another reason to toss DoF calculations

[adan]

Advertisement (gone after registration)

Well, I can understand the "artificially increased ISO" - in digital, "ISO" is just amplification of the image. What Dubuvoy is suggesting is not that much different than what Leica does to remove vignetting from M9 images.

 

The camera detects a "troublesome" lens - in Leica's case (via 6-bit coding), perhaps a 21mm, in Canon's case (via the lens CPU), an f/1.2 lens. Leica raises the amplification or gain in the corners - Canon raises the gain across the whole image (and maybe also in the corners), in both cases to correct for "missing" light that got lost due to sensor architecture.

 

It is an "ISO" increase like any other digital ISO increase or "push" - as Michael says.

[SJP]

Well..... yes. But non-coded lenses also work fine on Leica M8/9 and the camera does not seem to know what aperture is being used (they tossed that feature rather quickly) so how could they design for a sensible correction? (Answer = they (Leica) don't)

 

It is worth pointing out that DoF is not really related to "out of focus" it is just stating that for "normal" viewing you can live with a certain amount of slack. Focus or non-focus is related to Airy disks & Fourier optics etc. and then the criterion for "sharp" is completely different. The DoF is governed by the circle of confusion of ca. 31 micron on a FF camera, diffraction limited sharpness related to Airy disks of ca. 1-2 micron if you use a large aperture. These are completely different demands & length scales.

[adan]

"non-coded lenses also work fine on Leica M8/9" - hmm, ever tried a 21 Super-Angulon on an M9? Or compared results directly with coded/uncoded wide lenses?

 

Non-coded lenses often "work" - they also often "work" significantly better when coded, especially those wider than 35mm.

 

In fact I get estimates of shooting aperture showing up in Adobe Bridge EXIF from both my M9 and older M8 shots. They vary from reality by 0-2 stops - depending on a host of other factors - but that does not mean they are non-existent. I.E. my camera never reports I shot @ f/11 if I really shot at f/2 - or vice-versa.

 

Leica's corrections are most likely therefore also not precise, but that does not mean they are not "sensible." "Stopped down a lot" or "close to wide-open" are sensible enough for most vignetting corrections.

 

Just sayin'

 

As to DoF - no doubt you are right.

[pgk]

Leica's corrections are most likely therefore also not precise, but that does not mean they are not "sensible." "Stopped down a lot" or "close to wide-open" are sensible enough for most vignetting corrections.

I think that you are there Adan! My summation is that whilst results can be very acceptable with uncoded lenses, the RAW files are 'optimised' for 6-bit coded lenses and can look significantly better. With other manufacturers lenses we can't determine what effect their 'coding equivalent' actually has pictorially (unless part of the lenses' cpu could be disabled perhaps), however it doesn't look as if it will be extreme if we assume that differences will be similar to those from Leica.

 

So as you say, somewhat imprecise adjustments are probably more than adequate although in the case of cpu lenses these may include vignetting, light-loss factors and more, and these may well be applied variably over the aperture/focus ranges (I'd be surprised if they weren't) so they may be actually relatively complex? Whether. as the article suggests, this is disingenuous, is something else - personally I would say not.

 

However, I would suggest that one potentially worrying aspect of the article is that it might well be possible to adjust for some deficiencies in lens design by adjustment of the raw file (I adjust for chroma in photoshop as a matter of course for underwater shots taken through a flat glass port - this is a variable somewhat dependent on aperture and actual distance of subject, and doing so reduces distracting colour fringing on high contrast edges at the frame corners). Whilst I can see advantages in optimising an excellent lens, I would be disconcerted to think that poor designs could be made better by such adjustments although this may well be going on already.

[mjh]

But I am still puzzled by the suggestion (easy enough to check as Adan says) that faster aperture lenses show an increased depth of field wide open - although I can see how this might be assumed.

The light loss can be explained by the three-dimensional structure of sensor pixels imposing an effective aperture limit: there is a maximum diameter of the exit pupil that a photosite can actually “see”, so an even larger exit pupil doesn’t result in the photosites capturing more light. Up to this point Mark Dubovoy’s explanation made perfect sense, only then he was jumping to the conclusion that if a photosite only “saw” an effectively smaller aperture (exit pupil actually), the the depth of field would correspond to that smaller aperture. But depth of field depends on the sizees of circles of confusion projected onto the sensor, and if each sensor pixel did capture less light, this would render a circle of confusion dimmer – but not smaller. So the depth of field should stay the same.

[pgk]

Thanks Michael (interesting article too!).

