Pages: 1 [2]   Go Down

Author Topic: Hyper-reality for dummies  (Read 10266 times)

Rob C

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 24074
Re: Hyper-reality for dummies
« Reply #20 on: May 15, 2012, 04:22:28 pm »

it's an engineer thing....


Nancy, please be explicit. I have no idea to what you refer, a position that leaves me at a distinct disadvantage. It's not fair, your knowing and my not knowing.

Rob C

timparkin

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 66
Re: Hyper-reality for dummies
« Reply #21 on: May 29, 2012, 11:53:54 am »

The article was almost complete bollocks - yes there is a difference between MF and 35mm DSLR but it's not this 'wine tasting expert sense' thing he's talking about. Here's a list of differences that for me are important

1) Mostly better lenses - however good primes on 35mm SLR can do very well indeed. The simpler the lens design, typically the more rounded the result. Older lenses were designed by trial and error, with designers prioritising 'rendering' as well as resolution. Modern SLR lenses tend to be computer designed. Hence why leica lenses tend to have 'that look'. I have a slight problem with MF wide lenses as the need for strong retrofocus design makes for more complicated lens designs - the results aren't bad but they don't have the relaxed look of longer lenses or LF wide lenses such as the Super Angulons or MF Biogons.

2) Oversampling - not to be ruled out. bayer arrays and the conversion to image data is notorious for making up data. Think of it like the mp3 of the audio world - it works but there is something wierd at low bitrates. However, if you use a really high bitrate it doesn't sound bad at all. This is true of oversampling in terms of resolution in camera sensors

3) Colour rendition - this is the big one for me. In the hunt for better iso performance, manufacturers have been compromising on colour filter arrays. Making them more transparent lets more light in but also lets crossover colour in leading to the debayer having to do more to get the colour back. The Sony A900 is one of the only DSLR's that gets the colours as good as the best MF backs (it beats many, such as the P45)

4) Underexposure - MF backs underexpose by quite a bit by default. This typically helps in the smoothness of tones as you go toward mid-tones to highlights - especially for people who regularly expose to the right.

5) Sensors optimised for low ISO's - possibly this helps when making compromises in chip design?

You add these up and you can see how LF results always had a relaxed and rounded look - MF came close, especially with classically designed lenses and on 35mm you really had to use classic lens designs  on rangefinders to get the same effect. (even many modern optics are tweaks of classic designs)

Tim

p.s. I did a test recently comparing IQ180 and various other cameras including a Sony A900

http://static.timparkin.co.uk/static/tmp/cameratest-2/800px.html

If you look at the top comparison you can compare the IQ180 with the A900 and you'll see that there is very little difference. I personally prefer the A900 colour rendering but they're very close. We used a Pentacon 50mm f/1.8 lens on the Sony A900 which is a good old lens design.

also I have a feeling that the floating optic used in modern auto focus lenses adds significant extra design constraint to lenses - I haven't tested this though so it's just a thought at the moment.
« Last Edit: May 29, 2012, 12:26:10 pm by timparkin »
Logged

fike

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1413
  • Hiker Photographer
    • trailpixie.net
Re: Hyper-reality for dummies
« Reply #22 on: May 29, 2012, 01:12:43 pm »

BernardLanguillier. Nice work.  How many? X x Y x Z?
« Last Edit: May 29, 2012, 02:05:12 pm by fike »
Logged
Fike, Trailpixie, or Marc Shaffer

amolitor

  • Guest
Re: Hyper-reality for dummies
« Reply #23 on: May 29, 2012, 01:53:28 pm »

Wow, I hadn't seen that article on hyper-reality before.

It's hilarious that he starts out with the Audiophile Story and the Wine Story, both of which are completely debunked at this point, and then proceeds to compare a terrible photograph which happens to be shot with one camera to a less terrible photograph shot with another one, and finally draws some sort of conclusions.

This is exactly the sort of nonsense audiophiles and oenophiles use all the time: If the test is blinded, it's rigged, and if it's not rigged, it's not blind.

Tests that are both blind and fair instantly debunk most of what people say about Audiophile Equipment, and Fine Wines (not everything, just most of it). The test of MF versus whatever the other one is, is rigged.

