Talk:Film speed/Archive 1

Graininess
Hi, all, This is too much to just change in the main entry. May I have some comments so I can fix the proposed text up? My own comments are in [bracketed italics] 2B removed later. (I know there is an  !-- Invisible Note to editors --  mechanism, but these comments should be seen on the TalkPage.  I'll hassle getting them out later.)

PROPOSED TEXT

Individual silver halide crystals suspended in a film emulsion aggregate into clumps ("grains") of random size. [Grains are not as regular and orderly as the crystals which lie at their heart, and at the heart of the energy storage mechanism that makes it possible to store and later develop a latent image] Even when these are not individually resolved in an image, a texture ("graininess") remains. Higher film speed brings more graininess. [nice direct statement from original more or less] Fine-grain stock, such as film used for portraiture or copying, is "slow", meaning that the amount of light used to expose it must be high or the shutter must be open longer. Fast films, used for shooting in poor light or for shooting fast motion, produce a grainy image. [Easily grasped contrast of different practical situations, also from original, more or less]

Since each grain of silver halide develops in an all-or-nothing way, photographic images actually consist of a mosaic of developed and undeveloped areas. In this sense, film is a threshold detector rather than a linear detector. [''This is a delightful insight to present, at a time when so many of us are making the historic transition from "analog" to "digital". However, it didn't have enough of a "compare and contrast" discussion to clarify differences between the two ways of conceptualizing film's function.''] If the subject has an edge between light and darkness and that edge falls on a grain, the result will be an area that is all light or all shadow. Fine gradations of grayness are also quantized, so that graininess interferes with the perception of shape from shading as well as with object contours (edges). [''As the brain imposes structure on sensation to achieve perception, "shape from shading" has emerged in recent decades as an important mechanism of shape encoding from surfaces, just as contours and their elaboration in the brain is an important mechanism of shape encoding from edges, and of figure/background segmentation. Also, it makes for beautiful b/w photos :-) ] Photographers sometimes exploit film's quantization with high-contrast derivatives ("orthos") that really are only black or white, [These are typically made with photolithography film like Kodalith, hence the name.  I put this in to be sure readers understand that it's hard to get even grainy shots to lose their grey-ness, unless special steps are taken.  Perhaps it can be dropped to keep everything brief. Opinions?  ] but the issue of quantization is moot in most photos, because grains are random in size, overlap, and are not individually resolved in the image.  Under such typical circumstances, slow films have higher contrast, and faster films have not only lower contrast, but a much longer gray scale, beloved of available light photographers in journalism and movie making.  [There was a problem in the original on this point.  Maybe you can say it better, but we need some kind of clarification. An ASA 400 b/w film has what they call a "longer grey scale" than an ASA 25 film. Subjectively, I've always thought the grays of fast film "looked nicer", but technically it means that the negative film goes deeper into highlights before blocking/saturating, and it can throw a few grains into shadows before falling to baseline density at low image intensity.'']

[''Original: "Fast films are also relatively contrasty, for the same reason. That is, an area of the image will consist of bright areas and dark ones with few transitional areas of midtones." This statement is misleading at best. Slow films have a higher gamma. '']

END proposed text.

Please don't clobber me. If there are better ways to handle a suggested revision, just say. As for the revision itself, entries on technology are of necessity always a work in progress, so this won't be the last change, and not every change has to be made. Thanks! Jerry-va 17:09, 27 February 2006 (UTC)
 * I think that should be added as a separate section and not the intro, due to its detailed nature.

Does anyone know how films speed and the ISO ratings relate to digitalo cameras? I think this would be useful to add.

BEGIN remark Quote: "However, this loss is visible as image noise rather than grain." As someone from the area of digital image processing research I find this a completely non-technical and unsupportable statement. Maybe I am overlooking something. Please define the difference. If analog graininess and digital snow look different because of the square raster, I would understand it. However, from some distance artificial graininess in digital images cannot be discerned from the example given. Note that the figure on the page is also digital, with square pixels.

END remark —Preceding unsigned comment added by 94.212.221.215 (talk) 22:52, 5 August 2010 (UTC)

Delta 3200
A while ago, I removed Delta 3200 from the chart, and today an anonymous user, in what I assume to be a well-meaning edit, added it back (at 3200). The reason that I removed it is that Delta 3200 does not have a true ISO speed of 3200; according to the data sheet, the true ISO speed is 1000. Shooting it at the recommended speed of 3200 is actually "pushing" the film, and will result in higher contrast and less shadow detail than if it were shot at the ISO speed. For this reason, I don't think it's appropriate to list it on the chart. Listing it under 3200 is inaccurate, and listing it under 1000 would be confusing without a lengthy explanation. Similar arguments apply to Kodak's Tmax P3200. -- Coneslayer 18:00, 2005 Apr 17 (UTC)


 * That depends on what we mean by nominal film speed. The nominal speed of a number of popular films is different from the true ISO speed (which already contains a safety factor). Kodak TMZ and Ilford Delta 3200 are marketed and labeled as ISO 3200 films and can be shot as such.  For that reason, it makes sense to list them here and provide more detailed explanation in separate articles about these two films.  Your reasoning is not wrong, but by the same token you could challenge the inclusion of Fuji RVP 50, which has given rise to endless debate about whether it should be rated differently than its labeled speed, with recommendations ranging from ISO 40 to ISO 80. I'd say it's simplest if we stick with what's on the box (yes, that may be marketing hype) and explain what's really going on in separate articles for each film stock. --MarkSweep 20:25, 17 Apr 2005 (UTC)


 * Mark, to put it simply, I would say you're using the word ISO where it doesn't belong. The ISO specifies a method of testing film speed.  The 3200 speed does not come from this testing methodology, and you won't see the word ISO near that speed on the packaging or datasheet.  The ISO speed of Delta 3200 is 1000, period, no matter what you set the dial at, or how long you develop it.  In contrast, RVP 50 is an ISO 50 film.  You, and other photographers, might like the results better when it's shot at 40 or 80, but that doesn't mean the ISO speed is wrong, or that the speeds you like better are the true ISO speed.  It doesn't make sense to say I'm shooting an ISO 50 film at ISO 40--that's a misuse of ISO, because I'm not using the ISO methodologies.  The correct term would be EI 40 (exposure index).  If you want to list Delta 3200 under ISO 1000 on this page, that's fine with me, but I would strongly object to listing any film under a speed other than the ISO speed. -- Coneslayer 21:23, 2005 Apr 17 (UTC)


 * Good point. I don't disagree with your assessment of the facts. My only concern was that it would be better to explain this in the article, since these issues are generally relevant here and likely to come up again. So how about listing Delta 3200 under ISO 1000 in the table and adding a note explaining what's going on? --MarkSweep 23:33, 17 April 2005 (UTC)


 * I have added a paragraph on this topic to the article. -- Coneslayer 21:35, 2005 Apr 17 (UTC)

f-number and aperture diameter
I have edited the article slightly to remove the common misconception that the f-number is equal to the ratio of the focal length to the aperture diameter. It is not. The f-number is equal to the ratio of the focal length to the diameter of the entrance pupil of the lens. The latter is proportional to the diameter of the aperture. It is true that doubling the f-number halves the diameter of the aperture but it is not true that the aperture diameter is equal to the focal length divided by the f-number.--Srleffler 04:29, 14 February 2006 (UTC)

No mention of ISO 6 etc?
The description of ISO film speed method refers only to that for still colour negative film. Wouldn't it be more thorough to mention the ISO 6 method for B&W film; why there is a difference between the methods for B&W and colour film; why chromogenic B&W film doesn't quite fit with either; and the existence of a separate standard for colour reversal film? It might also be worth mentioning that there is no ISO for motion picture colour negative film. I would be happy to write the aditional material, as I have all the relevant standards.

--Helen Bach 15:50, 31 March 2007 (UTC)


 * Yes, please do. And if you can share copies of the relevant standards (just as fair use to help other editors, you understand), that would be fantastic, too. Dicklyon 16:23, 31 March 2007 (UTC)

Helen, I realize you wrote this 3-1/2 years ago, but I can't imagine why ECN-2 (and equivalent) motion picture films wouldn't have an identical standard to C-41 (and equivalent). My can of 5201 (Kodak Vision 2 50D) does only have EI ratings, but I would assume that is only because that Kodak assumes that Cinematographers will run tests on their filmstocks. This is similar to the practice of B&W photographers coming up with their own EIs for their favorite stocks, such as an EI of 320 for Kodak Tri-X 5063, or Portra 400NC. ISOGuru (talk) —Preceding undated comment added 00:45, 11 November 2010 (UTC).

Fallacy in illustrating digital sensitivity
Images of a flower taken at ISO 100 and ISO 1600 on a Canon 400D digital camera. Both images were shot under similar lighting conditions, varying only the ISO setting and shutter speed.

I can understand how they were shot under similar lighting conditions, but the illustration is incredibly fallacious, as the ISO 1600 shot would be overexposed compared to ISO 100. Shutter speed would need to be 16 times faster to achieve the same exposure, ie. 1/5600. This is a 2/3 stop of a difference, and that's assuming that the 1/4000th setting (fastest on that camera) is exposing correctly. Tests (in magazines) show that even shutters in professional cameras overexpose by as much as a whole stop at their highest speed, the problem is even more pronounced in amateur cameras (such as 400D). This test would be a good candidate to be repeated (possibly with the same camera), at shutter speeds of 1/100th and 1/1600th or slower. —The preceding unsigned comment was added by Special:Contributions/ (talk)

Use of Lux as a means of illustrating sensitivity of a digital video/still camera
I have a Sony HVR-A1J HDV camcorder with a 1/3" CMOS sensor which can be used as a still camera with images saved to a memory stick. The sensitivity of the camera is defined as: 7 lux at +18db with a normal shutter. Let's take this apart: 7 lux is a little less than one foot candle - sounds great, but that 18db is telling us that the noise level is going to be horrendous. Normal shutter probably refers to something between 1/30 and 1/48 second (25 or 30 frames/sec - this camera doesn't actually scan fields). If we move the gain back down to 0 db, sensitivity is more likely to be in the 400 to 1000 lux range. (Well you can do the math to get an exact number. I'm not going to bother.) This actually puts the camera in a fairly standard professional video camera sensitivity range - ISO 160 (or using our old friend Ektachrome 160 in the Bolex), where you can push the film in processing or turn up the gain as needed.

