[filmscanners] RE: Canon IDs vs Pentax 67II

["Austin Franklin" <austin@darkroom.com>]

Hi David,

> "Austin Franklin" <austin@darkroom.com> writes:
> >>>>>>>>>>>
> > Three things are wrong: digital pixels look better than scanned
> pixels in
> > one-on-one comparisons, i.e. comparing images with the same number of
> > pixels, and even in lp/mm comparisons, it just takes more
> pixles and lp/mm
> > in film to get the same apparent sharpness.
>
> That is not necessarily true.  A scan of a negative is very similar to a
> digital camera picture, if you are not resolving anywhere near
> grain...but...if grain becomes a sampling artifact, then you are correct.
> <<<<<<<<<<<<
>
> Dunno what scans you're looking at,

Er, my own ;-)

> but every scan I've ever seen has been
> uninspired at the pixel level compared to the better digital images I've
> seen.

Ah, you need a better scanner then.

> Even 4000 dpi Provia scans have noise levels that (while quite
> reasonable and not a problem at all) are off scale compared to
> what digital
> SLRs produce. And Velvia's a joke.

Aren't those slide films?  Slide films have a higher density range than
negative film...and it's that wider density range that challenges most
scanners.  Scanning negative film gives much better results.  I highly
recommend Portra NC.

> The best 8000ED scans I've produced
> aren't really acceptable at 300dpi, and need to be printed at 450
> dpi before
> I get the quality I want.

What do you mean by 300dpi and 450dpi?  That shouldn't have a thing to do
with scanning.  You should be scanning at FULL native optical resolution of
your scanner, and outputting without decimating the data, and let the DPI
fall where it may.  If you scan, say, a 6x6 at 4k SPI, you get a file that
has H and V pixels of ~2.25 x 4k or 9,000 pixels in either dimension.  To
print that to, say, a 10x10, you would simply change the dimension to 10" x
10", uncheck the box so there is no processing of the image, and therefore
get 900 PPI to the printer...and print.

>(I'm not complaining: I bought into
> scanned MF on
> the theory that 645 + the 8000ED would be adequate for 13x19, and my math
> seems to be holding up.)

Yes, it most certainly is adequate for 13x19.  I scan 35mm at 5080 and print
13x19s all day long with no problem.  MF scans at 2540 give very very high
end 13x19 prints, and I have no problem printing them up to 24 x 24.

> >>>>>>>>
> > The second thing that's wrong is that printing 1Ds images at 250
> > dpi (11x16)
> > results in prints that are about as good as prints get
>
> Not so.  I can send up to 720 PPI to the printer and get noticeably better
> prints.  So I dispute that claim, as I know first hand it isn't true.
> <<<<<<<<<<<
>
> With the Epson 950, if I take a large sharp MF scan, gradually downsample
> it, and print at various ppi settings, the appearance only begins
> to degrade
> at under 250 ppi. I find that good quality 250 ppi images max out the
> resolution of this printer.

I would suggest not downsampling it, just do as I suggest above and let the
PPI fall where it does.

> >>>>>>>>>>>>
> You are also involving things here, like the printer driver, that just
> because the one chosen may not take advantage of additional PPI when
> dithering, that doesn't mean that a "proper" print driver that
> does make use
> of this, won't make a better print!
> <<<<<<<<<<<<<<<
>
> I use Qimage Pro to print, and it automatically resamples to optimize for
> the printer driver.

Hum.  I don't know what it does, but I've made thousands of test prints to
find out what the best output I could get was, and I stated it above.  If
you resample, you are double resampling the image data...as the dither
algorithm will resample as well, and I believe you are degrading the image
further by doing so, at least for the printers I use (Epson 3000).

> >>>>>>>>>>>>
> > In other words, there's no way for _any_
> > technology to
> > make a print that any third party will see as better that an
> > 11x16 from the
> > 1Ds. (So there's no way for MF to win.)
>
> Hum.  I disagree, and know that it's just not true.  Digital will
> have less
> detail, because of the Bayer pattern imaging sensor.  Especially when
> talking about B&W, the tonality and tonal transitions will be FAR superior
> for a B&W image of that size...because it has more PPI to the printer than
> 250, and the difference is significantly noticeable.
> <<<<<<<<<<<<<<
>
> Again, if you take a loupe to your prints. (And B&W digital isn't under
> discussion, since it's a disaster.)

I don't need to take a loupe to see what I'm talking about!

> >>>>>>>>>>>>
> You seem to be basing your claims on resolution only, and that is not the
> only issue here.  First off, there is the Bayer pattern imaging sensor,
> which, will lose SOME data, simply because of architecture.  It may not be
> much, but the loss is real.  Second, is tonality.  I'm not arguing for or
> against, but just pointing out there is more than simple X by Y resolution
> that makes a print!
> <<<<<<<<<<<<<
>
> I think you overestimate problems with the Bayer architecture.
> Bayer sensors
> usually have an anti-aliasing filter in front of them, and exhibit
> resolution of about 70% of Nyquist. But getting more than 70% of
> Nyquist out
> of any digital imaging system is problematic because of aliasing. So there
> really isn't any more resolution to be had. (There are questions of the
> magnitude of the MTF response, and one might be able to low-pass
> filter and
> then downsample without introducing aliasing, so downsampling
> high-res scans
> ought to be able to compete with Bayer imaging.)

I'm not overestimating the problem, just stating that there is SOME
degredation, and there is.  No, it's not a huge number, but it is
significant, and VERY image dependant.

Remember, I have scanning backs and Bayer pattern backs of the same
resolution.  There IS a difference in image quality, no doubt about it!

> >>>>>>>>>>>
> > And finally, it's easier to make good prints from digital originals.
>
> What, exactly, do you mean by that?
> <<<<<<<<<<<
>
> Why do you think Reichmann found his 12x16" digital prints to have more
> detail than his MF scan prints? Because he failed to get the
> detail off the
> scan<g>.

Agreed.  I don't find much of anything he (MR) does can be taken seriously.
He claims to be an "expert", but at every test he fails miserably.

Regards,

Austin

----------------------------------------------------------------------------------------
Unsubscribe by mail to listserver@halftone.co.uk, with 'unsubscribe 
filmscanners'
or 'unsubscribe filmscanners_digest' (as appropriate) in the message title or 
body