[Leica] Re: Scanning Tri-X

[alex at zabrovsky.com (Alex)]

Sounds like a scientific-pretending crap by someone trying to boast his, 
probably hollow fundamental EE knowledge base.
I can try to put it simple what he might intend to impose (not necessary 
phtoography related, pulled out from just a plain, dry B.Sc.EE signal 
processing fundamentals about signal samlping theory)"

Sampling theory is the basis of the signal processing fundamentals and 
basically rules how the analog signal must be sampled (get converedt 
into digital form) in order to be able to reproduce it reliably and 
faithfully. I.e. if one is willing to convert analog signal into digital 
domain or whatever reason and then to get it converted back to analgo 
for thuthfull reproduction, the signal must be samlped acording to 
certain rules, the most fundamental of which is caleld Nyquist sampling 
theorem. To stay on topic, we may regard scanning of photographic film 
as sampling process in order to convert analog (continuous) media into 
digital (discrete) domain.
The Nyquist theorem claims (in its simplest form) that in order to 
preserve the original frequency (spectrum) content of the analog signal, 
its sampling rate must be at least twice as higher as the highest 
frequency component of the original analog signal. Just think of audio, 
music that you used to listen. You want to convert it into digital 
domain (say, for computer procesing, MP3 coding, whatever), so you must 
sample ot to get it represented as discrete signal values in time line. 
Assuming you know ahead that the highest audible frequency your ear is 
able to discern is 15 kHz, that means that to preserve that original 
quality that audio signal must be samlped at the rate no less then 30 kHz.
However, in practice, most real complex spectrum signals are featured by 
infinite frequency content, i.e. even though our ear isn't able to 
discern higher then, say, 20 kHz, the actual signal may contain much 
higher frequency components, even up to infinity (though with much 
lesser power). Obviously, analog signal of infinite spectrum content is 
impossible to sample obtaining truthful frequency content, just because 
theoretically is must be sampled at the rate which is at least twice as 
high then infinity, which is impossble. Thus, in such cases, the 
tradeoff must be made: we usually choose the highest meaningful 
frequency content per perticular application (the rest of the higher 
spectrum is regarded as undesirable noise) and then sample the signal at 
the rate that is at least twice as faster then that highest meaningful 
frequency.

Now back to our photographic "land". We regard the film as analog, 
continuous media featured by infinite "spectrum". Scanning posses 
digital sampling process, thus seemingly yet Nyquist rule can be 
applied. However, since we said the film's "spectrum" is not limited (at 
least in ideal world), in oder to samlpe it we must to choose some kind 
of meaningful "spectrum" content limit. For practical reasons it is fair 
to let that limit as film grain size, beyond of which there is no 
meaningful image information to look for. So that in order to assess the 
minimum necessary scanning resolution in order to attain the highest 
possible resolution in digital domain up to grain sizes, we have to take 
the grain size (say in mm of inches) and then obtain the necessery 
Nyquist sampling rate to be at least twice of that size (actually 
scanning resolution twice as high as the grain size).

I have no idea about grain sizes of particular films, but we can try to 
approach the problem from the opposite side: basing on our knowledge of 
available sampling resolution. Assuming our scanner is capable of true, 
optical 4000 dpi (makes 157.5 dots per mm), it is able to resolve 
details as small as 78 dots per mm or 2000 dpi, or in such case, the 
grain (our artificial meaningful informaiton limit) should be of size of 
1/78 mm (or 1/2000 inches). If the actual grain size is larger then that 
- we have set an excess scanning resolution that will pick the grain 
which may interfare with image details. On the other hand, if the garin 
size is smaller - our sampling frequency may be enough to resolve image 
details but will not be enough to clearly resolve the grain pattern, in 
which case, apparently this is good one. Having said that, is the garin 
is much smaller then that, we are at rist of loosing very fine details 
because our sampling resolution will not be high enough to resolve these.
In the case the grain actual size is very close to Nyquist limit of our 
particular scaning resolution - we may have the case of what is called 
grain alaising.

Hope what I tryed to convey is reasonably comprehensable, or I'll be 
glad for someone more knowledgable to cclarify the things even further 
and/or probably correct my ommisions in the aforementioned).

Alex



Lawrence Zeitlin wrote:

> On another list someone complained about the "posterization" and  
> aliasing encountered in scanning Tri-X film. Here is the quote:
>
> "both my 2880ppi and 4000ppi
> film scanners posterize my Tri-X negatives... which
> makes sense.  there is also nothing you can do to
> remove sampling errors after you scan.  My early
> analysis of Tri-X grain showed that it has strong
> frequency content around 4000ppi so when my scanners
> sample at around half the Nyquist frequency the
> aliasing is just a fact of life.  With finer grained
> films, the grain is still aliased, BUT since the
> signal to noise ratio is so much higher the actual
> useful *image* data is not lost and *luckily* the new
> aliased *grain*, while not an accurate representation
> of the original, is still aesthetically pleasing."
>
> Can someone explain what he means in plain language?  I've scanned  
> hundreds of Tri-X negatives with my Minolta Dimage 5400 scanner at  
> 4000ppi, and, apart from the inevitable dust spots on poorly stored  
> negatives, I have yet to see what he means. If there is a problem, is  
> the scanner software correcting it automatically?
>
> Larry Z
>
>
>
>
> _______________________________________________
> Leica Users Group.
> See http://leica-users.org/mailman/listinfo/lug for more information
>