What would be muchly appreciated is a Wikipedia entry that incorporates a cross reference table comparing all these different units of sensitivity to each other. Competitive manufacturers product literature rarely uses the same standards so very little makes sense - like mixing RMS and EIA watts! --Mccainre (talk) 07:32, 7 January 2008 (UTC)

Lux (meter candles) is also used to define ISO speed. And, since there are 10.76 candles at a meter (therefore 10.76 meter candles) (lux) as a candle at one foot ( 1 foot candle), that's only 2/3 of a footcandle.

According to a thread I read on Photo.net, the clearest explanation of ISO by the way, I've read thusfar, defines ASA/ISO as "0.8/the speed point in meter candle seconds. The speed poin is many times. 0.1 density units over gross fog. [base fog?]" and, as an example:  "An ASA/ISO 800 film requires an exposure of 0.8/800 = 1/1000 of a meter candle second[. . .]"  This is WITHOUT A LENS. So you'd have to compensate for light loss based on T-- (not F/-) stops. ISO and foot- or meter- candles are related. ISOGuru (talk) 00:56, 11 November 2010 (UTC)

Definition of digital ISO speeds
According to light meter, the ISO film speed S is given by S=Kn^2/Lt, where L is the luminance of a medium-brightness area (candela/m^2), n the aperture number, K a constant (usually between 12 and 14), and t is the shutter speed. Usually, medium brightness is taken to be an 18% reflective card. However, this doesn't say anything about the headroom. Should a sensor with a given ISO number be at exactly 18% of saturation for a given luminance L? Or is there a standard amount of headroom? I've searched a bit and I found numbers of 106%, 170%, and 141% for the saturation level. What does the ISO document say about it? This information is missing. See e.g., ,. Han-Kwang (t) 15:21, 10 October 2007 (UTC)


 * (copied from Talk:Light meter) Solving for S doesn't mean the meter is telling you that that's a definition of S, or a recommendation for S. I mean, yes, given a luminance and a meter calibration, etc., you can back-solve for S.  But there's nothing there that says the L value is taken from a "medium-brightness area", is there?  It's from wherever you aim the meter.  Now, there is actually a saturation-based digital camera ISO speed definition that works something like you say, but it has its own constant, not dependent on any particular meter calibration of course, and leaves about a half-stop of headroom above a diffuse white when average scene is something 18%, if I recall correctly.  But nobody rates their cameras by this definition anyway, so don't expect to find any such relationship in practice relative to any ISO setting or rating on a digital camera.  If you want to get technical ISO 12232 says "The saturation based speed, Ssat, of an electronic still picture camera is defined as:   Ssat = 78 / Hsat, where Hsat  is the minimum focal plane exposure, in lux seconds, that produces the maximum valid (not clipped or bloomed) camera output signal."  This may not be trivial to interpret as you want, but compare this to the noise-based ISO definitions of the form Snoise 40 =  10 / HS/N 40 where that H level is in some sense a "mid gray"; the ratio of saturation to this mid-gray is thus 7.8, which is more than 1/0.18 by a factor 1.404, or about a half stop.  I hope this helps. Dicklyon 18:53, 10 October 2007 (UTC)

(copied from Talk:Light meter) I see, so the equation n^2/t = LS/K is a simply a definition of what the reading that a light meter should produce for a given luminance, even though K is somewhat variable. But as a photographer, of course I would want to know how much headroom I will get if I point the light meter at a 18% card. As you say, officially the saturation level should be 141% (1/2 stop above 100%), but in practice YMMV (e.g. my digital compact camera seems to aim for 18/100 rather than 18/141). Maybe this subject really belongs on film speed, where I also asked this question. Anyway, I think the focal plane exposure is given by H=pi L t/4 n^2 (for objects far away from the lens), so Ssat = 78/Hs = 312 n^2/ (pi*Lsat*t) where Lsat is the luminosity that just saturates the sensor. I guess that 78 comes from 100*pi/4, so Ssat=100 n^2/(Lsat*t), which seems to be an elegant equation -- would it be appropriate to mention on film speed? Han-Kwang (t) 23:43, 10 October 2007 (UTC)


 * From Talk:Light meter: apparently, the saturation-based iso number Ssat is defined as Ssat=100*n^2/(Ls*t), where Ls is the luminosity (cd/m^2) that is at the threshold of sensor saturation for a given n and t. This may be nice to add to the article, but I'm not yet 100% sure that I understood everything correctly. Moreover, are there any statistics on to what extent digital cameras follow this equation? Do they typically inflate the ISO speeds or are they rather more conservative? Han-Kwang (t) 23:54, 10 October 2007 (UTC)


 * Interesting how you found that 100; I had not seen that before. Perhaps you're right; perhaps it's WP:OR.  Anyway, in typical good DSLRs you have a fair amount of headroom at ISO 100, meaning the saturation ISO is lower than that.  For example, the Sigma SD14 had so much headroom that Sigma recently introduced an ISO 50 mode via a firmware update; at ISO 50 it probably has not much headroom; I don't know what the saturation ISO is, but like I said, nobody ever specifies that. Dicklyon 01:18, 11 October 2007 (UTC)


 * Do you actually have the ISO documentation or are you like me dependent on secondary sources? Instead of 100, I could write 312/pi (=99.3), in which it's a straightforward derivation from the Ssat=78/Hsat equation. Would it still qualify as OR then? I read in some policy (can't remember which) that straightforward mathematical derivations are OK. The question is whether this one is as trivial to us as saying that 11 out of 40 is equal to 27.5% to others. :) (The reason that I'm interested in this is that I want to use my camera as a luminosity meter, but it seems that I need to calibrate it against a known reference.) Han-Kwang (t) 07:18, 11 October 2007 (UTC)


 * This is the story behind Ssat = 78/Hsat, according to ISO 12232:


 * The exposure index value (I subEI) is 10/Ha where Ha is the arithmetic mean focal plane exposure. The constant 10 is chosen to be consistent with 2721 and 5763.


 * Hsat is the focal plane exposure that would result from a 141% reflectance object (ie half a stop over a 100% reflectance object). Therefore the focal plane exposure resulting from an 18% reflectance object is Hsat * 18/141. Put that into the EI equation and you get IsubEI = 10*141/18/Hsat = 78/Hsat. The speed is the EI for an 18% grey card, hence Ssat = 78/Hsat Helen Bach 17:30, 11 October 2007 (UTC)

(copied from Talk:Light meter) I've posted something on Talk: film speed. Now that I have read Dicklyon's posts here I see that my post there is a re-write of the end of his previous post - ie that the 78 factor in the Ssat formula comes from 141/18. The formula for focal plane exposure used in 12232 (Eq 2) takes flare exposure (Hf), cos^4 loss, lens transmission factor (T) and vignettng factor (v) into account. The pi/4 combines with the other losses (T v cos^4theta, with Hf << H) to produce 65/100 (which is 0.83 pi / 4), so the formula for arithmetic mean focal plane exposure is 65 La t / (100 A^2) where La is the arithmetic mean scene luminance and A is your N. On axis the focal plane exposure would be 0.88 pi L t / (4 A^2). Helen Bach 22:47, 11 October 2007 (UTC)


 * OK, for off axis there's a factor cos^4(theta). I always thought that was the main cause for wide-angle pictures to be dark in the corners, but if I understand vignetting correctly, there is also a contribution from physical obstruction. It doesn't seem logical to me to assign an absolute value 0.83 to the combination of T, v, and , since these are dependent on the quality of the objective lens (T and v) and the focal length/sensor size (cos^4 theta). I think it would be nice to use the equations as discussed here in the article, although I'm not sure that film speed is the best place. I'm thinking of the general equation, H=T v pi L t cos^4(theta)/ n^2, the saturation-based ISO calculation, plus mentioning some typical values for v, T and averaged cos^4(theta), plus a few examples for the actual values of H for common photographical conditions (e.g. indoor 200 lux = 200/6pi nit (white surfaces), t=1/60, n=3.5, 38 mm lens (relative to a 35 mm sensor size), outdoor 10000 lux. More information about the noise-based ISO would also be nice, and conventions about the ISO numbers as indicated by digital cameras would also be nice. I did some experiments with a couple of compact and DSLR cameras and a calibrated lux meter. Using theta=0, T=1, v=1, and assuming that a saturated jpeg means that the sensor is saturated, and that the jpeg pixel values are with gamma=2.2, I calculated the saturation-based speed Ssat and compared that to the nominal speed as indicated by the camera. I find ratios Ssat/Snom between 1.5 and 3.2; even with the same camera different numbers depending on the nominal ISO speed setting. I really don't get how it is possible that something that is supposed to be a standard can be interpreted so freely by camera manufacturers. Han-Kwang (t) 11:25, 12 October 2007 (UTC)


 * A few comments. 1) The cos^4 falloff does not take physical obstruction into account. Cos^3 may apply to lenses that have tilting exit pupils - typically wide angle designs like the Super Angulon, but that is a subject for another page. 2) It is not correct to assume that a 'saturated' JPEG means that the sensor is saturated. Examining a RAW file will show that the JPEG may be at 255,255, 255 before the sensor is saturated. 3) From my own tests on various cameras I have found that the conversion to a JPEG does not happen at constant gamma, particularly at the shoulder, so there is a danger in assuming that it does. I'm just flagging this as a potential problem with your assumption, not saying that it is universally applicable. 4) ISO 12232 sets conditions that determine whether the noise-based speed or the saturation-based speed is reported as the camera's ISO speed. By the way, it also tells manufacturers to call the EI user control the 'exposure index' or 'exposure setting' control, rather than the 'ISO speed' control. Helen Bach 01:02, 13 October 2007 (UTC)


