Pages: 1 2 [3]   Go Down

Author Topic: Testing P65+ for architecture tomorrow.  (Read 19050 times)

bradleygibson

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 828
    • http://GibsonPhotographic.com
Testing P65+ for architecture tomorrow.
« Reply #40 on: March 24, 2009, 04:23:20 pm »

Quote from: BJNY
Thanks, Bradley.

However, I don't know what a line pair is.

What are you guys saying in non-scientific terms?

Sorry, Billy.

Like Eric says above, the whole goal of MTF testing is measure how well a lens (optical system) can copy a known pattern.  The pattern and the way it is measured is intended to be representative of how real world scenes are rendered (even though they look nothing alike), and the MTF is not a complete description of a lens' performance (factors such as bokeh, curvature of field, close-focus performance and other issues are not captured by the standard single MTF curve).

Still it is a useful tool in measuring lens ability to render fine contrast (what we'd call 'detail' in an image) and coarse contrast (what we'd call microcontrast in an image).

The fine detail is represented a chart with typically 30 or 40 line pairs per millimeter (40lp/mm or 40 cycles/mm).  Sometimes you'll see this written as 40 lines per millimeter, but this is incorrect--they mean 40 line pairs per mm.  This means that you will ask the lens to accurately reproduce a line that is 1/80mm in width, then reproduce a strongly contrasting line immediately adjacent to it, also 1/80mm in width, again and again, repeating across the frame. This is quite difficult, and even a very good lens might only reproduce around half or even less of the available contrast of the test target.

This level of performance would be shown by a line travelling across the the middle of the graph (representing 50% performance).  In practice, the lines are never perfectly straight, as it is progressively more difficult to achieve the same level of performance as you move away from the center of the lens' image circle.  Lines will typically curve down, indicating poorer reproduction.

The microcontrast is represented by the 'coarse' lines, at perhaps 10-15lp/mm.  Because these lines are wider, they're somewhat easier to reproduce (these lines are always higher on an MTF chart).

Hope that helps,
-Brad
« Last Edit: March 25, 2009, 11:21:34 am by bradleygibson »
Logged
-Brad
 [url=http://GibsonPhotographic.com

brentward

  • Newbie
  • *
  • Offline Offline
  • Posts: 37
Testing P65+ for architecture tomorrow.
« Reply #41 on: March 24, 2009, 05:57:26 pm »

Quote from: georgl
(remember "11MP-1ds beats 6x7 Velvia"!?)

I'm glad someone else remembers that farce...
Logged
___________________________________
brentwardphoto
splashpour.com

rainer_v

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1194
    • http://www.tangential.de
Testing P65+ for architecture tomorrow.
« Reply #42 on: March 24, 2009, 07:16:33 pm »

Quote from: brentward
I'm glad someone else remembers that farce...
was not a farce.
6x7cm is about 30% of the size of a 4x5".
11mp is 33% of 33mp, which is comparable to 4x5" drumscanned film in most situations.

i compared at that time also carefull my kodak slr ( 14mp) with my snacmate5000 drumscanned 6x9cm slides.
the slr didnt reached completely velvia resolution at 6x9cm filmsize, but nearly.
« Last Edit: March 24, 2009, 07:17:11 pm by rainer_v »
Logged
rainer viertlböck
architecture photograp

georgl

  • Full Member
  • ***
  • Offline Offline
  • Posts: 140
Testing P65+ for architecture tomorrow.
« Reply #43 on: March 25, 2009, 05:19:31 am »

Another important thing: one cycle (line pair) needs two pixels, when you have an image with 1000pixels height you can resolve (theoretically) 500 cycles, in reality, the contrast at 500 cycles already has to be zero to avoid alaising (AA-filter) and you lose additional resolution by bayer-filtering (you usually don't photography lines that align with your pixels and are black/white, do you?). Post-processing can generate very high results when measuring those test patterns (horizontal/vertical black lines on white ground) especially when using sharpening tools.

Sorry to come up with these numbers again, but when you don't reach the resolution of a 1ds with a 6x7 (56x72mm)-Velvia, something has gone terribly wrong! To match a perfect (!) 1ds-file you just need to resolve 25lp/mm (or cycles)!