 

So from your last answer, is it possible to calculate this limit in terms of maximum exit pupil for any given sensor or does this require 'insider' knowledge of the sensor's design? If so can an effective limit on maximum useful aperture be sorted out for each sensor/focal length. Or are there other complications which creep in?

[dpattinson]

Advertisement (gone after registration)

Seems to imply that the CCD chip allows better DOF subject isolation due to loss of light in CMOS chips.

 

An Open Letter To The Major Camera Manufacturers

[pico]

It makes no sense. DOF does not change with increased ISO and no change to the aperture.

[Nicoleica]

David, I'm not going to participate in an argument. But I think that your comment only goes to show that you can twist statistics any way that you like to get the answer that you want. The document referred to is bad enough in it's selective twisting of data, but to then take selected parts of that information and to then claim that it implies that CCD gives you more DOF control is, in my opinion, quite frankly astounding.

I will apologise in advance if you take offence at my comment above, but I am in mild shock at these conclusions and comments.

[mjh]

Seems to imply that the CCD chip allows better DOF subject isolation due to loss of light in CMOS chips.

And I thought everything had been said when we had the same discussion just four weeks ago: http://www.l-camera-forum.com/leica-forum/digital-forum/148146-another-reason-toss-dof-calculations.html.

 

And no, it doesn’t imply that at all.

[KevinA]

I may be misinterpreting what the article is saying. But here goes, I think he is implying that the wells are giving a pinhole effect, so some of the defocussed light is hitting the well sides and not recording, he implies that because of that there is less bokeh in the defocused areas as more light is scattered thinly across more sites at a more acute angle so a bigger proportion is missing the sensor. If that was the case would not the unfocused areas become darker and progressively so the more unsharp they are?

 

Kevin.

[sandymc]

I may be misinterpreting what the article is saying. But here goes, I think he is implying that the wells are giving a pinhole effect, so some of the defocussed light is hitting the well sides and not recording, he implies that because of that there is less bokeh in the defocused areas as more light is scattered thinly across more sites at a more acute angle so a bigger proportion is missing the sensor. If that was the case would not the unfocused areas become darker and progressively so the more unsharp they are?

 

The problem is that the original article is badly written to the point that it might be implying pretty much anything at all. But speculating about what it might mean is pointless, because, as became clear when the author responded to some queries, the data is basically meaningless. Sorry to be pointed, but the article would best just be forgotten.

 

Sandy

[jaapv]

I am very confused. For one thing the term increasing the ISO on the sensor is basically meaningless. A sensor, reading the ISO norm, has no ISO value as such, it has an ISO equivalent and is supposed to respond comparibly to a certain ISO-rated film. So the only thing the camera manufacturers are doing is adjust the gain to get the proper ISO equivalent response from the sensor. And what is wrong with that?

[KevinA]

The problem is that the original article is badly written to the point that it might be implying pretty much anything at all. But speculating about what it might mean is pointless, because, as became clear when the author responded to some queries, the data is basically meaningless. Sorry to be pointed, but the article would best just be forgotten.

 

Sandy

 

Yes, Sandy, the article has cropped up on a few forums and I think universally condemned. I was trying to point out what I thought was a flaw in the agument. It's a shame a site like Lunatic mindscapes got so far as to publish it.

I thought a conclusion had been eagerly jumped on that made little sense. A bit like the physio's my Dr. would send me to rather than put me on a waiting list to see a specialist, I eventually lost my patiences when they diagnosed the pain in my knee was because I walked with a limp. And no I am not making that up.

 

Kevin.

[ho_co]

... I eventually lost my patiences when they diagnosed the pain in my knee was because I walked with a limp. And no I am not making that up...

So that's the kind of thing they spend all that time in school to learn!

 

And the ones that flunk out get to work at DxO?

[ho_co]

For those still game on this topic, Mark has added an update to the LuLa Open Letter To The Major Camera Manufacturers.

[Bo_Lorentzen]

As Stephen said, lenses at the same setting should give about the same exposure.

 

The suggested ISO boosting "trick" may work on a dslr which read the f.stop - however it would be meaningless on a M camera as the camera do not read the f.stop.

 

if a picture is exposed properly with a summilux at 1/125 & f/2 then if their article makes sense re. the sensitivity the image should expose wrong at 1/250 f/1.4 - similar can be calculated for a Noctilux (i.e. 1/500 f/1).

 

Aside from that I'm entirely with Howard, in that I find the dxo data at best suspect, personally I find the data to be on the border to bogus. the only real useful aspect of that data is thousands of camera users can spend hours comparing numbers to numbers, much like when we would play with "exotic car trading cards" as kids, everybody knows a V12 beat a V8 yeah.?

 

.