 * Helen is absolutely right. You can't begin to get at the question of sensor saturation by studying a rendered JPEG out of a camera; the relationship is too distant, and not at all standardized.  Furthermore, the ISO speed or E.I. setting on a camera bears no relationship whatsoever to the ISO speed as specified in the ISO standard; well, maybe SOME relationship, like all the settings you can choose are hopefully between the highest noise-based ISO and the low saturation-based ISO, but even that is not specified or guaranteed.  You'd be better off studying raw data and raw converters if you want to find out where the headroom is, and you won't need any standards for that.  The latest ISO speed standard for digital cameras, ISO 12232:2006 I believe it's called, has added a new hack called "standard output sensitivity" (SOS) that tries to specify a bit about how a photo should be rendered as a function of exposure and ISO settings, but even that has no relationship to the ISO speeds specified in the same document, except that it's measured on the same scale and is a recommended exposure index; it's defined as 10/H for H being the exposure (lux-seconds) corresponding to an output code of 118 out of 256 (in an unspecified but presumably sRGB or gamma 2.2 color space).  The processing gain can of course be anything, so any SOS value is possible, unconstrained by the different definitions of ISO speed rating.  It's really a bit of a complicated mess.  —Preceding unsigned comment added by Dicklyon (talk • contribs) 06:06, 13 October 2007 (UTC)

(unindent) To Helen: 1) OK for the vignetting, but after some thinking I don't see why it is cos^4 rather than cos^2. One cos(theta) comes from the angle of incidence onto the lens, and the other cos(theta) from the angle of incidence onto the focal plane. Derivation:                     D    ,-| b             o         |,-'   +-                   ,-' |  f                ,-' o/cos(theta),-'          ,-' spatial angle dOmega       ,-'   |,-' a | a is a small surface with a luminance L; the surface has an intensity I=L a cos(theta) [candela] and sees the lens (at a distance o>>f) with a spatial angle d&Omega;=D^4 cos(theta) pi/4 o^2 . The flux as captured by the lens is
 * F = I d&Omega;=L a D^2 cos^2(theta)/4 (o/cos(theta))^2 [lumen].

The image in the focal plane has a surface b=af^2/o^2. The illuminance E at the focal plane is then
 * E = F/b = pi L D^2 cos^4(theta) / 4 f^2 [lux]

With D=f/N,
 * E = pi L cos^4(theta) / 4 N^2.

Of course, you can add extra factors for vignetting v(theta) and a lens transmission T, but that doesn't explain the cos^4.

2) and 3) - OK, that's the difference between gamma 2.2 and sRGB, but that's not going to be a big effect at 1 stop below JPG saturation.

4) It looks like (consumer) camera manufacturers are ignoring that part of the ISO standard.

To Dick: aha, 118/256 is 18.4% of saturation in gamma=2.2 and 18.0% in sRGB. (Are you sure it wasn't 18/255 by the way?) And then we are actually back to where I started: SOS rating S (100, 200, 400, 800, etc.) is equivalent to a ISO saturation speed, except that the SOS number refers to saturation of the output JPG image and the ISO number refers to saturation of the sensor.

If one of you can explain the cos^4 factor to me, I think it's time to expand the article.

Han-Kwang (t) 07:16, 13 October 2007 (UTC)


 * cos4 is a common approximation to the physics, but can't be even approximately correct for a super-wide or fisheye lens design. Here are some sources; the Doug Kerr paper is a good one. Dicklyon 07:30, 13 October 2007 (UTC)


 * Aahh, I forgot that the distance to the lens is o/cos(theta) rather than o. Thanks. I don't have time today, but I'll update the article then tomorrow, unless one of you feel like doing that now. I will mention that 0.83 is a typical value for on-axis transmittance. (Seems rather low, aren't AR coatings much better than that?) Han-Kwang (t) 08:29, 13 October 2007 (UTC)


 * 0.83 is not a typical value for transmittance, it is the combination of cos^4, v and T. The individual values used in the derivation of Eq 2 in 12232 are cos^4 theta = 0.94 (theta = 10deg); v=0.98 and T=0.9. Helen Bach 12:23, 13 October 2007 (UTC)

Arithmetic to logarithmic conversion formula different from that defined by ISO
How time flies. I'm still working on the draft I promised. Meanwhile, how about changing the arithmetic - logarithmic conversion formulae to the one implicit in the speed definitions given in ISO 6 and ISO 5800, and stated explicitly in ISO 2240; and the derived inverse? This is the conversion method derived from the two speed definitions in each of ISO 6 and ISO 5800, and stated explicitly in ISO 2240: S° = 1+10 log S This is the inverse of the above, which is not specifically mentioned in the Standards: S = 10^(S°/10) / 1.26 All base 10 logs.

Helen Bach 16:36, 10 October 2007 (UTC) —Preceding unsigned comment added by Helen Bach (talk • contribs) 16:33, 10 October 2007 (UTC)


 * How strange. I thought speed was measured in thirds of stops.  10log10(x) is close, but 3log2(x) would be exact.  What is that particular standard?  Is it still in force? Dicklyon 01:22, 11 October 2007 (UTC)

It is stated explicitly in ISO 2240, and can be arrived at by simple substitution in the speed formulae given in ISO 6 and ISO 5800. Its ancestry goes back further than that, but those are the three current standards. The relationship is based on the definitions of arithmetic and logarithmic speeds, not on a purely mathematical correlation between the integer pairs 100/21, 200/24 etc.Helen Bach 03:59, 11 October 2007 (UTC)


 * Interesting. I thought the numbers were tied to powers of two, and more specifically defined by tables such as those in ISO 12232.  But now that I examine the tables, the break points do follow powers of 10, just like decibels, with 10 one-third-stop steps per decade.  I guess that makes sense as a way to keep the decimal numbers nice.  Thanks for clueing me in. Dicklyon 04:48, 11 October 2007 (UTC)

Will the formulae be changed? What happens next? I'm new to all this. By the way, which table in 12232 are you referring to? Thanks. Helen Bach 22:11, 11 October 2007 (UTC)


 * WP:Be bold; WP:SOFIXIT. Table 1, section 8.2, "Method of Reporting" in the DIS; I presume this made it into the final. Dicklyon 22:26, 11 October 2007 (UTC)

Pencil sharpener images
I removed this:

Reasons: Han-Kwang (t) 00:19, 7 January 2008 (UTC)
 * The images suffer from motion blur - use a tripod for 1/5 s shutter time;
 * The thumbnailing process makes it even harder to compare the images. Crop the photos in the original resolution (i.e. the native resolution of the camera) to e.g. 250x150 size and save as jpg with sufficiently high quality setting.
 * What do you exactly want to illustrate with these images that isn't already illustrated by the existing image (the yellow flower)?


 * The problem with the flower shots was that they had unequal exposures; see Reference desk/Science. -- Coneslayer (talk) 15:10, 7 January 2008 (UTC)


 * (1) The motion blur was deliberate. I needed to show both the longer exposure time and the lower noise on ISO 100, and the motion blur in no way affects the noise.  The noise is caused by the sensor, and will exist even if the photo's subject is blurred.


 * (2) Is this what you mean? --Bowlhover (talk) 01:37, 8 January 2008 (UTC)


 * Not to put too fine a point on it, but the images suck. The inclusion of blur due to camera motion totally complicates the comparison, and it wasn't mentioned in the caption.  If you want to show the tradeoff between noise and blur, make it blur due to subject motion.  Take pictures of something that rotates, so you can see a sharp unblurred part as well as a clearly motion-blurred part, and compare that to a frozen but noisy picture of the same. Dicklyon (talk) 03:40, 8 January 2008 (UTC)


 * Can you explain how including blur complicates the comparison if there's a white featureless piece of paper in both photos? Viewers can compare the amount of noise on the paper and also on the featureless surfaces of the pencil sharpener.  Also, the photo gallery cut and pasted above by Hankwang contains the caption, "Both are approximately the same brightness, but the one at ISO 100 is less noisy and has motion blur from the lower exposure time."


 * Ah, yes, I was remembering the original caption. But the motion blur was from camera shake, not something that necessarily goes with exposure time. The fact that nothing is sharp makes the photo look a lot worse than such an ISO comparison should normally be expected to show.  The featureless background and camera shake contribute to the problem.


 * You and other users are welcomed to post high-quality photos at any time and insert them into the article. However, readers need to see what photos look like when shot at different ISOs.   --Bowlhover (talk) 05:19, 8 January 2008 (UTC)


 * Readers need to see these photos? I don't think so. Dicklyon (talk) 05:32, 8 January 2008 (UTC)
 * Most people are not photographic experts. They do photography pick up a camera, put it on the full auto setting, and take their photo.  It would be very useful to these people to show that the fact some photos are unacceptably noisy is because of the high ISO. Explaining increased ISO causes increased noise is not helpful if the person doesn't know the meaning of "noise".  --Bowlhover (talk) 23:58, 8 January 2008 (UTC)
 * While there's some doubt in my mind that this needs illustrating here at all, the above isn't the best way to show increased noise and the benefits of increased ISO in the same example. Motion blur is a perfectly fine way of illustrating what happens when you don't use higher ISO in reduced light, but it also makes a direct comparison of noise much more difficult for the casual observer. We might know that noise occurs, blur or not, but it would be easy to (wrongly) infer from this example that the lack of noise was due to blurring, and not lower ISO. It's not "scientific" enough; there's no control. If we want to show the reason for higher ISO as well as the noise penalty, I agree with Dick that a moving subject, like a vehicle or sports shot, would be much better. The caption could then point out a static camera and the static background would provide the control with which noise levels can be directly compared. --mikaultalk 09:33, 8 January 2008 (UTC)

I think the article would benefit from a photo comparison of different ISO settings. I like Dicklyon's suggestion to have a picture with something that moves against a steady background. It would be nice, though, if the picture had some aesthetic value as well when cropped (not scaled) to 250x150 px. Han-Kwang (t) 09:40, 8 January 2008 (UTC)


 * I'll try to do that, but if you think you have the necessary equipment, please take the photos yourself too. --Bowlhover (talk) 23:58, 8 January 2008 (UTC)

ASA 6
Where does the information that the original Kodachrome was ASA 6 come from? I've read in some sources that it was 8 and in some that it was 10, but never 6. —Preceding unsigned comment added by 130.232.17.50 (talk) 09:23, 6 May 2008 (UTC)


 * Here's a book that says it was 6, and doubled after the war to 12. Dicklyon (talk) 23:20, 3 February 2009 (UTC)

Since ASA was in flux before the 1950s, a speed of "6" in WWII does not necessarily correspond to a speed of "6" ASA today. While this wasn't the case with reversal, to be sure, ASA speeds for B&W were effectively doubled in the 1950s. I have no source to back this up, but I've seen it mentioned many many times by those knowledgeable in such things in Photography Fora on the internet. —Preceding unsigned comment added by ISOGuru (talk • contribs) 23:52, 10 November 2010 (UTC)

Readability
Parts of the article seem to be written more to convince the reader that the author knows the subject than to explain the subject to the reader.