Note that demonstrating a test to be rigged does NOT disprove the result. It merely proves that the test is meaningless. The result may or may not be true.
Logged

BernardLanguillier

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 13983
    • http://www.flickr.com/photos/bernardlanguillier/sets/
Re: Hyper-reality for dummies
« Reply #24 on: May 29, 2012, 11:48:08 pm »

BernardLanguillier. Nice work.  How many? X x Y x Z?

I'd have to checked, probably 3 focus points per pano position, 5x3 images, so a total of around 40 images?

This was shot with a Zeiss 50mm f2.0, so DoF was plenty.

There is also a bit of HDR masking for some of the rocks on the upper left.

Cheers,
Bernard

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: Hyper-reality for dummies
« Reply #25 on: May 30, 2012, 01:37:43 am »

Hi Tim,

Much appreciate your comment on the issue. I'm of course familiar with your test.

I would expect that a larger sensor always helps, but reducing exposure would loose much of the advantage. Regarding the lenses I assume that they are not always good enough. Regarding the Leica lenses, I got the impression that new Leica lenses are quite aggressively optimized, any comments on that?


Best regards
Erik


The article was almost complete bollocks - yes there is a difference between MF and 35mm DSLR but it's not this 'wine tasting expert sense' thing he's talking about. Here's a list of differences that for me are important

1) Mostly better lenses - however good primes on 35mm SLR can do very well indeed. The simpler the lens design, typically the more rounded the result. Older lenses were designed by trial and error, with designers prioritising 'rendering' as well as resolution. Modern SLR lenses tend to be computer designed. Hence why leica lenses tend to have 'that look'. I have a slight problem with MF wide lenses as the need for strong retrofocus design makes for more complicated lens designs - the results aren't bad but they don't have the relaxed look of longer lenses or LF wide lenses such as the Super Angulons or MF Biogons.

2) Oversampling - not to be ruled out. bayer arrays and the conversion to image data is notorious for making up data. Think of it like the mp3 of the audio world - it works but there is something wierd at low bitrates. However, if you use a really high bitrate it doesn't sound bad at all. This is true of oversampling in terms of resolution in camera sensors

3) Colour rendition - this is the big one for me. In the hunt for better iso performance, manufacturers have been compromising on colour filter arrays. Making them more transparent lets more light in but also lets crossover colour in leading to the debayer having to do more to get the colour back. The Sony A900 is one of the only DSLR's that gets the colours as good as the best MF backs (it beats many, such as the P45)

4) Underexposure - MF backs underexpose by quite a bit by default. This typically helps in the smoothness of tones as you go toward mid-tones to highlights - especially for people who regularly expose to the right.

5) Sensors optimised for low ISO's - possibly this helps when making compromises in chip design?

You add these up and you can see how LF results always had a relaxed and rounded look - MF came close, especially with classically designed lenses and on 35mm you really had to use classic lens designs  on rangefinders to get the same effect. (even many modern optics are tweaks of classic designs)

Tim

p.s. I did a test recently comparing IQ180 and various other cameras including a Sony A900

http://static.timparkin.co.uk/static/tmp/cameratest-2/800px.html

If you look at the top comparison you can compare the IQ180 with the A900 and you'll see that there is very little difference. I personally prefer the A900 colour rendering but they're very close. We used a Pentacon 50mm f/1.8 lens on the Sony A900 which is a good old lens design.

also I have a feeling that the floating optic used in modern auto focus lenses adds significant extra design constraint to lenses - I haven't tested this though so it's just a thought at the moment.
Logged
Erik Kaffehr
 

torger

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3267
Re: Hyper-reality for dummies
« Reply #26 on: May 30, 2012, 04:11:40 am »

Thanks for the very interesting replies.

One aspect which may be important is in short depth of field photography. With bokeh it seems like to get the most pleasing you should not have a lens that is too much corrected to get optimal sharpness.

MF lenses at least traditionally like RZ system had simpler designs and not very large max apertures, but due to the larger film area the DOF could be very short anyway. In other words you can get short DOF with less corrected lenses and thus more pleasing bokeh.

