A bit of lens testing:
I'm still comparing my Canon 50mm Compact Macro, with the 1:1 converter, against my previously tested EL-Nikkor 105. The Canon has a significant advantage--autofocus confirmation.
But the shorter macro lens suffers from excessive fall-off, in the range of about 15%, at all apertures. I suspect this is a combination of the shorter focal length and the response of the sensor. The Nikkor's falloff is more like 3-5%. I've written in the light-source thread that I'm using an LED-backlit computer monitor panel to provide the backlight, and it's very even in an of itself. Using that for a light source also provides the means create a correction pattern that can be displayed on that monitor. More experimentation needed there, but I'm expecting that to be fairly easy. The panel is parked well behind the film being scanned so that it's pixel pattern is completely out of focus.
One test I used to determine lens performance: The size of the resulting files. I could not tell by inspecting the files of a normal negative which aperture was the sharpest, but the JPEG files sizes were: 3361, 3757, 4402, 4064, 3498, and 3225, for f/5.6, 8, 11, 16, 22 and 32, respectively. This outcome is nicely monotonic, peaking at f/11, which is completely reasonable. The EL-Nikkor file sizes ran about 15% higher than the Canon macro sizes, which were also nicely monotonic, peaking (unhappily) at f/22. That the Canon lens needed such a small aperture was not a positive statement in its behalf--aberrations took a lot of correcting before diffraction overcame them.
Remember that my camera is a 12.5MP Canon 5D when judging these files sizes. Raw files were in the 13-14 MB range, followed the same curve and were proportional to the JPEG file sizes.
Looking at the best of these files, which was the f/11 file for the EL-Nikkor, I see that the scan is not sharp to the pixel. Dust specs on the negative, which show up as white after inversion, have edges smeared over two or three pixels. They are sharper than the image on the negative, which I take to be limited by the photograph itself, though I was able to make very sharp 16x20 prints from this negative back when I had my darkroom. The taking lens was a Schneider Super Angulon 121mm f/8 at f/32. The original image is posted here.
I don't think I'm doing any better than the Epson than I was with my first quickie test, to be honest. I still cannot resolve beyond a hint of grain.
This picture is the whole 24x36 scan frame:
Here is a 100% detail (or it will be after you click it). The staple is about 1mm on the 4x5 negative. I've adjusted levels and curves to match my conception of the image (as linked above), but this is without any sharpening.
Here it is with a bit of sharpening to overcome the effects of the anti-aliasing filter on the camera. The sharpening setting is 0.6 radius, amount 200, zero threshold. I can just see a hint of grain, maybe. The detail in the negative isn't, I don't think, sharp enough to benefit much from sharpening at this level, but this is basic corrective sharpening, not the sharpening I might do when targeting the image to a particular print size.
Maybe this is just as good as my 5D can do at 1:1. Or maybe I just need a better lens.
At least there's no distortion. I photographed a clear plastic ruler, and drew a line along the edge. At the resolution of my camera and lens, the edge was straight within a pixel along its entire length.
Rick "taking baby steps" Denney
You might try downloaded Zerene stacker, http://zerenesystems.com/cms/stacker, which has a 30 day trial. Get the focus at the confirmation spot. Now back off the Velmex a small bit, about a 90* rotation. Take a series of photos, moving the velmex knob about 10* with each step. Go past the confirmation point by a good number. Bring the images into Zerene and run a Dmap stack. Now compare the image to what you took with a single frame. This will not only get rid of any focusing error, it will also help if the lenses' plane of sharp focus isn't really planar, and it'll alleviate any misalignment issues.
Rick "back to the basement" Denney
Rick, I wonder about the difficulty of aligning the plane of the sensor to that of the film. Maybe it would be possible to make use of a cheap laser pointer. How about shooting the beam through the film mounting glass and reflecting it off a mirror taped to the front of the DSLR lens. Then adjust the DSLR plane to shoot the beam exactly co-linearly back through the film plane. A bit of fussing with this, but if the DSLR lens axis is precisely orthogonal to the sensor (high quality DSLR should be I suppose) then you should have a pretty good alignment. As I recall in your setup you may not be using a glass plate for film mounting - but maybe you could substitute one. The laser pointer would be temporarily substituted for the normal light source. Just thinking about alignment issues when I start fiddling with this.