Could it be simplified - perhaps by adding more explanations and links? Perhaps by adding very short summaries to each section?

One thing I had problems with was "clipped or bloomed". To me they are not well defined terms. If it is standard terminology, it would be good to explain them, either here or in separate articles. Mlewan (talk) 07:12, 31 May 2008 (UTC)

Lay explanation request
I'm an amateur who understands film sensitivity and was looking for an explanation about digital sensitivity. I do not pretend to understand the math or the formulae for either film or digital. From my perspective, what is missing in this article is a layperson's summary of what generates digital noise. The key is on "layperson". That will probably lead to a major simplification and result in long and endless debates about how much simplification is appropriate! I don't know the answer to that. but I do know that this article as it exists today (3 Feb 2009) is way over my head and provides me with no useful information.


 * Including such a discussion would be off-topic. Noise in digital images is unrelated to film speed. Noise in digital images comes from a number of sources, but film speed (ISO setting) isn't one of them. For most current digital cameras, most of the noise is shot noise. Shot noise is a function of the number of photons captured by the sensor. Using a higher ISO setting is typically used in low-light situations, where the number of photons captured is low and the level of shot noise is therefore high. But it's the low light, not the high ISO setting, that causes the increase in shot noise. Wikipedia has a specific article on image noise. Doug Pardee (talk) 19:29, 17 June 2009 (UTC)

Also, the eraser images above. For me, the blurring confuses the demonstration of noise. I do understand why the image is blurred. Showing two effects in one photo will lead to confusion among those who come here seeking info. Three shots would be a better demo. But make the slow shot at a reasonable speed that people would use for hand held. I believe that few point and shoot cameras would let one take a shot at 1/5th without some warning, and most users of DSLR would probably understand that 1/5th is too slow for hand held - even with vibration reduction. (Though I'd probably still push it and hope!) 207.236.137.2 (talk) 18:47, 3 February 2009 (UTC)Peter in Canada]


 * The images above are there as examples that didn't work; ignore them.


 * The lay explanation is this: it takes a certain amount of light to be make a decent-looking image; the ISO speed is inversely proportional to that amount of light. The exact factor in that proportionality is determined by standards of how much noise looks decent, how light is measured, etc.  What causes the noise is not relevant.  How much more detail would you like?  Dicklyon (talk) 23:17, 3 February 2009 (UTC)

Exposure index
I am struggling with the relationship between "exposure index" and film "sensitivity". The exposure index section confuses me:


 * Exposure index, or EI, refers to speed rating assigned to a particular film and shooting situation, and used in the exposure meter or equation, to compensate for equipment calibration inaccuracies or process variables, or to achieve certain effects. Exposure index may or may not be the same as manufacturer's film speed rating for that particular film ...

That description and the lengthy example that follows it seem more concerned with "arbitrary exposure compensation" than "film speed" calibration. I've got it into my head that:


 * 1) ISO, ASA, DIN, etc. are all species of "exposure index". So, for example, "ISO 100" would be an "exposure index value".
 * 2) Therefore the primary function of an exposure index value is to 'calibrate' an exposure meter to the sensitivity of the film being used.
 * 3) Therefore any intentional offset from the actual film speed (ie. the recommended EI for that film's sensitivity) should be called an "exposure compensation".

Where am I going wrong? --Redbobblehat (talk) 22:43, 16 February 2009 (UTC)

Some simply express this as a new ISO rating (EI800) which is the effective sensitivity; calling this the EI isn't exactly incorrect, but it is extremely confusing! More properly, it's used as an index (surprise surprise) and expressed in factors corresponding to the shift in exposure required (exposure compensation) So a ISO400 to ISO800 shift means sensitivity rating is indexed upward and hence has an EI factor of 1. Viewed this way, an EI of 800 equates to an 800 stop difference in exposure... it's unfortunately become common to express EI factors as effective ISO ratings, helped no doubt in great part by this article! I can find at least two references in my library for the factor-based expression but it seems the net is full of the conflated version; I've even seen it used to describe manufacturer's sensitivity ratings, which is plain wrong. I'd certainly like to see this properly addressed here, anyway; sorry for the confusion. Hopefully you're a bit clearer now? mikaultalk 01:54, 17 February 2009 (UTC)
 * You're conflating film rating with EI, and to be fair, so is the article to a large extent. An ISO/ASA rating is an expression of light sensitivity; exposure compensation is something photographers do which may require a deviation from the manufacturer's rating. An exposure index is required to quantify that deviation.


 * Exposure index just means the speed setting that you set your meter to. Like in this book.  It can include "compensation" and "calibration" effects, but tends to start close to your rated ISO speed. Dicklyon (talk) 02:12, 17 February 2009 (UTC)


 * Properly, ISO speed, EI, and exposure compensation are distinct entities. The ISO speed is a sensitivity under specified standard conditions; an EI is a bit looser. In some cases, it can represent an effective speed; for example, black-and-white film processed in a given developer to specified minimum density and contrast. The conditions may differ from those in the ISO standard for the film, but they work toward the same end. This is common practice among users of the Zone system. In other cases, an EI is a more arbitrary adjustment, such as a photographer “pushing” film.


 * The usual meaning of “exposure compensation” is adjustment to a meter reading of a subject that isn't statistically average. With the Zone system, the adjustment is systematic; in other cases, it can be, well, less systematic. JeffConrad (talk) 02:22, 17 February 2009 (UTC)


 * Thanks guys, this is really helpful :) I also found some good explanation in the push processing article. This is definitely getting clearer but I still have a couple of loose ends:


 * Would you use the term "exposure index" only in the context of the push/pull workflow ? ie you would not use it to describe a deviant-subject-metering-compensation ? So you might say An "exposure index" is used in place of "ISO rating" when the exposure is offset from the ISO rating in anticipation of pull/push processing ? ... If this is the case, it may be better to #REDIRECT exposure index to push processing and deal with it there rather than in film speed ... ?


 * No, it doesn't matter a bit if you're pushing or pulling. You might set exposure index 200 and process normally for 100, for example, to get darker image of a scene that you feel would otherwise be overexposed; in that case it's called "exposure compensation"; on many camera you'd set 100 and -1 stop to do that, but if you have a regular meter you might just set EI 200.  Or maybe you know your shutter is slower than it should be, and that's what you're compensating.  It can be used for lots of things; the point is, it's the speed you meter for, as opposed to the speed of the material or the processing. Dicklyon (talk) 06:27, 24 February 2009 (UTC)


 * IMO Film speed#Conflated ISO rating would be a useful section, though Effective ISO rating might be more prudent ;-). It might mention "exposure index", "effective ISO", "equivalent ISO", etc, etc, with links to whatever appropriate articles give fuller explanations ... a handy "heads up" section without adding too much to film speed's already groaning burden ?


 * Mikaul, apart from the conflated "Effective Iso" expression "EI 800", how else would the EI be written ? I'm more familiar with expressions of video sensitivity such as "Senstivity: 0.5 lux (15dB)", so I guess (for 2 stop push) eg "ISO 100 +2" ? or "ISO 100 +2EV" ? or "ISO 100 +2EI" ? ... something like that ?


 * Don't be introducing "effective ISO" and such without a source, nor confuse it with exposure index (EI) please. Dicklyon (talk) 06:27, 24 February 2009 (UTC)


 * Alot of my confusion came from underestimating the ambiguity of the term "exposure compensation" ... (see Talk:Exposure_compensation) --Redbobblehat (talk) 15:18, 23 February 2009 (UTC)


 * I don't think that begins to explain where your confusion comes from, but it's a good example anyway. Dicklyon (talk) 06:27, 24 February 2009 (UTC)

“Improper” rounding of ISO speeds
As it stood prior to the 8 January 2010 edit by 75.118.66.209, the table agreed with ISO 12232:1998 for speeds up to 10,000; absent a persuasive argument to the contrary, I think we need to treat that standard as authoritative. Both ISO 12232:1998 and CIPA DC-004 2004 stop at 10,000. I don't have ISO 12232:2006, but if it gives speeds beyond 10,000, those values should be used; if not, we should defer to values given by camera manufacturers. ASA PH2.12-1961 (Table 2, which compared arithmetic and logarithmic parameters) showed a value of 12,500, and this table was included in Appendix 2 of ANSI PH3.49-1971; the table was not included in ISO 2720:1974. Because that value hasn't been repeated in almost 40 years, it's difficult to see it as controlling.

More often than not, standardization lags practice, and accordingly, standards often simply document practice. Canon and Nikon have stated the higher ISO speeds as they have, and absent solid authority to the contrary, those are the de facto values. The role of Wikipedia is to document, not advocate, and we simply can't arbitrarily replace a widely accepted value with one of our own, no matter how logical a particular editor may think it.

