HansDerHase
I’ve now tried out various exposure sequences and lo and behold:
I had significantly over-exposed the film using the manufacturer’s recommended 4+4 × 3 sec. tilt. I get a negative that scans well at 2.5+2.5 × 3 sec. tilt when I expose FP4+ at ISO 200. ISO 400 works perfectly well too. At 3.5+3.5 × 3 sec. tilt, the negatives are already too dense for scanning.
I also get good results at 3.5+3.5 × 1 min. tilt.
I.e. manufacturer’s settings vs. mine =
4+4 vs. 2.5+2.5
or
6+6 vs. 3.5+3.5
Can anyone explain why ‘I’ deviate so significantly from the manufacturer’s specifications?
Kind regards,
Bernd
FrankJBeckmann
Hi Bernd,
Such short development times are hardly reproducible. Even pouring the developer in or out a little faster or slower can lead to a significant deviation. And am I right in thinking that you only scan the negatives, but never put them in an enlarger? Scanners often prefer ‘thin’ negatives, which can’t really be used properly with an enlarger. Have you had a look at the shadow detail on the negatives you exposed at 400 ASA? I’d accept 200 ASA for the FP4+ in Emofin, but at 400 ASA there probably isn’t much left to see in the shadows.
HansDerHase
That’s right, I’m only planning to scan the negatives.
At 400 ISO, I still get enough shadow detail (for my taste). But then again, I’m more of a film noir sort of guy :-)
My main problem when I develop for longer than 3+3 × 3 seconds is the increasingly annoying grain. At 4+4, the grain is downright obscene. But that’s probably also because the scanner can barely shine its dim light through the negative at that point and then only captures ‘a few clumps’.
FrankJBeckmann
Hi Bernd,
Scanners that use directional light, in particular, struggle with black-and-white film and tend to magnify the grain. Chromogenic colour films work better in this respect.
If you’re happy to sacrifice detail in the shadows, you can achieve an exposure with as little light as you like.
Renate
Hello,
Scanners are a bit of a tricky business. What looks on paper like perfectly normal film grain can appear completely different after scanning, and the result depends heavily on the settings and the scanner itself. An enlarger cannot simply be replaced by a scanner.
First of all, there’s the question of which scanner was used? How does the transparency unit work? What software was used? What resolution was set?
Many programmes automatically apply sharpening during scanning. This causes the grains to become blurred. The grain then looks very coarse.
If the scanner’s resolution is roughly the same size as the grains on the film, the sampling theorem is violated. This results in low-frequency aliasing, which then appears as a ‘monster grain’ in the image. The ‘monster grain’ isn’t actually on the negative; it’s only created by the transformation during scanning.
As a rule of thumb, you should ensure that the scanner’s resolution is either significantly smaller or significantly larger than the size of the developed grain.
Best regards,
Renate
cfb_de
Hi Bernd,
Ooh... sounds like physics.
Yes. What chemistry is to analogue Duka, physics is to scanning. Without a basic instinct for the fundamentals, neither will work.
But how do I know how large my developed grain is?
A magnifying glass with a scale. Aka a ‘thread counter with a scale’. Those handy stereo microscopes from, say, Zeiss or Leica would be ideal. You don’t have to buy one; some chemists have them, and goldsmiths and watchmakers usually do too.
I’m only marginally interested in these issues, though: I work purely with chemicals in the classic Duka process. There, I don’t have to worry about a scanner’s resolution theorem, no ‘ICE’, no ‘automatic creation’ or any other rubbish ;-)
Best regards,
Franz
Wolfgg
So let me explain the sampling theorem:
The sampling theorem is nothing mysterious. Anyone who owns an audio or video CD or DVD has it right there in their living room, so to speak. Without it, top-class quality would simply be impossible. Take audio, for example: to record frequencies up to 20kHz without distortion, the sampling theorem stipulates that the sampling frequency must be at least double the highest signal frequency – in this case, at least 40kHz. Why? Because, mathematically speaking, sampling is nothing more than multiplication, which always results in the sum and the difference between the sampling frequency and the signal frequency(ies). And the difference is where the problem lies. If, for instance, a 20kHz sound were sampled at only 30kHz, the resulting difference frequency of 10kHz would overlap with the original signal at 10kHz and thus cause interference. Or, to take an extreme example: sampling 20 kHz at just 20.1 kHz produces 0.1 kHz. These new frequencies, created solely by the sampling process, are called beat frequencies. In practice, the sampling frequency is set to at least three times the highest signal frequency, so that there is still some margin for filtering.
