Exposure basics, lesson two point one (& ISO)

Started Mar 19, 2013 | Discussions thread
bobn2 Forum Pro • Posts: 60,603
Re: Exposure basics, lesson two point one (& ISO)

texinwien wrote:

bobn2 wrote:

Firstly, good idea starting a thread on a sub-theme from the last discussion. This question of the effect of REI and multipattern metering is really an interesting one, and deserves to be discussed in more detail. With luck, a thread that can result in new knowledge.

As I say, worthwhile doing this thinking in the open, because maybe people will join in who have hard information on the topic, and we will all learn.

I fully expect to learn some new things in this thread - thanks for sharing your thoughts.

First, when you say 'dial in more exposure,' I take it to mean that the camera would lower the shutter speed. Assuming that's what you mean, is there any evidence that would suggest that the E-M5 does, indeed, do this?

That evidence would require a systematic investigation. If someone with an E-M5 wants to do it, then it would be most interesting to see the results. I would think you'd want to calibrate the E-M5's meter against a control meter of the integrating variety, and then go through a varsiety of scene types noting the exposures given by the control meter and the E-M5.

It has just occurred to me that one forum member who owns an E-M5 and has, apparently, been photographing at least semi-professionally for four or more decades shoots only in full Manual mode and meters based solely on his trusty light meter (think it's a sekonic). I wonder if there would be a way for him to generate a report based on his RAW files that would be of any use.

What you'd need to know is how the camera metered the scene. That might be captured in the EXIF. If course, he'd need to have kept his metering set to multi-segment.

Not systematic, but perhaps of interest? I'd run systematic tests, but Spring is here, and I will be outdoors as much as possible for the next 8 or 9 months.

Meters are cheap on ebay. The little extra time to meter some shots and note the meter, and you can combine the experiment with your days to day photography, after all, what you want is a variety of shots.

A note about my standard E-M5 usage practices - I always used spot-metering. I usually use aperture priority mode with ISO 200 in well-lit situations. I use either shutter priority mode or manual mode either with fixed ISO of my choice or Auto ISO limited to 1600. I also try to ETTR whenever I have the time to be deliberate. My subjective gut feeling is that I almost always have a good deal of room to add +EV - often well over a full EV.

Ah, then if you're not using pattern metering, that won't do.

Also, in your good example stratagem, the camera would be programmed to only lower the shutter speed when it recognized low light situations. With that in mind, I have a couple of questions:

What would you expect to happen in well-lit situations? I assume that you would expect to see 'correct' metering in these (correct based on the camera ISO setting).

Well, remember that the point of multi-segment metering is to find a 'better' exposure than you get by simply using an integrating meter. On the one hand, one might say that the optimum exposure (from the output brightness point of view) is that which you'd get from an incident light meter, so you'd think perhaps that a multi-pattern meter should try to compute and replicate the incident light meter result. But another way of looking at it is that it should just give a more 'pleasing' result. So far as I know, no manufacturer reveals the algorithms behind their multi-segment metering, so we don't know in practice. But I would think they might do things such as lowering the exposure in high contrast scenes to try to avoid blown highlights, increasing the exposure where there seems to be a preponderance of dark to avoid noisy shadows and so on. When you start to think how multi-pattern metering would work, it seems pretty natural to think it's going to give more exposure in dark scenes (such as night shots) and less in bright high contrast (snowscapes and beaches, etc) because that is what most photographers would do.

Just to be clear - is it your position that any 'cheating' on ISO tests that could possibly be done would have to be done through 'mis-metering', giving the camera more exposure than it should get, compared to other cameras, based on the scene?

I'm not sure it is 'cheating', just allowing what the system and standards allow. So far as running a high headroom, that is if anything a relative disadvantage for nominal exposure for that ISO, since it results in a read noise contribution relatively higher than it could be with respect to the recorded raw levels.

I'm not trying to pin you down here - I'm just trying to make it clear for myself and others, since you seem to be saying that the only way to 'cheat' here would be by manipulating the exposure, and that if such manipulation were occurring, it's something we should be able to detect by running systematic tests.

In the end, what we are talking about is how different evaluative metering systems evaluate different kinds of scene, and it's only really a problem for the photographer who hasn't got their exposure management sorted out. After all, generally the exposure in low light is set by the shutter speed you can bear, the f-number you have available and the light. The ISO should follow along, not be dictating the exposure.

Do you think most ISO test scenes used by DPReview, imaging-resource.com and other well-known testing outfits would count as low-light scenes in such a stratagem?

Mostly they look like mixed contrast scenes. Since DPReview, IR and the like don't use the camera metering for their tests (which invalidates the ISO rating, really) any meter effect would not affect them. However, I would not be at all surprised if the metering didn't get gamed for test scenes. There was a piece on the radio yesterday in which it was reported that motor manufactures are setting up their ECUs to recognise being driven to the statutory test procedures and lean out the engines to achieve apparently lower carbon emissions, wile in normal driving the emissions are unchanged.

It's good to be reminded that this kind of behavior occurs. As a software guy, it's interesting to try to imagine a way to program a camera to recognize when it's being used in this sort of test scene. The quickest, easiest way I could think of would be to have the camera recognize various test charts, which most of these shots contain.

With a little more work, you could get it to recognize the test scenes used by the most popular testing sites and agencies with high certainty - so, recognize with high certainty known test scenes and, as a fallback, recognize common test charts. If you see either, switch to 'multi-pattern metering for best ISO test results' mode.

Yup, the real problem is you soon start thinking along the lines 'how could I manipulate this system', then you find lots of opportunities. The other thing is also that most cameras don't seem to run an absolute constant ISO, that is there is also adaptivity in the in-camera processing. After all, why not do that? It just helps the photographer think they are managing exposure better, which will make the camera seem better. I first noticed this when I started using the simple open source raw conversion tools, and suddenly my processed raws started showing quite big brightness variations with respect to the in-camera JPEGs.

If we saw 'correct' metering in well-lit scenes (correct based on the camera ISO setting), we would see more noise in them than we would have seen had the E-M5's ISO settings been more closely aligned with its saturation sensitivity. Would you expect consumers not to notice this excess of noise in well-lit images?

Most modern cameras have an excess of SNR (for most photographers) in well lit situations. Generally, an advanced photographer will trade a bit of that SNR for highlight protection, and I would expect that is what multi-pattern metering will do to.

-- hide signature --


I asked this because I have seen multiple people complain about noise in daytime skies in photos taken with the E-M5. I ran some of my own quick and dirty tests here, and found that I could, in some cases, dial in a full +3EV of exposure compensation without blowing any highlights and cut way down on the appearance of noise in the skies.

I haven't thought about noisy skies, but I can imagine that there are some factors in play. Firstly, you have a very flat textureless area, which means what shot noise there is will be apparent. For that the higher the exposure the better. Then there is a relatively low red channel content, so red channel noise could cause perceptible noise. Same solution. The other thing is, just blur the blue sky. There is a technique I use to clean up skin which could work for sky. Make another layer, blur it. Give it an alpha channel based on the sky (skin) colour, invert the alpha channel and flatten.

-- hide signature --


Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow