PDA

View Full Version : When camera manufacturers lie? When ISO is not what you expect.



Kym
28-10-2010, 7:02pm
http://www.luminous-landscape.com/essays/an_open_letter_to_the_major_camera_manufacturers.shtml

A very interesting article about how cameras increase ISO without telling you.

old dog
28-10-2010, 7:45pm
that`s quite interesting Kym. Wonder what will eventuate with some more homework on this.

OzzieTraveller
28-10-2010, 9:45pm
G'day Kym

Quite extraordinary - never appreciated this before, but I can see that it could make a lotta sense

Regards, Phil

Watchamacallit
29-10-2010, 10:52am
That is actually very interesting, from a variety of stand points.

arthurking83
29-10-2010, 12:27pm
I don't really understand this 'beat up'? Sounds like a storm in a tea-cup in reality as digital camera ISO values are not to be considered in the same vein as film ISO values have been.

there's a pretty good explanation of ISO values on the Wiki, that explains ISO values pretty clearly.

Basically, most photography folks still think of ISO sensitivit ratings as comparable to film ISO values, and the ISO system is pretty strict and well defined in the film world. It has to be!! it's cross platform comparable and for the sake of consistency to the user, this makes sense. That is load a roll of ISO100 Agfa film into your Nikon body and it will work the same(or very similarly depending on the lenses used) to the same roll of film in a Canon body, with Canon lenses, or a 8x10 field camera. The point of the ISO rating is that the film has to be within a defined specification or the user will get inconsistent results!

This is easy to understand.. and makes perfectly good sense! :th3:

Digital cameras(for the most part.. not taking into consideration digital backs and suchlike).. we're referring to the smaller format, inseparable digital camera bodies here.
The requirement for an ISO rating on the part of the manufacturer is entirely at the discretion of the manufacturer. The standard is not as strict as the film ISO ratings. Canon can label the relative ISO rating of their sensor as randomly as they want(to within very loose tolerances).
Too many people confuse ISO ratings on digital cameras as having a direct relationship to film ISO. While they may be comparable or similar.. they not the same, unless the manufacturer wants them to be!!

The ISO standards are numbered for each set of specification requirements. That is film ISO standards can be ISO 5800:1987 for colour neg film, ISO 2240:2003 colour reversal film, and ISO 6:1993 for B&W.

Digital sensor ISO standards have to conform to ISO 12232:2006.. and the major point in this particular satndard is that the manufacturer can basically rate the sensitivity as they see fit.. not compared to any of the film ISO standards.
(note: for this exact reason, I found myself having to argue a point made by an ex member on the need for using an external light meter for digital photography a while back.. They're nto going to be as accurate on digital as they will with film, unless the manufacturer of the external light meter calibrates the device to each and every digital camera ever made, as seen in this LL article and known by any tech head.. digital cameras(or more specifically the sensors all basically have some slight difference in light sensitivity when compared to each other and more importantly film!

With digital sensor ISO ratings(in fixed digital cameras) all that matters is the relevance it has to itself. that is, is ISO200 twice as sensitive as ISO100 is.
The difference between a Canon 5D's interpretation of ISO200 and Nikon's D70s idea of ISO200 is meaningless and useless... how the manufacturer then tweaks the sensor and electronics to determine an exposure is up to the manufacturer only.. they designed the electronics to provide an image.

LL's parting comment that the manufacturers need to explain anything is a cheap shot(I think looking for some kind of controversy for whatever purpose). They don't need to explain anything! :confused013 You either like the camera and what it does, or you don't. That's what camera reviews are all about. The inner electronic workings are a side distraction, and if they have an adverse impact on the resultant image, then this should be noted in the review.

Also!.... wide open usage of a fast aperture lens is always going to have some impact on exposure. fast aperture lenses are notorious for light falloff!! I don't get their point. Are they implying that light falloff has suddenly become an issue, even tho the same condition has impacted on film for many years before digital cameras came into existence! They haven't specified in their 'Open Letter' where this light fall off occurs within the frame! :confused013 We know that th ecorner of the frame can be affected as much as 2-3Ev, so are they rating the entire frame or only a central portion, or what? We also know that some cameras are setup to counter this effect with vignetting compensation ability.
I don't know much about Canons cameras, but I know that the D700 and D3 series have variable settings for this. Did they switch this feature off?.. leave it on? some software compensates automagically.. it may depend on the camera used, lens used... etc.. etc.

this point that they're trying to make here doesn't[ seem to make any sense(at least to me). I really don't get what they're trying to achieve here.. parity with some film standard? if so., which one?
Digital sensors are capable of re producing any one of the three film types listed above.. as well as others such as UV and IR. Are camera manufacturers supposed to rate each camera to be comparable with each and every one of those ISO standards? or just colour reversal.. or maybe just B&W film.
it's a silly notion! makes no sense to me, and is not really worth the effort in debating.