Although there are no sounds on the film, the same applies to spatial frequencies. These are nothing more than regular line patterns (e.g. 100 line pairs per mm consists of 100 lines followed by an equally sized gap per mm; in other words, a black and a white line, each 1/200 mm wide, together constitute a line pair). Such a grid can also be created by the film grain. If a scanner were now to accidentally capture such a line grid with 100 line pairs per mm at 110 samples per mm (equivalent to 2794 DPI), it would produce a beat frequency of 10 line pairs per mm (110 minus 100). It should be noted, however, that this structure with 10 line pairs per millimetre was not actually present on the film! It only arose due to the violation of the sampling theorem in the scanner and is also transferred to the PC. Here, one would need to scan at a minimum of 300 samples per millimetre (equivalent to 7620 DPI) in order to transfer the information from the film to the PC reliably and without distortion.
A few more figures from earlier measurements: the film grain on TMax400 is approx. 1.7 µm, on Technical Pan approx. 0.7 µm, both estimated as accurately as possible under a microscope at 1000x enlargement (Franz: you can’t see anything with a thread counter). It was assumed that no silver-based developers were used. Nor have any grain agglomerations been taken into account, which can occur depending on the developer (Rodinal!). Assuming a regular grain arrangement, i.e. grain plus a gap of the same size as the grain, this then leads to 1/(2*1.7 µm) = 294 line pairs/mm for TMax400, and for Technical Pan to 1/(2*0.7 µm) = 714 line pairs/mm as the ‘grain grid’. One can then calculate the critical scanner resolutions. A TMax400 should be scanned either at less than 150 samples/mm (3810 DPI), corresponding to 2 times the grain plus gap per sample (undersampling, grain then disappears) or at over 900 scans/mm (22,860 DPI!), corresponding to 3 scans per grain plus gap (oversampling, grain scanned cleanly); this ensures you are on the safe side. For the Technical Pan, use either below 350 scans/mm (8890 DPI) or the rather extreme 2100 scans/mm (53340 DPI). These resolutions refer not only to the sensor line alone, but to the entire system, i.e. including the scanning optics and any post-processing such as unsharp masking. Of course, these are not scientifically precise figures, but they are sufficient guidelines for practical purposes.
I hope the explanation was understandable even for those who have no interest in physics.
Regards, Wolfgang
cfb_de
Hello Wolfgang,
Franz: a thread counter won’t do the trick here
. Decent microscopes have a scale superimposed in the image capture or can overlay a grid on the image :-) Or they have a digital camera screwed on top, which opens the door to all sorts of shenanigans. Including the fact that, despite the Imacon camera, I see more in the real image than is left in the digital image afterwards. Which is why there’s still a film camera mounted on top of the state-of-the-art Leica microscope. I’m talking about work here.
Besides, with silver film, microscopy is only of limited use for determining spatial frequencies on the negative. Unfortunately, the grain size is only statistically distributed, but by no means always the same at every point... Probably *the* problem with these colour-image-optimised digital scanner crutches called "ICE".
The idea is to patch up from behind with a fist what the plastic mechanics can’t manage. So that the sharply defined dust can be detected in the colour image cloud – that’s a nice side effect. A sort of "reverse sampling theorem", so to speak. The dirt just shows up less :-)
*Edit*: Blimey. That should read: "The dirt just shows up better :-)"
Best regards,
Franz