I agree with 75.118.66.209 that the ISO rounding is somewhat arbitrary, but it follows long-established practice, and in any event, ‘tis as ‘tis. Whether the rounding is “improper” is a matter of personal opinion. In theory, speeds progress in steps of the cube root of 2, and I suspect (but cannot document) that meter calibrations make that assumption; the notes to Table 2 of ASA PH2.12-1961 stated as much. But the notes to that table stated that the “exact” value of ASA 25 was 26.7 due to the arcane conversion between ASA logarithmic and arithmetic speeds. If that is still the case, insisting on “proper” values for every speed would lead to quite different from those to which we are accustomed. Because APEX never really caught on, I'm not sure that table is still meaningful; more recent standards (e.g., ISO 2720-1974) make no mention of it.


 * A minor correction here: ASA PH2.12-1961, Table 2, stated that the value for ASA 25 speed used in computing the tabular luminance values was 26.7; the precise value for ASA 25 speed was $$32 / \sqrt[3]{2}$$, or approximately 25.4. In other words, all speeds that were integral powers of two were exact. With those rules, ASA 100 was actually approximately 101.6. Those rules apparently applied only to the values when used as APEX parameters; the speeds that obtained from the midpoints of the ranges for log10 Hm would seem to imply that none of the values was exact. For example, in ASA PH2.5-1960 for black-and-white film, values for log10 Hm of −1.65 to −1.56 were specified as ASA 32; using the midpoint, −1.605, the formula


 * $$S = 0.8 / H_\mathrm{m}$$


 * gives S ≈ 32.2. The current standard, ISO 6:1993, gives the same range for log10 Hm. We could speculate endlessly about the “precise” values, but speculation is all that it would be. This would seem all the more reason to accept the ISO values as stated, recognizing that they aren't exact, and that actual imaging media could have sensitivity anywhere within the allowable ranges for log exposure. JeffConrad (talk) 06:24, 11 January 2010 (UTC)

Absent an authoritative source, I'm not sure it's possible to determine the “exact” values for each speed. If indeed it were possible, insisting on “exact” values would require changing almost every value in the table. Given that speeds are rounded to the nearest 1/3 of an exposure step anyway, the quest for exact values seems pointless.

Accordingly, I've restored the ISO speeds to their previous values. If 75.118.66.209 can provide an authoritative source for the values he suggested, the matter is of course open to discussion; otherwise, I think he should refrain from further such edits before discussing the matter here. JeffConrad (talk) 02:20, 11 January 2010 (UTC)

Despite repeated requests for sources, 75.118.66.209 insists on adding unsourced material that in some cases is essentially original research (e.g., “higher speeds can be extrapolated”) and in other cases is unsupported personal opinion (e.g., “corrupted”). When material is challenged, the burden of evidence is on the editor adding the material, and this editor defiantly refuses to provide any supporting material or even respond on the talk page. Accordingly, I've removed the unsourced material, in accordance with WP:V. JeffConrad (talk) 20:44, 11 January 2010 (UTC)

I think most of us can work out progressions of speeds, but we cannot displace de facto standard values with our personal preferences. In any event, material like “If they called the values "100,000,000 and 500,000,000" that would make the numbers right too :-/ Learn some basic math. . .” is clearly inappropriate for an article, and it's difficult to see this as a good-faith edit. And the comment “Its on” seems an obvious challenge to an edit war. Accordingly, I suggest we treat this as vandalism, and I have so warned ISOguru. Anyone else? JeffConrad (talk)

As the article stands, it includes unsupported personal opinion that's at odds with de facto standard values. I think we should follow WP policy and insist that controversial matherial be supported, but I am not going to waste time in an edit war with ISOguru/75.118.66.209, who apparently sees no need for sources or discussion. JeffConrad (talk) 23:48, 11 January 2010 (UTC)

I've made some changes in attempt to satisfy ISOguru while keeping to what can be supported by reliable sources. ISO 12232:1998 does not specify a speed greater than 10,000, but the upper limit for Snoise 10,000 is given as 12,500, suggesting that ISO may have envisioned the progression that ISOguru claims. But this is speculation, and at present, the only devices with speeds greater than 10,000 have specified speeds using a power-of-2 progression that continues from the highest previously realized speed of 6400. Because these speeds are not given by ISO, they are ISO equivalent speeds, as the comment in the table already notes. Because values using the Canon/Nikon progression have actually been specified, I've restored those values to the table rather than speculate on what ISO intended or may eventually adopt. If another manufacturer announces a product with a speed using the progression that ISOguru proposes, I suggest we address it at that time by changing the entry to something like


 * 12,500/12,800[Note 3].

I've tried to address ISOguru's issue in Note 3. Although there is some speculation, I've made it clear that such is the case. Without some speculation, the alternative progression cannot be addressed until we have a reliable source. Hopefully, this will be acceptable to all concerned.

I've removed references to “proper” value and “corrupted”, because these terms are non-NPOV, and accordingly have no place in a Wikipedia article. If ISOguru takes issue with Canon's and Nikon's decisions, this is not the place to express that dissatisfaction.

It's perhaps helpful to examine several progressions of ISO speed, as shown in the table below. The columns show, from left to right


 * The ISO logarithmic speeds
 * The ISO arithmetic speeds, with the values proposed by ISOguru in parentheses
 * The ISO equivalent speeds specified by Canon and Nikon
 * A power-of-2 progression for which the speed is exact for integral powers of 2 (e.g., 32)
 * A power-of-2 progression which the speed is exact for ISO 25 (and hence also ISO 100)

To keep the size digestable, I've elided most of the values that don't correspond to integral power-of-2 progressions from ISO 32.

The progression with exact speeds for integral powers of 2 is arguably the most logical, but the progression with exact speed at ISO 100 is a better match to the current integral power-of-2 ISO values, and is an exact match to the values chosen by Canon and Nikon. Without support from a reliable source, however, we'd simply be guessing at which values are exact. In practice, I don't think it matters all that much, because the actual values are rounded to the nominal values, and may not exactly correspond to any progression.

Anyway, see if this works. JeffConrad (talk) 11:11, 12 January 2010 (UTC)

Dispute tag
I've added a tag, and  tags for specific items. My previous comments should suffice, but in case they do not, I ask ISOguru/75.118.66.209 to provide a source, other than his opinion, for the “proper” values, as well as the “standard ISO progression”. Because the current ISO progression apparently stops at 10,000, any continuation is speculation. We could forever debate whether Canon and Nikon make the “proper” choices, but both have stature to effectively set standards. Neither ISOguru nor most other Wikipedia editors can make the same claim. It's perhaps much like using EV for ISO 100 speed to indicate luminance. Though some editors here disagree with this practice, it's been established for 40 years, and to rail against it is pointless. Standards organizations, and often equipment manufacturers, set standards and practices; Wikipedia editors simply document them.

If it is thought that this article should comment on Canon's and Nikon's possible deviation from an established pattern, it should be discussed here before such comment is arbitrarily added to the article. JeffConrad (talk) 02:05, 12 January 2010 (UTC)

So, you are basically making the same argument as "The winner of a war can rewrite history." Just because Nikon and Canon have chosen to disregard convention established by Eastman Kodak, Mees and James, the American Standards Association and the subsequent ISO convention does not mean these numbers are correct. Frankly I couldn't care less Canon and Nikons whims upon the subject, since these are only "ISO equivalents." ISO is a film measurement system that has been crudely modified to express digital exposure ratings. ISOGuru (talk) 00:14, 11 November 2010 (UTC)

No one has commented on this edit, so I assume the dispute is more or less resolved. If there is no objection, I'll remove the tag that I added. I believe ISOguru's concerns have been adequately (and perhaps excessively) addressed. I also think the Note 3 reference in the last column of the table suffices, and that removing the references from the first column would reduce clutter. If no one objects, I will do this. The situation might change if another manufacturer specifies a speed rating using the progression that ISOguru proposes. JeffConrad (talk) 02:39, 15 January 2010 (UTC)


 * I have no objection; you've handled this well. I always thought ISO/ASA speeds were a constant in regard to films, which is why I questioned the 'rounding up'. Progression beyond a set ISO/ASA speed limit may be speculative, but if DSLR manufacturers have implemented this and documented it I have no cause for concern. I'm no expert on sensitometry, BTW. Your work is appreciated. Baffle gab1978 (talk) 04:52, 15 January 2010 (UTC)


 * I'm not sure I understand the comment “thought ISO/ASA speeds were a constant in regard to films”; perhaps this section is still not as clear as it needs to be. JeffConrad (talk) 09:41, 15 January 2010 (UTC)
 * I meant that I thought an ISO, ASA or DIN speed number is a set measurement of a film's sensitivity, it doesn't change although one can up- or downrate a film, where the new speed is termed Exposure Value). So if I buy a roll of FP4+ it will *be* 125 ISO/ASA, not 128 ISO/ASA as the other editor had suggested. But then I'm no expert on the subject, just a some-time user of some very consistent modern emulsions. Baffle gab1978 (talk) 06:57, 16 January 2010 (UTC)


 * No number of magical incantations can change the ISO speed of a film, but the precise actual value is not so cut and dried. The standard for B&W films, ISO 6:1993, gives the formula for arithmetic ISO speed as


 * $$S = \frac {0.8} {H_\mathrm{m}}$$ ,


 * where Hm is the exposure that gives a density of 0.1 above film base plus fog. Any measured log Hm from −2.25 to −2.16 is reported as ISO 125. If we apply the formula to each limit, we get measured speeds of 115.6 and 142.3; the mean value for this range is 128 or 129, depending on whether we take the mean of the logs or the mean of the calculated speeds. The reported value of 125 is a “convenient” rounding that may or may not better represent the actual speed than would 128. The difference is 0.03 step, so the effect on exposure calculations is negligible. In theory, the actual speed could differ from the reported speed by 1/6 step, and that two films with the same reported speed could differ in actual speeds by as much as 1/3 step. I added a clause mentioning the rounding to the section Determining film speed in attempt to make this more clear.


