Topic: ISO on digital camera (Read 20586 times)

maggieddd · « **on:** October 06, 2005, 09:09:32 am »

Does anyone know how a digital camera handles the ISO setting? I mean is this something that is done using software inside the camera after the photo has been taken? Or the sensor's sensitivy changes somehow? I have no clue, anyone know?

Graeme Nattress · « **Reply #1 on:** October 06, 2005, 10:17:05 am »

I think it's the gain on the ouput amplifiers on the sensor that gets changed. The sensor is an analogue device, and by changing the gain before it gets digitized you can get different sensitivities.

Graeme

Jonathan Wienke · « **Reply #2 on:** October 06, 2005, 10:19:54 am »

Graeme is correct.

maggieddd · « **Reply #3 on:** October 06, 2005, 10:22:53 am »

great thanks, does anyone of you have any written documentation supporting this? I argued with my teacher and I want to show him a proof. He claimed that it was software controlled and I disagreed and told him that it was not but couldn't really explain.

Thanks

Jonathan Wienke · « **Reply #4 on:** October 06, 2005, 10:29:33 am »

Do a search on the Rob Galbraith forums for posts by Chuck Westfall. He's a senior Canon employee (Director of Technical Information for Canon USA or something like that) and he's posted on that and similar topics.

Graeme Nattress · « **Reply #5 on:** October 06, 2005, 10:32:05 am »

Well, you could look from the point of view that given a CCD has limited dynamic range, that changing in the ISO post AtoD in the DSP just wouldn't work! You'd end up with posterisation artifacts at high ISO, and of course, we don't see these artifacts, only noise.

Graeme

maggieddd · « **Reply #6 on:** October 06, 2005, 11:29:27 am »

thank you

BJL · « **Reply #7 on:** November 11, 2005, 06:43:43 pm »

I agree about ISO Exposure Index value being adjusted by adjusting degree of amplification before A/D conversion, with one qualification. Some very high EI values might be done in the digital domain: shift bits one place left for each extra stop of speed. Hopefully only the "Hi" or "boost" speed settings like 3200 and sometimes 1600.

Anon E. Mouse · « **Reply #8 on:** November 14, 2005, 10:03:41 pm »

FYI, there are two ways of determining ISO on a digital camera - saturation based and noise based. Saturation based is the "natural" sensitivity of the sensor where the exposure provides the maximum saturation (camera signal) possible and give the highest quality. For increasing the ISO or having a changable ISO, a noise based ISO is used. The limit of noised-based ISO is based on the signal to noise ratio.

Most consumer cameras change gain to change sensor sensitivity, but you can also bin a sensor to increase it as well. Binning is combining pixels to increase the collection area, but at a cost of resolution. A 2x2 binned sensor increase sensitivity by four, but with just a lose of resolution by 1/2.

Anon E. Mouse · « **Reply #9 on:** November 14, 2005, 11:02:11 pm »

Just a little more trivia on ISO. Unlike film cameras, digital cameras don't actually use ISO to determine exposure since they can get a signal directly off the CCD.

BJL · « **Reply #10 on:** November 16, 2005, 01:48:52 pm »

Actually, there are at least three ISO standards relating to measurement of sensor speed. (The ISO publishes thousands of standards!)

1) What we used to call Exposure Index (EI), which determines the combinations of shutter speed and aperture ratio that will give the right level to metered mid-tones. That is the speed or sensitivity that is adjusted by the "ISO" setting on digital cameras, and I will call it EI from now on.

2) A saturation based lower limit on EI, or "base ISO speed": the lowest EI that can be used without excessive risk of blown highlights, requring about three stops of headroom above metered mid-tones (170/18 times the luminosity). This is like the limit to deliberate underexposure and push processing. (Compact cameras often fudge on this, giving a minimum EI with cramped highlight headroom in order to offer a low minimum EI and thus less shadow noise at minimum EI. Some DSLRs also lose a bit of DR at their minimum EI, indicating that the value is a bit below the sensor's base ISO speed.)

3) A noise/grain/shadow handling based upper limit on EI. This is the one that people think of as the ISO speed of a film. It is also one that is hardly ever mentioned explicitly by digital camera makers!

Anon E. Mouse · « **Reply #11 on:** November 16, 2005, 06:59:10 pm »

I will refer you to ISO 12232, "Photography - Digital Still Cameras - Determination of exposure index, ISO speed ratings standard output sensitivity and recommended exposure index." That documment only list saturation and noise based ISO. BLJ, can you cite the other standards?

Where did you get you information about saturation-based ISO? ISO 12232 uses a different definition. which is simply the minimum exposure required for the maximum useful camera output. I don't think this can be compared to pull processing no more that any ISO can be compared to media response.

Film ISO is determined differently from digial ISO. Comparisons are hard to make. However, ISO is simply an exposure index based on media response and in that regard digital and film ISO is equivalent. Just like film ISO, noise-based ISO is not a single point. It is defined based on output and so is neither an upper or lower limit, but a sliding scale depending on the result that is to be acheived.

And manufactures do tell consumers about the results of noise-based ISO. I can't think of a single one which states that ISO has no impact on noise. Kodak gives quite a bit of detail. Try a search for "Kodak image sensors - ISO measurement." That is certainly more than a photographer needs.

maggieddd, the ISO standard does not include post-processing to determine the speed of the sensor. Like film, it is simply the minimum exposure to acheive a certain output. However, like film, this is hard to separate ISO from processing as it does affect the output.

