Digital BW, The Print

I disagree with your application of Nyquist's theorem.  Nyquist's theorem
applies to periodic signals.  An individual line (white or black) is not
periodic, but a line-pair (one black and one white) is.  So, Nyquist's 2x
sampling theorem applies to the lp/mm number, not the lines/mm number.

----- Original Message -----
> From: "Austin Franklin" <darkroom@...>
>
> 40 lp/mm is 80 lines/mm.  But...to actually reliably resolve 80 lines/mm,
> you need to sample at 2x that, or 160 lines/mm...and that only is on axis,
> off axis you need 1.4x, or 224 lines/mm...  24mm x 224 x 36mm x 224 =
> 43,352,064 or 43M pixels...and this doesn't take into account that a
single
> shot sensor only contains %25 real color information...so to get 43M REAL
> color pixels, you'd some multiplication of that...

Hi Thomas,

> I disagree with your application of Nyquist's theorem.  Nyquist's theorem
> applies to periodic signals.

I disagree, Nyquist's theorem does not only apply to periodic signals.

> An individual line (white or black) is not
> periodic, but a line-pair (one black and one white) is.  So, Nyquist's 2x
> sampling theorem applies to the lp/mm number, not the lines/mm number.

Well, of course I disagree and this is common sampling theory...and is
correctly applied.  In order to RELIABLY detect a single line of n width,
you need to sample at 2x n.  Very simply explained if you sample at n,
instead of 2n, and the line falls dead between two sensors you will get only
grey, hence the reason that Nyquist does apply in this application.

Austin

> ----- Original Message -----
> > From: "Austin Franklin" <darkroom@...>
> >
> > 40 lp/mm is 80 lines/mm.  But...to actually reliably resolve 80
> lines/mm,
> > you need to sample at 2x that, or 160 lines/mm...and that only
> is on axis,
> > off axis you need 1.4x, or 224 lines/mm...  24mm x 224 x 36mm x 224 =
> > 43,352,064 or 43M pixels...and this doesn't take into account that a
> single
> > shot sensor only contains %25 real color information...so to
> get 43M REAL
> > color pixels, you'd some multiplication of that...
>