Archived posting to the Leica Users Group, 2002/12/17

[Author Prev] [Author Next] [Thread Prev] [Thread Next] [Author Index] [Topic Index] [Home] [Search]

Subject: Re: [Leica] digital versus film
From: Johnny Deadman <lists@johnbrownlow.com>
Date: Tue, 17 Dec 2002 14:25:09 -0500

On Tuesday, December 17, 2002, at 01:25 PM, Randy Jensen wrote:

> A simple calculation I did (based on some leaps of faith, of course) 
> is that
> to match slow film (ISO 50 or so) you would need an equivalent 
> megapixel
> amount of about 28MP.  This is how I came to that number.
>
> With a decent lens and film such as Fuji Velvia you can expect a lpm of
> about 45.  Multiply this by the long side of a 35mm neg (36mm) and you 
> get
> 1620 lines across.  This is theorhetically the maximum number of lines 
> you
> could discern from the far left to far right of the neg (or slide).
>
> Nyquist theorem dictates that you need to double sampling resolution to
> record detail.  But this only works for things that are in phase, such 
> as
> audio.  That's why CDs sample audio at 44.1kHz.  People can hear up to 
> 20k
> (at best) so to record detail that high you need to at least double 
> that
> (and there's debate that 44.1k isn't high enough, but that's WAY off 
> topic).

Where do you get this from? What do you mean by 'in phase'? A thing 
can't be 'in phase' on its own, it has to be 'in phase' WITH something. 
Do you mean a continuous oscillating signal?

> Anyway, for resolution of an image, doubling is not nearly enough.  
> You need
> to multiply it by 4 (can anyone say "Reciprocity Law"?  lol.  It's like
> doulbing in both directions (2x2=4).  This link hints that it may be 
> more
> like a factor of 6, but I'll be conservative with 4:
> http://clarkvision.com/imagedetail/sampling1.html

The Nyquist criterion is to do with anti-aliasing, where the sampling 
of a repeating signal at too low a frequency produces an artefact of a 
signal at a lower frequency. It's no good just squinting at lines and 
saying 'hmph, doesn't look sharp to me'. In the examples you give the 
image sampled at the Nyquist frequency clearly resolves the lines at 
the correct frequency, which is all we care about. It may not be pretty 
or perfect but the lines are resolved. It's always nice to have more 
pixels, but since you are basing your calculation on resolution of lpm, 
the Nyquist frequency, or preferably a bit above it, is just fine.

BTW the 'doubling in both directions' is a canard. You are already 
'doubling in both directions', because when you double the pixel 
density of a sensor you double it in both directions, thus increasing 
the actual number of pixels by four.

Boy, I feel like Austin today. Next!
- --
Johnny Deadman

http://www.pinkheadedbug.com

- --
To unsubscribe, see http://mejac.palo-alto.ca.us/leica-users/unsub.html