A 35mm-slide would just contain 2MP at this frequency!

My M8 (sensor is 60% smaller) is very similar to my MP loaded with Velvia resolution-wise (depends on contrast), although the "look" is entirely different.

Here is one of the few comparisons with a decent scanner (Imacon @8000ppi downsampled to 4500ppi): http://www.boeringa.demon.nl/menu_technic_..._resolution.htm

The lines beside the "20" are 43lp/mm. And there should be MF-(maybe even LF?)lenses that reach the MTF of a Minolta 1.7 50mm...

Bloody theory, were are the 8x10"-scans?  
« Last Edit: March 25, 2009, 05:22:30 am by georgl »
Logged

eronald

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6642
    • My gallery on Instagram
Testing P65+ for architecture tomorrow.
« Reply #44 on: March 25, 2009, 05:31:40 am »

Quote from: georgl
Bloody theory, were are the 8x10"-scans?  

I guess we'll see them when Jonathan gets back in a couple of weeks.

Edmund
Logged
If you appreciate my blog posts help me by following on https://instagram.com/edmundronald

brentward

  • Newbie
  • *
  • Offline Offline
  • Posts: 37
Testing P65+ for architecture tomorrow.
« Reply #45 on: March 25, 2009, 09:45:53 am »

Quote from: rainer_v
was not a farce.
6x7cm is about 30% of the size of a 4x5".
11mp is 33% of 33mp, which is comparable to 4x5" drumscanned film in most situations.

i compared at that time also carefull my kodak slr ( 14mp) with my snacmate5000 drumscanned 6x9cm slides.
the slr didnt reached completely velvia resolution at 6x9cm filmsize, but nearly.

A very nice MF  lens (pentax 67 100 macro for example) should be able to render velvia's 80+lp/mm where as a non digital 4x5 lens should be in the 40-50 lp/mm range. I have 40x50's from Pentax67 + 100 macro scans that the original 1ds couldn't touch and even the 1ds II doesn't match it.
Logged
___________________________________
brentwardphoto
splashpour.com

georgl

  • Full Member
  • ***
  • Offline Offline
  • Posts: 140
Testing P65+ for architecture tomorrow.
« Reply #46 on: May 06, 2009, 01:18:16 pm »

Any news?
Logged

Snook

  • Guest
Testing P65+ for architecture tomorrow.
« Reply #47 on: May 06, 2009, 02:16:59 pm »

Yeh he decided to buy a Nikon D3x...:+}
Snook
Logged

eronald

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6642
    • My gallery on Instagram
Testing P65+ for architecture tomorrow.
« Reply #48 on: May 07, 2009, 04:22:00 pm »

Quote from: Snook
Yeh he decided to buy a Nikon D3x...:+}
Snook

We both already had Nikon D3x when we did the test. Jonathan seems to have disappeared, probably on a work trip, I was thinking about asking him whether he wanted to buy my DB

Edmund
Logged
If you appreciate my blog posts help me by following on https://instagram.com/edmundronald

Jonathan Wienke

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 5829
    • http://visual-vacations.com/
Testing P65+ for architecture tomorrow.
« Reply #49 on: May 23, 2009, 07:58:50 am »

Quote from: georgl
Another important thing: one cycle (line pair) needs two pixels, when you have an image with 1000pixels height you can resolve (theoretically) 500 cycles, in reality, the contrast at 500 cycles already has to be zero to avoid alaising (AA-filter) and you lose additional resolution by bayer-filtering (you usually don't photography lines that align with your pixels and are black/white, do you?). Post-processing can generate very high results when measuring those test patterns (horizontal/vertical black lines on white ground) especially when using sharpening tools.

Sorry to come up with these numbers again, but when you don't reach the resolution of a 1ds with a 6x7 (56x72mm)-Velvia, something has gone terribly wrong! To match a perfect (!) 1ds-file you just need to resolve 25lp/mm (or cycles)!

A 35mm-slide would just contain 2MP at this frequency!

My M8 (sensor is 60% smaller) is very similar to my MP loaded with Velvia resolution-wise (depends on contrast), although the "look" is entirely different.