 * It is quite possible that the reported value of 125 is indeed rounded from 128, as ISOguru maintains and ASA PH2.12-1961 implies. But we don't know this; the reported value could also be rounded from 125.9 (a 1/3-step increment from ISO 100), a power-of-2 progression that would make the Canon and Nikon values exact. My issue was with stating speculation as though it were fact. I mentioned the power-of-2 progression from ISO 32 in the note I added to the section Applying film speed, but went no further than saying that ISO PH2.12-1961 so stated. As I mentioned, though, the differences among the several possible precise values are minor compared to the amounts by which the actual speeds can differ from the reported values. In theory, I don't necessarily disagree with ISOguru, but we need to support what we claim as fact, especially if it differs from what is commonly accepted by the majority of the photographic community. JeffConrad (talk) 09:44, 16 January 2010 (UTC)

The Canon and Nikon iSO values are NOT exact, because they are taken from doubling OTHER ROUNDED NUMBERS. There is nothing exact about ISO anyway. It represents, in an easy-to-understand manner, changes in negative film density of log [10] 0.30 shifts on negative film. 0.30 has to be a stop on a negative film (3-1/3 stops change per base ten order of magnitude). There is nothing exact about any of the ISO speeds. The only "exact" ISO number, I think it was 100 when the standard for ASA was originally written is a basis for deriving other numbers. ISO values are no different than F/stops, which not only do not always take into account the actual size of the stop, but also do not account for light loss due to internal element refraction and due to zooms and other conditions. ISO film speeds similarly do not give actual film speeds. Manufacturers have routinely called ISO 200 films "100," ISO 640 films as "1600" films, and 4- to 500 films as "800" films. ISO film speeds do not in any way represent actual numerical relationships between film speeds, they are rough denominations that relate to sensitometry-ignorant photographers 0.30 progressions (negative film log 10 density) under or over 100 ISO speed. ISOGuru (talk) 05:08, 16 November 2010 (UTC)


 * I'm not certain that the 100, 200, ..., 6400, 12,800 sequence represents a rounding, though each speed applies to a ±1/6 step range of speeds, so in that sense, nothing is “exact”. As I have mentioned several times, ASA PH2.12-1961 stated fairly clearly that ASA 32 was exact, and absent reliable information to the contrary, we need to accept their statement. It is difficult to tell whether film speeds really were in cube-root-of-two progression as ASA indicated, or in 0.1 log steps as the log density ranges in the speed tables seem to suggest. I don't think I've ever encountered disparities between claimed and actual speeds that you cite; in some cases, my effective speeds for monochrome film differed from ASA speeds, but the differences were mainly the result of different developers, processing technique, and contrast criteria. Incidentally, lens transmission losses are accommodated in meter calibration equations.


 * With regard to the precision (and consistency) of film speeds, the difference between 0.3 and log&thinsp;2 is insignificant, and accordingly, ASA probably just ignored it. In some instances, at least ostensibly, ASA were pedantically precise, yet in almost the same context were quite sloppy. A good example was the conversion between ASA arithmetic speed and the stillborn ASA logarithmic speed; when originally described in the foreword to ASA PH2.5-1960, it was given precisely as


 * $$S_\mathrm{v} = \log_2 \left ( NS_\mathrm{x} \right ) \,,$$


 * where Sv is the logarithmic speed, Sx is the arithmetic speed, and N is a constant with exact value of 2−1.75 (0.2973). In later publications, they apparently recognized that most people would not know how to calculate base-2 logarithms, and they gave it as


 * $$S_\mathrm{v} = 3.32 \log \left ( NS_\mathrm{x} \right ) \,,$$


 * with N = 0.30, and no explanation of where the magic numbers came from, making determination of the exact values almost impossible for anyone but those mathematically inclined and extremely curious. The constant N was apparently so arcane that they added a detailed explanation to ASA PH2.12-1961 (without which I'd probably not have figured it out). The differences may seem small, but they are enough to have thrown off one of the exact values mentioned in ASA PH2.12-1961 by 0.2 step.


 * So I guess in a sense I'm touting their precision and criticizing their lack of it almost in the next breath. The point is that in light of considerable apparent slop, taking ASA at their word on the cube-root-of-two progression is hardly a stretch, especially in this article, which is more about photography than sensitometry.


 * It now appears that ISO 100 is the standard for most things, but the ISO standards, unlike their ASA/ANSI predecessors are short on explanatory material, and make no comment on which numbers in the sequence are exact. Again, it probably is not unreasonable to treat the power-of-two progression from ISO 100 as exact; in the case of digital cameras, that well may be the case, but we don't know because the manufacturers don't get into that much detail.


 * Quite honestly, I fail to see the point of this discussion, because even if we think Canon, Nikon, and perhaps others made the “wrong” choice, we simply cannot make such comments in the article, and we cannot state “exact” values from which others are converted without support from a reliable source on that specific issue. Accordingly, I don't see what we can change in the article except perhaps for stating more clearly that speeds above 10,000 are technically not ISO values because ISO have yet to assign them. And even there, we'd need someone who has a copy of ISO 12232:2006 to confirm that they aren't mentioned there. We've rehashed the same arguments three or four times, and I see no reason to further beat a dead horse. JeffConrad (talk)

“Exact” value to double logarithmic speed
I removed the note


 * Strictly speaking, the correct quantity to add to achieve a doubling is equal to ten times the base-10 logarithm of 2, which is approximately equal to 3.0103,

which was appended to one of the references. Ostensibly, this seems reasonable, but as previously discussed, we don't have an authoritative source for exact values, so the statement is a bit speculative. Because ISO logarithmic speeds are rounded to the nearest integer anyway, the point would seem moot (in the American sense). If it's felt this comment should be included, it should have a source and be a separate note appended to the previous sentence so it's clear to what it applies. JeffConrad (talk) 21:14, 11 May 2010 (UTC)

What's the obsession with powers of 2?
The argumentation above bases on two things: The first is clearly correct. The second is demonstrably incorrect. Look at the ISO rating: there are linear and logarithmic values, though for some reason few people use the latter. But they also make it very clear that the steps are related to powers of 10, not powers of 2. The steps are the tenth root of 10. 10,000 ISO linear is 41° logarithmic. 51° logarithmic is 10 times 41° logarithmic, 100,000, and not 102,400. This makes the table inaccurate. It also means that the current section 1.2 is wrong: doubling the sensitivity adds approximately 3° to the logarithmic scale (closer to 3.0103°).
 * Canon and Nikon are now doubling speed ratings rather than sticking to the progression that has been in place up to the current limit of the ISO range.
 * Claims that the speeds are in some way related to powers of 2.

I believe that this correct procedure is to state this information, and that the camera manufacturers haven't adhered to the intention of the scales. Groogle (talk) 07:25, 23 February 2010 (UTC)


 * Obsession? Sensitometrists may work in powers of 10, but photographers, like camera controls, work in powers of two.


 * Though the ISO standards give conversions in terms of powers of 10, I'm not sure it's settled that those are “exact” values, because unlike the earlier ANSI standards, the ISO standards don't discuss this. And even the ANSI standards were notorious for sloppy rounding and embedding of “magic” numbers in formulas (the conversion between ASA arithmetic speeds and the stillborn ASA logarithmic speeds was a good example). It could simply be that the drafters of the standards assumed that most people would not know how to compute base-2 logarithms. But absent a definitive statement in one of the ISO standards, any such interpretation is just a guess.


 * There's indeed a very good correspondence between the former DIN logarithmic progression and the rounded arithmetic values up to 10,000, but claiming that those values are “exact” would be an inference on our part. The arithmetic speeds included in the ISO standards derive from the former ANSI arithmetic speeds; ASA PH2.5-1961 indicated that the exact speeds were integral powers of 2. Is this still the case? Absent additional information from a reliable source, I don't see how we can make the call. I've tried to cover this as far as possible while sticking to supportable statements. If someone has a better idea, I'm all for it. But we need to stick to what we can support with reliable sources. JeffConrad (talk) 02:42, 24 February 2010 (UTC)

Thanks for your take, Groogle. While "Photographers" work in powers of 2 with stops, lab technicians, and those practicing the science of densitometry (where ISO/ASA/DIN is determined) work with log sub 10 powers. A D-max of 4.0 above base fog means that a film only lets through 1/10,000 the amount of light that the D-min would admit on a piece of transparent film.

I find it funny that those who criticized, censored, and deleted the correct progressions back in January expected me to easily find a source on a website backing up my claims. I have since misplaced my Petersen's Photographic Library, 1978 Ed. "Increasing Film Speed," which clearly indicates EIs of 10,000. I have a scan from that book for anyone who doesn't believe me, but not hte article. However, with quite a bit of searching, I have found two references to film speed that do include the proper progressions:

"[With reference to the new Kodak TMZ 3200 5056 film:] Can be exposed at speeds of EI 800 and EI 1600 with excellent results. Allows a high degree of enlargement. ... T-MAX P3200 Professional Film Multi-speed panchromatic negative film that combines high to ultra-high (EI 3200 to EI 25000 ) ... "Kodak scientific imaging products" Eastman Kodak Company. Photographic Products Group - 1989 Ed.

"2475 Recording Estar +DK50 [Using this B&Wphotographic developing agent]. Probably the fastest combination in the world.  2475 Recording Estar was designed for low-light surveillance work by bodies such as the FBI.  It is a very grainy emulsion only suitable for special effects or for use when no other stock will produce the required results.  The basic film speed is 1600 ASA and it can be pushed up to 10,000 ASA, although at this speed [. . .] "   From a Kodak leaflet on 2475 Recording film, undated.

So, if someone wants to correct the article, fine. I am done "ranting" on here. But I am not full of_ _ it just because the "infinite" knowledge of the internet only contains incorrect speeds propagated by camera manufacturers who wish all photographic sciences were dead so they could sell more software and printers, and other uneducated photographers who make the same ignorant rounding mistakes.