BJL · « **Reply #12 on:** November 17, 2005, 05:06:27 pm »

Quote

I will refer you to ISO 12232, "Photography - Digital Still Cameras - Determination of exposure index, ISO speed ratings standard output sensitivity and recommended exposure index." That documment only list saturation and noise based ISO. BLJ, can you cite the other standards?

Where did you get you information about saturation-based ISO? ISO 12232 uses a different definition.[{POST_SNAPBACK}][/a]

ISO 12232 also defines the linear and logarithmic scales (the linear scale is the old ASA, the logarithmic is the old DIN), which defines the units of measurement used to state both those other sensor speed measures. This is the "ISO Exposure Index" I was talking about, defined in terms of the combinations of aperture ratio and f-stop that give normal exposure of the gray card (or whatever) used for reflective metering.
That Kodak document "Kodak Image Sensors ISO Measurements" you mention also briefly describes the ISO definition of Exposure Index with the formula "EI=10/H where H is the exposure in lux."

I have not seen the full ISO 12232 standard, but that Kodak document describes two variants of their saturation based sensor speed measurement (aka "base ISO"). Both are based on reflected light metering with an 18% gray card. The first more demanding variant requires handling highlights up to 170% reflectance, or 170/18 times brighter than the gray card reading (to allow for metering of something of reflectance as much as 2/3 stop less than 18%?) The second, suggested for snapshot photography, requires handling only up to 106% reflectance.

About manufacturers telling users about noise based ISO speed, I meant that they do not tell people such measurements for their cameras. Who tells you that "our ISO dial goes from 100 to 3200, but the noise based ISO maximum speed is only 1000"? Who even tells you that the saturation based ISO minimum speed is 200? (With some digicams, the noise based speed might be even be lower than the saturation based speed!)

That Kodak document gives no details about noise base (maximum) speed mesurements for any particular camera; it just briefly states the standards of "SNR=40 for an excellent image, SNR=10 for an acceptable one".

Do you know if those SNR measurements refer to mid-tones? I.e. meter light reflected from a gray card, photograph that gray card, and measure S/N from that? From data for the KAF-5100 sensor in the Olympus E-1, "excellent" would give roughly EI 400-800, and "acceptable" EI 1600-3200, which does loosely fit with the way that sensor is described by Kodak and Olympus.

Reference.

That Kodak document is linked as "ISO Measurements" at [a href=\"http://www.kodak.com/global/en/digital/ccd/publications/applicationNotes.jhtml]http://www.kodak.com/global/en/digital/ccd...tionNotes.jhtml[/url]

Anon E. Mouse · « **Reply #13 on:** November 17, 2005, 06:34:09 pm »

You say the ISO 12232 define both linear and log scales. On which page? I had a quick look though my copy and I can't see that.

Yes, the SNR measurements refer to 18% gray with a scene luminance of 160:1. The SRN=40 and SRN=10 are directly from ISO12232. But noise-based ISO is on a sliding scale. As sensitivity increases, so does noise. There is no maxium noise-based ISO. Just the limit where you think noise is too much - very subjective. The SRN=40 and SRN=10 is a benchmark; any value for the SNR can be used. A noise-based speed is always equal to greater than saturation-based speed as saturation-based speed effectively maximizes the SNR to the sensor response.

BJL · « **Reply #14 on:** November 18, 2005, 03:53:58 pm »

Quote

You say the ISO 12232 define both linear and log scales. On which page? I had a quick look though my copy and I can't see that.

You might be right about the ISO dropping the logarithmic (DIN) scale, which was there in ISO 5800:1987 for film speed, and which leads to speeds listed on film boxes as "ISO 100/21º" combining ASA 100 with DIN 21º. I was relying on a summary of ISO 12232 since I do not have the full text of ISO 12232, only the initial part that I can get for free from www.opsci.com.

Quote

There is no maxium noise-based ISO. Just the limit where you think noise is too much - very subjective. The SRN=40 and SRN=10 is a benchmark ...

Agreed, noise based ISO speed at least depends on subjective choices like SNR=40 or 10. It would be nice if at least cameras described as "professional" by their makers (see the recent thread on that slippery marketing-speak "professional"!) came with speed measurements at one SNR value, and I would vote for SNR=40. Kodak does seem to sometimes give SNR=40 Exposure Index limits for its sensors, if you dig around.

Quote

A noise-based speed is always equal to greater than saturation-based speed as saturation-based speed effectively maximizes the SNR to the sensor response.
[a href=\"index.php?act=findpost&pid=51555\"][{POST_SNAPBACK}][/a]

I was mostly joking about noise based speed lower than saturation based, but doesn't the measurement of saturation base (minimum) speed rely on a characterization of adequate highlight headroom? Kodak talks about needing to hold highlights up to 170% reflectance, meaning 170/18 times the metered level, but is that 170% a user selected figure, in the same way the SNR=40 is?

(If that 170% is part of the standard, the saturation based ISO Exposure Index could give SNR less than 40, and so strict noise based ISO EI would be less than saturation based ISO EI.)

Robert Spoecker · « **Reply #15 on:** November 18, 2005, 06:17:21 pm »

When I push my shuttedr button on my DSLR it closes an electrical circuit and from then on the internal computer circuitry controlls everything else that goes on, including among many other things, software selects the seting of the imaging pixel gain. Todays cameras depend so much on electronics (software) that there are murky lines between software and hardware controls. The answer is that many things depend on harware and software controll.

Robert

Author Topic: ISO on digital camera (Read 20586 times)