Granted it'd be useful to have a comparative database of how the ISO values relate to other ISO standards, but to turn this into some kind of controversy is just plain silly!
(or are they trying to load a sensor from a D3s into an F4 or something? :p)

just to give you an idea on how open the ISO standard is for digital cameras:

The Recommended Exposure Index (REI) technique, new in the 2006 version of the standard, allows the manufacturer to specify a camera model’s EI choices arbitrarily.
Quoted form the Wiki.
Pretty simple huh? they basically only have to do what they want with the figures, and the need to be comparable with any film rating is purely up to the manufacturer.
Note also! that the ISO rating is only relevant to the jpg file in most cases(of determination) and in one other method of reporting the ISO rating(of the five available!!) the need is to rate a tiff file. raw images are not a part of the issue. If DxO are using raw images for testing purposes, then shame on them for stuffing the test procedure up!(I normally have lots of faith in DxO Labs tests)
All files for digital camera ISO ratings only need to be in sRGB. If there are any exposure differences between aRGB and sRGB in camera, then stiff! The (ultimately)lower quality sRGB is the file to rate ISO values upon.

this is a tech spec that th eend user really shouldn't need to concern themselves with.
The only concern with respect to ISO values on any digital camera that the user should have(this is what I have anyhow!) is how it performs relative to other values on that camera.
ie. is it noisy? is the noisiness within acceptable limits? is upping the value going to maintain the same Ev for the scene, if I compensate with other variables.
Forget light meters(flash meters may be another story tho.. never used them). Best light meter for your camera is the one built into the camera. Best ISO value to use is determined by you the operator, who has tested over a series of different conditions what works well for a given condition. How accurate that value relative to other values, or how loose the interpretation that your cameras manufacturer decides to use.. makes no difference.

T ratings in lenses is more of an issue than any 'surreptitious' meddling by the camera manufacturers with twiddling of ISO in camera.
Why does my Nikon lens underexpose compared to my Tamron lens, and other Nikon lenses?
(answer!!.. because it's a cheap consumer kit lens, and it does so in manual exposure mode. Set it to any of the auto exposure modes, and if exposure is not the same, the metering system is then to blame instead.)

anyhow.. I've written too much(and Kiwi is going probably having a fit over the length of my reply :D)

JM Tran
29-10-2010, 12:41pm
anyhow.. I've written too much(and Kiwi is going probably having a fit over the length of my reply )

yes, please narrow down to summaries or dot points, hate reading long winded essays. I think you might have also missed the point too.

I agree in some ways with this article, as I have seen this exhibited by the old 5D, which was notorious for overexposing at unecessary times.

DxO has always been highly regarded for its scientific results regarding photography. But I am not a pixel peeper nor do I care much, I'll just keep on trying to make beautiful photos:)

wolffman
29-10-2010, 2:34pm
Not sure its important at all, and certainly not to most of my photography. The way I read that last graph is that you can expect an increase in effective ISO when shooting wide open by 1/2 a stop, but at f4 there is no effect to speak of.
If I am shooting moving things in the dark and wide open I will be at high ISO anyway and stopping the movement is more important than an extra bit of noise, when the photo is likely to be noisy anyway.
If I am shooting wide open for DOF effect, then I am normally down at around 400 so if the camera takes it to 640 at worst, then thats no big deal.
I agree with Arthur - storm in a teacup, if you don't like a camera's electronic innards works then go back to film and scan it yourself.

Tannin
30-10-2010, 2:01am
Over at DPR, where there are many ultra-geeks and at least some of them are very, very knowledgeable indeed, the consensus is that the article is a beat-up of something quite unimportant at best, and downright loopy at worst.

A few examples:


Sigh, not another DxO garbage measurement (measurement overkill of things nobody really cares about) coupled with an incompetent interpretation of the data presented. The guy interpreting the data is quite off his trolley blaming the CMOS sensors and coming to the conclusion that CCD sensors are so much better - the f/1.4 graph clearly shows that the worst offenders are all Nikon CCD based cameras.


I ... believe that the explanation of the data that the author complied is a bit misleading in terms of the light gathering capabilities of a larger aperture. The plain fact of the matter is that for any given camera, a lens with a larger aperture will result in MORE light hitting the sensor and a narrower DOF. If you look at many different cameras and sensors (as the author did) and compare their performance at a given large aperture, then sure, the results will vary. This is hardly news. For me, using a 40D only, a 35mm/1.4 lens is significantly different in terms of light gathering and DOF than a 35mm/2 lens. If I need that difference for a photo, then the 1.4 lens is certainly not a waste of money!


lol

This is what happens when people read data instead of looking at photos because anybody, and I mean anybody, even my 5 year old kid (I just asked her) can see that the there's a shallower DOF and more blur from fast lenses.

This guy is an idiot


the 'findings' of DxO are rather suspect in that what they measure is the output from the camera in the form of a RAW file, which is a FAR cry from the true RAW sensor output (something that no-one other than the manufacturers of the cameras and POSSIBLY a very few engineers with a LOT of time on their hands) can get their hot little hands on. DxO simply ignores that ALL cameras do some processing (signal interpetation) before writing the values to a 'RAW file'.