Another hint for my esteemed colleagues: If log 2 powers were being used here in ASA film speeds, why are they calculated in THIRDS of a stop? Seems rather unusual to me. ISOGuru (talk) 00:15, 11 November 2010 (UTC)


 * The suggestion of “censorship” completely misses the mark. The “correct” progressions were a matter of speculation, which we cannot do in Wikipedia. The authoritative source for ISO speeds is ISO, and they simply never formally assigned speeds greater than 10,000. Though it's not illogical to extend the sequence as was suggested, ISO did not do it. So while it might be reasonable to say that the higher values apparently didn't follow the progression that may have been intended, it is not reasonable to say that they are incorrect. Whether or not the camera manufacturers made the correct choices is a matter of pointless debate—absent values assigned by ISO, a decision was needed, and for good or for ill, the manufacturers did what they did. Though perhaps they don't have quite the authority of ISO, they speak with considerably more authority than do Wikipedia editors. Whatever is regarded as the “correct” progression, it's obvious that some values in any of the sequences are “rounded” to “convenient” values probably governed more by history than anything else.


 * A “source on a website”, per se, would be meaningless. Wikipedia requires reliable sources, and as previously indicated, ISO are the authoritative source. When ISO are silent, camera manufacturers speak with considerable authority, certainly as much as Kodak.


 * It's quite likely that 1/3 log2 steps are inspired by (or perhaps even converted from) 0.1 log10 steps, but they're stated as they're stated. Absent something specific from a reliable source, the origin again is speculation. As I indicated in more detail previously, the ASA standards claimed an exact power-of-two progression. The ISO standards merged ANSI and DIN speeds; however, the standards that show both arithmetic and logarithmic speeds list only the rounded values, so it's impossible to tell which values are meant to be exact. Given that the rounded values represent ranges of speeds, the whole discussion of what value is “exact” seems a bit silly.


 * Once again, we state what we can support with reliable sources, not what we think may be reasonable. JeffConrad (talk) 01:37, 11 November 2010 (UTC)

I just posted not one, but two reliable sources. I don't know what more I need to provide. YOu deleted all of my hard work in January, expecting me to know all of the inner workings and procedures of finding your posts here. Now it appears you don't accept the sources I've provided because they don't have a URL or a digital camera maker's name attached to them. There are plenty of weak points in this article that are not adequately backed up by solid citation, yet you delete mine because it is not formatted correctly, or is a rant? Do you think I made these sources I have posted now up? Who cares what is in the ISO standards. They have a very clear progression. You make arguments similar to "Almost everyone in the world media announced that 2000 was the start of the new millenium, not 2001, so this lends organizational credibility to this argument." ISOGuru (talk) 22:30, 12 November 2010 (UTC)

Here are some more sources:

"KODAK T-MAX P3200 EI 800-6400 Grain: RMS 18 RP: 125/40 lpm ID: TMZ, 5054 T-Max P3200 has excellent speed characteristics—it can be pushed to EI 50,000 for surveillance work and exposed at EI 3200-6400 for general shooting—with good grain and sharpness at the lower film speeds. We've used T-Max P3200 for many subjects in all kinds of lighting and have found it to be an excellent choice whenever film speeds above EI 800 are required." Source: http://www.photographic.com/film/37/index7.html  Retrieved today.

"Delta 3200 Type: Black and White Speed: ISO 1000/DIN 31 Available formats: 35 mm, 120 Granularity: Latitude: EI 1600/33 to EI 6400/39, up to EI 25000/45 with push processing Resolving power: History: Available since 1998 as a competitor to Kodak's T-Max 3200. Unlike Kodak's emulsion, it is available in 120 format. Primary usage: Low light and extreme low light depending on what speed it is exposed and developed at." Source: http://en.wikipedia.org/wiki/List_of_photographic_films  NOTE THIS IS RETRIEVED FROM WIKIPEDIA, making it the least reliable source, from what I've experienced here. But it is interesting that this site isn't even consistent with sister articles published by some of the same people.

http://www.apug.org/forums/forum216/34406-ektachrome-e-i-20-000-processing.html Here is a link to formulae that are almost certainly from the exact same EI I mentioned from the 1978 article in "Increasing Film Speed," giving EIs of 10,000 and 20,000. ISOGuru (talk) 23:16, 12 November 2010 (UTC)


 * Let me say it one more time: the only reliable source for ISO speeds is ISO, so we necessarily care what is in the standards. And n None of the sources you cite qualify as reliable sources for this issue.


 * I don't see the point of what you've quoted for 2475 recording film, because it doesn't appear to mention any speeds greater than 10,000. The claims for Kodak T-MAX P3200 are easily verified, but they don't establish the correct values for speeds greater than 10,000; Kodak have no more authority to do so than Canon or Nikon. As a small detail, the Kodak data for both films refer to EI rather than ISO speed, so they arguably can have any appropriate value. I don't suggest that the speeds greater than 10,000 assigned by Canon and Nikon necessarily establish the “correct” progression, either; again, ISO are the only ones who can speak to that with authority. ISO may choose to follow the lead of Canon or Nikon or they may not; if they don't, we can deal with it at that time.


 * As I have said several times, we simply cannot insert conjecture, however seemingly reasonable, as if it were fact. And in any event, what's the point of suggesting that Canon and Nikon have used the “wrong” speeds? They have given the speeds they have given; we indicate them as “ISO equivalent”, because they aren't values that ISO have actually assigned, so I think we properly cover it. Table Note 3 addresses the apparent departure from the existing progression, and goes about as far as we can go with support from reliable sources.


 * Once again, the whole point is that we don't insert material that isn't specifically supported by reliable sources, and we don't do op-eds. JeffConrad (talk) 00:20, 13 November 2010 (UTC)

You are doing an op-ed on my credibility in every one of your smarmy posts. EI and ISO are the same exact numbers. So is the first ISO number identical to the numbers used in ASA. There is a document I've posted there that says "25,000 ASA." You keep praising ISO standards (with the 12,500 ISO you don't list), yet you used digital camera speeds as "proper" listing. How is that objective? ISOGuru (talk) 04:48, 16 November 2010 (UTC)


 * Insulting other editors is usually a bad idea unless your objective is another block. Though we obviously disagree, I'd prefer to leave the personal insults out of it.


 * ISO speed and EI are most assuredly not the same for films. The former is a sensitivity determined in accordance with specified test methods; the latter is a camera setting (or value used in calculating exposure) that may give acceptable results for a given application, as Kodak clearly describe in the literature for T-MAX P3200. It is simply impossible to make a significant change in actual threshold speed; pushing a film simply increases contrast, and consequently, highlight density. As nearly as I can tell, the highest actual speed of any film ever produced was 1600. If you look at Table note 3, I specifically mention the reference to ASA 12,500 in an old ASA standard; again, I think this note goes about as far as we can go without getting into original research. I don't see anywhere in the article or this discussion where I've described any speed as “proper”; we've noted that the digital camera speeds are “ISO equivalents”. To me, this seems to cover it, but we probably could go further and specifically mention that no ASA/ANSI/ISO standard to date has listed a speed greater than 10,000. The edits you made earlier this year repeatedly referred to “proper” values and “corrupted” values; such comments are extremely POV and Wikipedia policy strictly forbids them. And that was the whole point of my objection; we can perhaps note that Camera manufacturers declined to extend the existing sequence, but we cannot chastise them for doing so. JeffConrad (talk) 06:27, 16 November 2010 (UTC)

Progressions, logarithmic speeds, and “exact” values
An examination of some of the ANSI and ISO standards reveals several things:


 * The ranges for each ISO film speed progress in 0.1 log steps (e.g., Table 1 of ISO 6:1993), consistent with sensitometric practice. At the same time, many ASA/ANSI standards state a cube-root-of-two progression, so I think it's tough to say which one is exact. The difference between the two progression is fairly minor, and given the extensive rounding inherent in film (and perhaps electronic sensor) speeds, it probably doesn't matter anyway. ASA PH2.12-1961 (notes to Table 2) states that tabular values of speeds and other parameters “are rounded off to a uniform series of numbers which are easy to remember”, which really isn't much help in explaining how the published sequence of speeds was derived. In sum, I think any explanation we might give would be no more than speculation, which we cannot do.
 * I can't find any ASA/ANSI/ISO standard that goes higher than 3200/36°, so I question whether we can reasonably include them here. ISO 12232:1998 doesn't even mention logarithmic speeds, so I'm not even sure it makes sense to include them with a qualifying not. I don't have the latest standards for color reversal and negative film or ISO 12232:2006, so I can't say for certain that what I've just said holds for them as well; perhaps someone who does have these standards can confirm or deny. It's possible than DIN actually gave higher values; I don't have a copy of DIN 4512, so I must defer to someone who does.
 * Though ASA PH2.12-1961 (notes to Table 2) implied that an arithmetic speed of 32 and values on a power-of-two progression therefrom were exact, the ISO standards are silent on the matter. A glance at speeds of most films available for the last 50 years, and especially in the last few years, shows that most films match a power-of-two progression from 100. In that light, I'm not sure Canon's and Nikon's decisions to continue that sequence are unreasonable. To the contrary, it could well be argued that they chose to go with a reasonable progression rather than continue with one that was flawed as well as arcane. Whatever the case, words such as “correct”, “improper”, “corrupted”, and the like clearly violate WP:NPOV, so we can't go there. And I'm not sure there's even any reason to go beyond what's in Table note 3.

JeffConrad (talk) 04:01, 16 November 2010 (UTC)

Ah, but "arcane" is objective. Your argument is totally flawed. 1250 is a rounded ISO/ASA/EI speed. If ISO is going to cave in and introduce ugly numbers, that doesn't seem to prove your point. It just shows that they are willing to cave into pressure from companies that got the numbers wrong.

As for taking Kodak, Fuji, and other film coating companies' use of speeds, I take them over Nikon and Canon, because they were working in some cases at it a CENTURY before Nikon and Canon were.

