file will never quite preserve the detail of your RAW files. As for DR, film still wins in my books. Shooting architectural interiors, for example, I always find myself layering different digital exposures where one sheet of film would capture the brightness range.

Gordon, your point about blur is well taken. It cuts both ways: if you want blur with digital, you'll sometimes need a ND filter in order to get sufficiently long exposure times. I haven't had much luck with stitching because there always occurs some important change during the exposure series that I can't rectify later: the light, the wind, people, cars... One-shot capture is also much easier to compose.

rvhalejr, it would be fun to tinker along the lines you suggest. But it wouldn't be very practical in the real world of impatient bathers waiting to get into the pool. I have never managed to set calculated tilt angles even remotely precisely enough, nor have I been able to estimate Merklinger's parameters. In real life, I start out zeroing the standards with a mirror device and then adjust by ground-glass inspection with a 7x loupe. Doing that off-center opens another can of worms: the tricks played by the Frensnel lens, light falloff...

[bglick]

> I start out zeroing the standards with a mirror device and then adjust by ground-glass inspection with a 7x loupe. Doing that off-center opens another can of worms: the tricks played by the Frensnel lens, light falloff...


Nice post... This is just one of MANY reasons why there is limits in how accurate you can try to focus or align anything with view cameras.

In addition, a 40x loupe on a gg which is only capable of ~ 3 lp/mm resolution, is well..... not providing any added useful information. This is why viewer camera makers sell gg loupes in the 3 - 6x range, as more magnification serves no purpose.....if it did, they would offer us 7 - 40x loupes as well.

[rvhalejr]

... behind ideally perfect ground glass (intentionally staying away from plenums here) ...

Loupe Stands and Plenum (in place of GG when focusing with 35-40x Loupe)
...largeformatphotography.info/forum/showthread.php?p=423824#post423824