The larger magnification factor maybe also have some effect on in-focus to out-of-focus transitions.

However, perhaps much of that aspect has been lost with the smaller 645 format and more modern digital lenses?

Concerning color rendition I am a bit skeptical, but also curious about it. The difference in light loss between different color filters are afaik quite small, like 1/3 stop or so, if correct it seems unlikely that one would sacrifice color just to gain 1/3 stop. Also, to get accurate color you may actually need quite much color overlap since the eye works like that, not sure though. Dxomark that actually does some sort of measurement on color accuracy show very small differences between cameras, but it is hard for me as a non-expert to interpret that.

I see it likely though that the MFDB manufacturers put quite some effort into color profiling in their raw converters (capture one, phocus) to get natural color like professionals need, while Nikon/Canon etc may be more focused on getting some sort of pleasing look that might not be as natural.

What makes me most skeptical about the color rendition issue is that all the talk about it seem very similar to the "MF has better dynamic range" talk, which people took for granted and saw huge difference (some said six stops!) but now when we actually compare side-by-side with a modern sony exmor sensor the MF DR is actually a bit inferior.

Concerning the oversampling aspect, I've noted that demosaicing false color artifacts is (unsurprisingly) more of a problem if the sensor don't have an AA-filter, and well MF sensors don't have that...
Logged

timparkin

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 66
Re: Hyper-reality for dummies
« Reply #27 on: May 30, 2012, 05:50:18 am »

Concerning color rendition I am a bit skeptical, but also curious about it. The difference in light loss between different color filters are afaik quite small, like 1/3 stop or so, if correct it seems unlikely that one would sacrifice color just to gain 1/3 stop. Also, to get accurate color you may actually need quite much color overlap since the eye works like that, not sure though. Dxomark that actually does some sort of measurement on color accuracy show very small differences between cameras, but it is hard for me as a non-expert to interpret that.

I see it likely though that the MFDB manufacturers put quite some effort into color profiling in their raw converters (capture one, phocus) to get natural color like professionals need, while Nikon/Canon etc may be more focused on getting some sort of pleasing look that might not be as natural.

What makes me most skeptical about the color rendition issue is that all the talk about it seem very similar to the "MF has better dynamic range" talk, which people took for granted and saw huge difference (some said six stops!) but now when we actually compare side-by-side with a modern sony exmor sensor the MF DR is actually a bit inferior.


The best example of the colour differences is in the test shown. The biggest difference is between the Phase P45 and the A900 - try opening these in photoshop and making one look like the other.

The differences seem to be to do with metameric effects of the combination of light source, colour filter and sensor response.

Also I may be wrong but a very strong red colour filter has a filter factor of about 8 or 3 stops of light. Supposedly you can get away with a lot less and so gain about a stop or more of light advantage.

Tim
Logged

torger

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3267
Re: Hyper-reality for dummies
« Reply #28 on: May 30, 2012, 06:20:19 am »

The best example of the colour differences is in the test shown. The biggest difference is between the Phase P45 and the A900 - try opening these in photoshop and making one look like the other.

The differences seem to be to do with metameric effects of the combination of light source, colour filter and sensor response.

Also I may be wrong but a very strong red colour filter has a filter factor of about 8 or 3 stops of light. Supposedly you can get away with a lot less and so gain about a stop or more of light advantage.

Thanks for the reply. Actually looking at dxomark again, the A900 is one that have significantly higher metamerism index than the other cameras, also the IQ180. I guess it has the same sensor as the D3x, which indeed has a lower metamerism index but better dynamic range, so there's definitely something to the color filters. Dxomark cannot present any indications that MF sensors would be better than the best DSLR sensors though, rather that they have nothing special concerning color response. So maybe it is about color profiling too, or some factor that doesn't show up in measurements.

Looking at the P45 test image of yours, it is quite easy to get about the same look for everything except the grass plains in the middle of the image, so there's definitely non-linearity going on. Colors that haven't been registered cannot be tweaked. Not having seen the scene with my own eyes I cannot know which one is more correct though, but the A900 rendering look more believable.
« Last Edit: May 30, 2012, 06:41:56 am by torger »
Logged
Pages: 1 [2]   Go Up