[APML] film grain and scanner resolution

[hermperez@worldnet.att.net (Herm)]

Sorry to you guys who also subscribe to the Filmscanner list, I found this
message very interesting.

Herm

>To: Filmscanners@halftone.co.uk
>Subject: Re: 'Grain' and scanning (was Averaging Scan Noise)
>From: TonySleep@halftone.co.uk (Tony Sleep)
>Date: Wed, 19 Jan 2000 14:38 +0000 (GMT)
>
>> Aside from electrical noise in the CCD, what is the contribution of the
>> slide/negative to this noise?  I've heard "film grain" mentioned, being
>> traceable to the size of the photo-active chemicals in the various 
>> layers.
>> But if I recall, another person replied that film grain was extremely 
>> small,
>> and what was seen were "dye clouds," which I guess is the accretion of 
>> the
>> dye during the development process.  Anyway, they're easily seen under a
>> microscope.  
>>     - What are they?
>>     - What size is they for various film types (also pos/neg)?
>
>In colour films, and also 'chromagenic' B&W films using colour processing 
>(XP2, TMax 400CN), the image silver is chemically bleached from the image 
>leaving dye 'clouds' which had been coupled to the grains. These form the 
>visible image. In colour films, the dyes are R,G and B, in layers which 
>were sensitised to the same colours (reversal films create a positive 
>image via a second exposure or chemical action prior to removal of the 
>sensitised silver but after initial development). They occupy the same 
>locations that the silver grains did, but have more diffuse boundaries - 
>dyes diffuse into the surrounding gelatin some - which is why colour 
>images seldom have the acutance of silver B&W. Overlapping dye clouds tend 
>to reduce the impression of graininess. Which is why colour materials such 
>as ISO800 colour neg can look absurdly grain-free alongside conventional 
>B&W of similar rating, albeit at a cost of some image sharpness. This also 
>gives them the characteristic of less apparent 'grain' in densely exposed 
>areas, and is why colour neg shadows tend to be more 'grainy' - the exact 
>opposite of conventional B&W where silver grains clump together in dense 
>areas.
>
>Grain or dye clouds fall in the range ~5-30um, depending on the material, 
>speed, exposure and processing. Pixels in a 2700ppi scanner will be ~9um, 
>in a 4000ppi scanner ~6um. Nyquist requires at least 2x the sample rate, 
>so a 2700ppi scanner cannot resolve grain of ~<18um, and a 4000ppi scanner 
>cannot resolve grain ~<12um, even theoretically. 
>
>In practice, it is worse than that, since grain/dye clouds are not the 
>same shape as pixels but irregular, with boundaries that require a much 
>higher sampling frequency to resolve.
>
>For this reason, the sort of scanners we are using cannot resolve grain 
>clearly with most materials. From empirical tests, a Polaroid 4000 
>achieves a good resolution of grain shape with very fast film with a sharp 
>and coarse grain structure (TMaxP3200), but slower materials 
>give grain images which are less precise because of aliasing. 
>
>From the above, 2700ppi scanners can only begin to resolve grain in some 
>of the coarser (faster) materials, or where there is heavy exposure or 
>processing which creates locally enlarged grains. The same applies, 
>somewhat less, to 4000ppi scanners. Yet grain is commonly seen in scanned 
>images.
>
>OH NO IT ISN'T <g> What is mostly seen is aliasing artefacts which look 
>like grain, and often like quite coarse grain, even from materials which  
>have a fine and subtle grain. 
>
>This is quite clear if you zoom in enough to inspect the structures. It's 
>worst when scanning conventional B&W, as the sharp grain boundaries create 
>the most violent aliasing errors. The result is often unuseably 'grainy' 
>scans from film which prints with low grain - an almost reticulated look.
>
>Because of the variables (grain or dye cloud size and topography) it is 
>impossible to predict which films will scan well, and doubtless different 
>CCD's (and probably internal processing) doubtless complicate matters 
>further. But IME every scanner has particular films which will upset it 
>and produce accentuated grain effects. 
>
>These will not always be the fastest films either; if the particular 
>characteristics of an emulsion exposed and processed in a particular way 
>happen to clash with the characteristics of the scanner, you'll see 
>'grain' even where there really isn't much. Even more upsetting is seeing 
>the effect in some parts of the image and not in others. 
>
>These effects are inescapable until we get CCD's which are sub-2.5um, 
>around 10,000ppi, beyond the Nyquist limit for even the finest-grained 
>materials and capable of rendering higher-frequency topological features 
>too. Whether all this is really worthwhile to faithfully capture grain 
>information we don't need anyway is another matter, but a while back 
>someone sent me comparative samples done at 8,000 and 12,000ppi. 
>Unfortunately I seem to have lost them since, but the difference was 
>startlingly large, not only in terms of grain but also inner detail of the 
>image itself. Drumscanner mfrs. need not throw in the towel just yet.
>
>Regards 
>
>Tony Sleep
>http://www.halftone.co.uk - Online portfolio & exhibit; + film scanner 
>info & comparisons
>====================================================================
>The filmscanners mailing list is hosted by http://www.halftone.co.uk
>To resign, send a msg to ListServer@halftone.co.uk with UNSUBSCRIBE FILMSCANNERS in the 
>title.

--  APML Archives at <http://www.system.missouri.edu/apml/>  ---
                Unsubscribe at <majordomo@seds.org>