Here is one of the few comparisons with a decent scanner (Imacon @8000ppi downsampled to 4500ppi): http://www.boeringa.demon.nl/menu_technic_..._resolution.htm

The lines beside the "20" are 43lp/mm. And there should be MF-(maybe even LF?)lenses that reach the MTF of a Minolta 1.7 50mm...

What you're forgetting is that film has a built-in dithering pattern due to the grain structure that digital does not. It's similar to the difference between an inkjet printer and a continuous-tone dye-sub printer. The inkjet has to have a much higher nominal resolution (~1440 DPI) to match the ability of a 300 DPI  dye-sub to comparably render continuous tones and smoothly blend color gradients. The inkjet can thoroughly kick the dye-sub's ass when it comes to printing fine high-contrast black-on-white subject matter such as text or your resolution test patterns, but when it comes to rendering real-world images, the resolution gap between them is insignificant due to the loss of resolution caused by the inkjet's need to dither a pattern of ink droplets to print colors that do not exactly match an ink color.

The conclusions you draw from your scan comparison are bogus because you are only considering scans of high-contrast test patterns where the dithering effect of the film grain affects resolution the least. If you were to make a low-contrast RGBCMY version of your test chart and photograph and scan that, the results would favor film much less.
Logged

Conner999

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 932
Testing P65+ for architecture tomorrow.
« Reply #50 on: May 23, 2009, 11:45:06 am »

This may sound like a simplistic question, but what of say a higher-end DSLR + (to pick one) Mamiya TS Auto Bellows M645 and choice of M645 glass a as budget option (vs. MF) as opposed to a  dedicated CaniKon TS lens?
Logged

Conner999

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 932
Testing P65+ for architecture tomorrow.
« Reply #51 on: May 23, 2009, 12:22:40 pm »

Excellent link provided. Thanks

Quote from: georgl
Another important thing: one cycle (line pair) needs two pixels, when you have an image with 1000pixels height you can resolve (theoretically) 500 cycles, in reality, the contrast at 500 cycles already has to be zero to avoid alaising (AA-filter) and you lose additional resolution by bayer-filtering (you usually don't photography lines that align with your pixels and are black/white, do you?). Post-processing can generate very high results when measuring those test patterns (horizontal/vertical black lines on white ground) especially when using sharpening tools.

Sorry to come up with these numbers again, but when you don't reach the resolution of a 1ds with a 6x7 (56x72mm)-Velvia, something has gone terribly wrong! To match a perfect (!) 1ds-file you just need to resolve 25lp/mm (or cycles)!

A 35mm-slide would just contain 2MP at this frequency!

My M8 (sensor is 60% smaller) is very similar to my MP loaded with Velvia resolution-wise (depends on contrast), although the "look" is entirely different.

Here is one of the few comparisons with a decent scanner (Imacon @8000ppi downsampled to 4500ppi): http://www.boeringa.demon.nl/menu_technic_..._resolution.htm

The lines beside the "20" are 43lp/mm. And there should be MF-(maybe even LF?)lenses that reach the MTF of a Minolta 1.7 50mm...

Bloody theory, were are the 8x10"-scans?  
Logged

georgl

  • Full Member
  • ***
  • Offline Offline
  • Posts: 140
Testing P65+ for architecture tomorrow.
« Reply #52 on: May 23, 2009, 01:35:01 pm »

"The conclusions you draw from your scan comparison are bogus because you are only considering scans of high-contrast test patterns where the dithering effect of the film grain affects resolution the least. If you were to make a low-contrast RGBCMY version of your test chart and photograph and scan that, the results would favor film much less"

Yes, film is much more contrast-dependent than digital but this test is also partly in favor of digital, because it's not color (a bayer-filtered sensor interpolates color information) and the line pattern align with the pixel-structure.

There are also some interesting degrain-algorithms coming from the movie-industry (I'm not sure if we will ever see them as consumers):
http://www.digitalfilmcentral.com/docs/ARR...s_April2009.pdf

My point was that most comparisons are based on crappy scans with inferior processing methods - otherwise we wouldn't come up with the conclusion that 35mm has 2 Megapixels resolution (1ds =6x7)!

Anyway, were are the scans!?!  
« Last Edit: May 23, 2009, 01:37:19 pm by georgl »
Logged
Pages: 1 2 [3]   Go Up