Scanning and Resolution Annotated

From Tigwiki

(not ready for publication)

Notes on Peter Swinson's "Scan Sampling to capture 35mm Film at its 50% MTF point" chart.

Peter assumes the scanning size of Super 35mm film to be 24.0 mm. This is a good even number and matches the SMPTE 96M for TV scanned area of a type C aperture. Thus, 5 cycles per mm fills the 24 mm scan area with 120 cycles (also line pairs.)

Note that 5 lp/mm in standard definition TV scanning results in a frequency of:
120 / 56.656 uSec = 2.12 MHz

A scan of 8.45 line pairs per mm will place the NTSC waveform at the subcarrier frequency of 3.58 MHz (resulting in lovely rainbow colors.)

Peter does not indicate if the top period of the bars or the squeezed middle portion represents 5 lp/mm. The ratio is about 1.3:1 so the rate is either 5-6.5 or 3.8-5 lp/mm.

The 4K scan (edge to edge of 24mm) should result in 170.66 samples to describe 5 cycles (I don't know where 128x128 comes from.) The camera aperture for Academy is 0.981" x 0.735" (I don't have a copy of SMPTE 95 handy). This is 24.92mm x 18.67mm which agrees with EBU Tech 3297-E http://www.ebu.ch/CMSimages/en/tec_doc_t3297-2004_tcm6-12768.pdf

The Nyquist rate for 5 lp/mm should be the equivalent to scanning the 24.92 horizontal aperture with 249 samples. The requirement to band limit frequencies below half of the sample rate (colloquial "Nyquist") must render the "240 scan" example as completely flat grey. You can't ignore the requirement to band limit the source before (re)sampling. Any evidence of 5 lp/mm in this example is an indication of alias frequencies created by the failure to band limit.