I still don't understand how it is "speculation" to see the 10 12[.5]* 16 20 25 32 40 50 64 80 *round or drop pattern that is repeated through three iterations in respected, documented literature. ISOGuru (talk) 04:56, 16 November 2010 (UTC)


 * ANSI/ISO 1250 is presumably rounded, but we don't know from what. A case can be made that it more logically follows from a progression of 0.1 log steps than cube-root-of two steps, but we don't really know, so we can't speculate.


 * It is one thing to note that recent values given by Canon and Nikon depart from the apparent sequence, but quite another to say that they're “wrong”, especially since ISO never assigned values to the contrary. In any event, we cannot make such comments in an article without violating WP:NPOV. JeffConrad (talk) 06:04, 16 November 2010 (UTC)

Speeds to be used in exposure calculations
At the end of Applying film speed, we state (full disclosure: I think I added this),


 * Commonly stated values for exposure parameters are rounded to convenient values that are easy to remember; however, the actual values to be used in exposure calculations are in exact power-of-2 progressions.

citing the notes to Table 2 of ASA PH2.12-1961. As has been discussed here and in Talk for other articles, the APEX proposal was essentially stillborn, so I’m not sure it’s reasonable to still take its recommendations as authoritative. Though I have no reason to suspect that equipment manufacturers ever did otherwise, a fit of the rounded speeds to an exponential function gives a better fit to power-of-ten progression than to a power-of-two progression. The difference is so minor as to be of no practical significance; the difference between actual measured speeds and the rounded values is probably far greater. Nonetheless, the sentence quoted above states as fact what, absent support from a reliable source indicating that this indeed is what manufacturers do, is only a reasonable assumption. Should we keep this statement, revise it, or eliminate it? JeffConrad (talk) 08:12, 19 December 2010 (UTC)


 * Since APEX never caught on, I think we should stick to the values in the ISO speed standards for film and digital. But they don't go high enough...  Dicklyon (talk) 04:44, 20 December 2010 (UTC)


 * So basically, get rid of this statement? I don’t think we would lose anything by doing so. As I’ve said several times, I think we should stick to ISO standards and avoid our own interpretations (which probably would be WP:OR anyway) to the extent possible. I do think we should accept the manufacturers’ higher speeds, because we really don’t have much choice. The 10th ed. of the Manual of Photography does so, and if it’s good enough for them, it should suffice for us, should it not? JeffConrad (talk) 05:10, 20 December 2010 (UTC)


 * Yes, as much as I hate what Canon and Nikon did, it's the reality we have to live with. Dicklyon (talk) 05:22, 20 December 2010 (UTC)


 * Is it really that bad? I strongly suspect that film speeds used to compute exposure have always been power-of-two progressions, especially with analog controls (it would have been difficult to use a power-of-ten progression in the simple linear resistors used in shutter speed/film speed dials, and the accuracy of those dials was probably less than the difference we’re talking about, anyway). An earlier version of this article indicated that most common films have been in the sequence that would exactly match the speeds Canon and Nikon have chosen, so we may eventually conclude that, in retrospect, ASA/ANSI made the wrong decisions and C&N were wise to end the charade. And if we note that the difference from the presumed ISO rounded values is 0.03 exposure step, we may conclude that far more serious issues confront us. In any event, C&N have written what they have written.


 * It breaks the easy pattern: 10, 100, 1000, 10000, 102400? Dicklyon (talk) 06:02, 20 December 2010 (UTC)


 * Yeah, it does. But we put up with all manner of silly things in photography, such as “shutter speed”, the progression thereof, and the progression for relative apertures, so I’m not sure this is any worse, though the sequence is longer and is almost certain to increase. But C&N are, well, Canon and Nikon, and we are but lowly customers . . . and lowly Wikipedia editors. Had the ISO people better anticipated what was coming, or been more nimble in responding, they might have called the shots. But they weren’t and they didn’t. Onward . . . JeffConrad (talk) 07:56, 20 December 2010 (UTC)


 * I’ve removed the statement. We cite ASA PH2.12-1961 in the notes to Table 1, so we still need it under References. I’ll add an author–date-type link if it’s thought necessary. JeffConrad (talk) 05:50, 20 December 2010 (UTC)


 * I’ve moved most of the commentary in that table note to References, which I think is a better approach because it removes clutter that probably is not of interest to the average reader. Moreover, the clickable link matches the others in the table notes. See if this works. JeffConrad (talk) 07:56, 20 December 2010 (UTC)

Including speeds for digital cameras in Table 1
Most digital cameras have ranges of ISO speeds that match film speeds of 3200 or less; if every digital camera is included in the list, it's going to be an unreadable mess. Is it reasonable to include the PowerShot SX200 IS but not the others? JeffConrad (talk) 22:23, 16 January 2010 (UTC)

High ISO vs Low ISO (on Digital Cameras)
In its current state, the article does not elaborate on the ISO ratings of digital cameras.

I recall from a Consumer Reports study on digital cameras that a high ISO is a good thing for several reasons. One, a high ISO means less light is needed. In other words, a high ISO means one doesn't need to use a flash to get a good photo of a candle-lit birthday cake. It also means that there is less blur when the camera and/or subject is moving because the picture is taken in less time. This is why cameras like the Nikon D3S and the Canon EOS-1D Mark IV, which are professional grade cameras, have been designed to be able to take pictures at 102,400. This article, however, is quite biased against high ISOs. Now, I do know that using a higher ISO can lead it poorer quality (ex:Panasonic DMC-FS15), but this article fails to mention that cameras rated at higher ISOs have less quality degradation at higher ISOs than cameras rated at a lower ISO.

In short, the intro (and the article in general) needs to talk more about the ISO ratings of digital cameras and not focus so much on film because the two are so different that they need separate treatment.--Humanist Geek (talk) 08:09, 16 April 2010 (UTC)


 * How should the article elaborate? And how is it biased against high ISOs?


 * Certainly, the performance at high ISOs of recent cameras has greatly improved. Dunno how much I'd rely on CR as a source for this article, though; all other things being equal, the availability of higher ISO allows shooting with existing light without going to an unreasonably long exposure time. But things are seldom equal, especially with most point-and-shoot cameras. I agree that the performance of better digital cameras at high ISO speeds is far better than can be achieved with film. But even the top Canon and Nikon cameras are only marginally usable at their top ISO speeds.


 * There are some constraints on this article, such as the need for reliable sources to support whatever is added—perhaps that's one reason why there' more about film. Can you suggest some that would support what you think should be added? JeffConrad (talk) 08:50, 16 April 2010 (UTC)


 * The problem is that ISO ratings and ISO settings are very different things, and camera manufacturers seldom tell you the ratings; the ISO standards define several signal-to-noise-ratio-based rating (SNR 10 and SNR 40 criteria, iirc), but manufacturers don't measure or report those; and they actually don't make much sense, since you can also increase your rating by blurring the image or applying other types of noise reduction.  So "a high ISO means less light is needed" would be correct in theory, but in practice, a high ISO just means less light is used.  Not many sources talk about this problem, and some just lie, like this one and this one.  Dicklyon (talk) 04:42, 20 December 2010 (UTC)

Sensor specs: mV/(lux-sec)
"Sensitivity" includes a disambiguation to this page. CMOS and CCD sensors list "sensitivity" in millivolts or volts per lux-second. This isn't touched upon on this page (although it comes close). What does that measurement mean? Don't these sensors "count" charge (electrons) not voltage? Is more volts per lux-second more sensitive or less sensitive? Thanks. —Ben FrantzDale (talk) 18:58, 15 November 2010 (UTC)


 * That's actually a measure of Responsivity. It's often associated with, or confused with, sensitivity; but it's really just a gain, and can be increased by using an amplifier, without making your detector any more "sensitive" in terms of more sensible criteria such as an SNR threshold.  You're right that the sensor detects charge, usually, but then there's a charge-to-voltage conversion amplifier that determines the responsivity.  Film speed and the roughly equivalent ISO sensitivity of digital sensors are actual sensitivities, not responsivities, as they measure the number of lux seconds that makes a threshold difference in the response, in terms of a fog level in the case of film, or an SNR threshold in the case of digital.  Where did you encounter the mV/(lux-sec)? Dicklyon (talk) 00:47, 16 November 2010 (UTC)


 * Thanks for the reply. I've seen it from two different CMOS vendors, OmniVision and Cypress Semiconductor, so I was starting to think I was crazy not to be able to make sens of the units as a sensitivity. On a spec sheet for a sensor very similar to the iPhone4's it says: "...unparalleled low-light image capture with low-light sensitivity of 960mV/(Lux-sec)" . Aah: looking for examples I found one, www.cypress.com/?docID=24434 [not a link 'cause it set off WP's spam filter] which has a glossary with an explanation (in addition to listing a sensitivity of 1256 V·m2/W·s):
 * Sensitivity A measure of pixel performance that characterizes the rise of the photodiode or sense node signal sensitivity in Volts upon illumination with light. Units are typically V/(W/m2)/sec and are dependent on the incident light wavelength. Sensitivity measurements are often taken with 550 nm incident light. At this wavelength, 1 683 lux = 1 W/m2; the units of sensitivity are quoted in V/lux/sec. Note that responsivity and sensitivity are used interchangeably in image sensor characterization literature so it is best to check the units. [emphasis added]
 * Even then, I don't understand how V/(lux-sec) is useful as that gain is, in my mind, within the black box that turns photons to digital number counts, and since it says nothing about the number of electrons (and hence shot noise). It seems like sensor specs are designed to be obtuse. I'd think they could start by just giving QE*FF, well depth, dark current/noise info, and ADC noise. In my mind I'd just like to think "The photosite was exposing for time t and was hit by 100,000 photons, what is the probability density function of the digital number that comes out?" Anyway, thanks again. —Ben FrantzDale (talk) 12:50, 16 November 2010 (UTC)
 * Yes, "sensor specs are designed to be obtuse" is a correct observation. The responsivity is almost completely useless.  You need the QE and the read noise (in input-referred electrons rms) to make any sensible performance estimates, and responsivity is irrelevant.  Dicklyon (talk) 05:59, 17 November 2010 (UTC)