Nate Potter, Austin TX.
Okay, experiments have continued.
The focus was correct with the EL-Nikkor. I made a series of exposures and looked at each of them at high magnification, and I had to turn the Velmex handle a whole turn (0.025") to noticeably degrade the image at f/11 (indicated, f/22 effective) and 1:1. Everything in the range of +/- 0.015", at the very least, looked identical. So, I stopped worrying about focusing. I just focused one way until the the image, blown up on the camera's LCD, was noticeably less sharp, turned it the other way past focus until I got the same result, and then split the difference. Just like I used to do with the enlarger.
I also set the camera to lock up the mirror, pause three seconds, and then open the shutter, using a remote cable. Made no discernible difference at any shutter speed.
What was not correct with the EL-Nikkor was the EL-Nikkor. It's okay, but I doubt it's optimized for this magnification. And it's just a wee tad long to get much greater magnification with my setup. And my Canon macro lens only goes to 1:1.
So, I tried something different. I took a cheapie 80mm f/2.8 Arsat lens, put it on an Arsat reversing ring, and mounted it on my bellows arrangement reversed. The Arsat has a 62mm filter ring--not common--so that's the lens I have that will fit that reversing ring. I had bolted my bellows standards down onto an Arca plate, so I just slid the whole thing forward, focused it, and made some photos. Magnification is greater--about 1.8:1. The reversing ring provided considerable additional extension, as it turned out. So, the field of view is 13mm by 20mm. I can pretend I'm using my wife's D300, heh.
For the curious, the Arsat is a 6-element double-gauss lens, multicoated, and a good performer used normally except at apertures larger than f/5.6. It is, of course, a medium-format lens intended for 6x6, and this application is just using the sharpest part in the middle of the field. In this case, the greatest detail was recorded at f/11 according to the file size, but I could not see any improvement over f/8. f/5.6 was visibly worse, as was f/16.
And there is the grain. Once I could see grain, I decided to exaggerate the sharpening, partly to demonstrate Ben's point. So, I used a radius of 0.7, a threshold of 0, and an amount of 500. Dust spots on the negative became crisply resolved to the pixel, and I think I'm now getting everything there is to get from the sensor. Note that what made this possible was seeing grain--seeing grain allows radical sharpening, because the sharpening is increasing the apparent acutance of the grain. Given that there is no signal at that scale, it does not increase the signal/noise ratio at larger scales. Sure enough, even viewing the 13x20mm piece of the original negative as a single image does not show anything untowards with that much sharpening at the grain level.
This suggests to me that a key to this approach is providing enough magnification or resolution to resolve grain well enough to be able to really apply sharpening to it. That improves the MTF at larger scales without the grain becoming more apparent. But if you look at the image at much greater than 100% on the monitor, it will look a bit freaky.
Comparing this image with the one from the EL-Nikkor is interesting. The EL-Nikkor resolved some grain--there are larger grain clumps visible in both images--but not enough to allow that sharpening strategy without it revealing artifacts at larger scales.
So, instead of making multiple 1:1 exposures to reduce noise, my intention is to make the same number of exposures overall, but smaller tiles instead of redundant tiles.
Here's the 13x20 sample (on my monitor, the image, when you click below, is a piece of what would be a 24x30" print):
And here is the detail at (if you click it) 100%--a 50x enlargement if your monitor has 100 pixels/inch:
Both are post-sharpening.
Rick "making progress" Denney
Don't be confused by the apparent distortion--the door on the front of the Espada Mission does not have one single right angle on it, and the photo itself was not corrected for laterl perspective convergence. Since the subject was not parallel to the film when the negative was exposed, there is a touch of change in apparent sharpness across the face of the door. I was also trying to keep the stone door frame in acceptable focus--the shapes of those stones are the design point of that doorway. I'm evaluating sharpness mostly using dust spots and grain, as best I can.
Rick "noting that the camera axis is normal to the film holder within a thousandth or two over a little more than a foot" Denney