It were a serious issue if
a) DxO would demonstrate the issue on real life images (which they don't)

b) DxO would document the method of measurement (which they don't - so nobody can check their methodology - very shoddy in terms of scientific merit)

... they only needed another publicity stunt to keep their sales of their software up and pulled that white rabbit out of their hat of many irrelevant measurements...

A poster called jenbenn goes pretty close to expressing my own view when he writes the following (I have added emphasis to underline what I see as his main points):


Sorry, but this article is a joke. The author finds half a stop differnce of the "real" iso vs. the iso shown by the camera at f/1.2. At f/1.4 its a quarter of a stop. Seriously, I fail to see the issue. At f/2 the whole phenomenon dispappears. Where is the practical relevancy, please? An f/1.2 lens still has an advantage over an f/1.8 lens. .....

Furthermore his article is based on DXo measurements ... their results are the least correct of all testing sites. For example, they claim that the corner/edge resolution of the Canon 16-35mm lens decreases as you stop the lens down. Not by a little but by a lot. Sorry ,but the lens is cetainly sharper at f/8 than it is at 2.8. If you dont believe it, just shoot it.

Also: have you ever compared differnet DSLRs by looking at big prints from raw files? You'll find that their sensor rankings are often upside down. Really, as good as DXO's lens correction tools are, their sensor and lens testing site is a bad joke, at least from a practical point of view.


Nevertheless, an interesting topic and one to watch for further developments on.

arthurking83
30-10-2010, 2:49am
I did a few quick tests using my 50/1.2 on the D300, and there is a very slight increase in exposure at f/1.2 and f/1.4. Their finding may be real, but the premise is a beatup. The inaccuracy could be due to more than simply ISO increases, such as not entirely accurate aperture reporting at both f/1.2 or f/1.4, or inaccuracy in the mechanical links at those apertures.
The difference between f/1.2 or f/1.4 compared to f/4 was approximately a 1/3rd unexplained increase in exposure at both f/1.2 and f/1.4.

But then I played with the Sigma 50/1.4 lens, which has a much better light falloff performance level(ie. less of it), and is CPU'ed(where the 50/1.2 is full manual) and the difference in matrix metering compared to spot metering using the centre spot meter was nearly 1Ev in exposure anyhow.

* I'd be damned if I could see any noise performance decrease in the f/1.2 and f/1.4 exposures if Nikon are jimmying the ISO values.
* That 1/3 Ev increase is barely noticable even side by side. Yeah it's there, but you'd be pushing the envelope to say it's an issue.
* differences, inaccuracies, and errors in metering ability(as most folks use auto metering and all such modern conveniences anyhow) will make a much bigger difference to exposure than this (non) issue is making!

LOL! on their CCD vs CMOS findings .... :lol:

if there was a real correlation in sensor type, it'd have to be across the board. D40x D60 D80 and D200 all share the same sensor(maybe a tweak or two difference here and there, but for all intents and purposes the same sensor). D40x, 60 and 80 all fare pretty badly according to DxO, whereas the older tech D200 fares pretty well by comparison.. and the newer tech CMOS D300/D90 all manage about the same as the D200.
Looking at the DxO graph displayed on LL(and me focusing squarely at the Nikon data) the only direct correlation between this (non)issue and sensor performance is directly related to the market level of the camera. Where the D200/300/300s(and D90) and higher level models all fare less badly than the lower consumer level bodies such as D70/80/50/60 and 40's(the anomaly here is the D90). The conclusion in Nikon circles is that the camera itself( ie. the inner mechanics) are probably less accurate on the cheaper models(makes more sense to me!!)
D70 series(and D40 too) were both notorious and advantaged by their cheaper shutter mechanisms when compared to the same sensor'ed D100. The shutter system was once rated to operate at only about 1/90s mechancially(makes it a lot cheaper!!) but with the electronically gaited sensor still had the ability to operate at up to 1/4000s(D100 was good up to 1/8000s), but on the plus side, the D70 was capable of 1/500s sync speeds.
The main point here is that this could easily be a lot more than simple ISO boosting, if the eact accuracy level of the aperture is not known as well(ie. probably assumed to be accurate by DxO)

I can show this anomaly easily using my Tammy 300/2.8 which is also full manual, and has full and half click stops using the diaphragm. I can shoot at f/4 at anywhere between 1/200 - 1/320 and still get the same exposure due to the fact that the half stops are not reported correctly by the camera.

Where they saw an exposure difference of more than 0.4Ev on their D300, I can't get anywhere near that with my D300.. maxing out at 0.3(I suspect 0.33). If I set exp compensation to -0.4 on the brighter image(the faux f/4 image) it becomes too dark by comparison either of the f/1.2 or f/1.4 images.

Adjusting WB to an appropriate value can have a much more profound impact on exposure than this(non)issue.

Oh! and welcome back Tony.. it seems that for once I seem to agree with your assessment of DxO(which I normally find informative).... Sheesh!.... I must be losing my mind.