Learning about (your) camera(s) in text and video
September 13, 2013 9:00 AM   Subscribe

If you wanted to understand how cameras work, you could spend some time with your camera manual. If it's not handy (or not helpful), you might opt to read through Wikipedia pages about film speed, ISO and digital equivalents, F-numbers (f-stop* or relative aperture), and depth of field (DOF). If you prefer, you could read "tedious explanations" of f-stop and depth of field and other photographic topics from Matthew Coles. Or you can spend 45 minutes watching 3 videos from YouTube user Dylan Bennett, as he explains ISO, F-stop and depth of field. Then you can join Joe Brady for a few hour-long sessions on setting up your camera for Portrait Photography with Ambient Light and Landscape Photography, and mastering exposure for landscape photography posted by filthy light thief (30 comments total) 152 users marked this as a favorite
 
And then with a few lessons under your belt, spend awhile with the photography teacher nobody wants.
posted by cribcage at 9:29 AM on September 13, 2013 [3 favorites]


Michael Reichmann's material (here and elsewhere on LL) is extremely good.

I am currently somewhere on the right side of this, though I don't know where.
posted by dickasso at 9:44 AM on September 13, 2013 [1 favorite]


We're rapidly reaching the point where "film camera" has as much relevance as "typewriter". And we're getting stuck with things like "ISO", which has as much relevance nowadays as the layout of the QWERTY keyboard -- historical artifacts which no longer make any sense, but which continue because of network effect.
posted by Chocolate Pickle at 9:44 AM on September 13, 2013 [1 favorite]


Civil_Disobedient had two amazing AskMe answers in an old, old question about lenses and f-stops here and here. Worth a read if you're interested in this stuff.
posted by Blazecock Pileon at 9:47 AM on September 13, 2013 [3 favorites]


we're getting stuck with things like "ISO"

Earlier this summer, a third-party camera software called Magic Lantern was updated with an ISO hack that "samples half of the sensor at ISO 100 and the other half at ISO 1600" to "get almost the entire dynamic range the sensor is capable of (around 14 stops)." It's a pretty amazing hack, and I have to assume it's being looked closely by camera manufacturers. Not to be overly optimistic, and surely it'll take a couple more years at least, but it's possible our old conceptions of ISO are on their way out the door.
posted by cribcage at 9:54 AM on September 13, 2013 [7 favorites]


Chocolate Pickle, why isn't ISO still relevant? While I'm sure a new camera's ISO 800 has less noise then my d70's ISO 200 and may not have any benefit at a lower number, it is surely still useful going the other way for higher sensitivity (ie ISO 25600 on that new camera will give you a properly exposed picture, but it will be much noisier than if you were able to take it at ISO 800).
posted by any portmanteau in a storm at 10:00 AM on September 13, 2013


ISO in recent digital sensors is largely irrelevant anyway, you can capture what looks like a totally black image in raw at ISO 80 then boost the exposure in PP and get back virtually the same image you'd have got if you shot at ISO 1600.

At very high ISOs there is some special magic going on beyond that but you can get pretty extreme dynamic range in a single shot already just by shooting raw with the right sensor and underexposing.
posted by dickasso at 10:02 AM on September 13, 2013


Never forget: the camera on the back of your phone that you have with you, is better than the expensive dSLR setup you left at home. Every time.

I laugh at reviews that dismiss 8+ megapixel phone cameras nowdays; the 5MP camera in my Nexus 4 is a better camera for general purpose use than the 2.1MP dedicated digital camera I used as my "main" camera just a few years ago. Some of the best pictures I've ever taken came from a Sony Mavica FD50. It stored pictures on 3.5" floppy disk and had a maximum resolution of 640x480.

My most famous photo (so far) was a lucky shot with a Canon S3 IS point-and-shoot. It's been featured as a magazine cover, used on numerous websites, and people have bought prints to frame. I still consider myself a rank amateur.

My "nicest" camera is a ten-year-old Canon EOS 300D / Digital Rebel dSLR. "only" six megapixels; I have a couple different lenses for it and so forth. The entire setup costs less than what some friends have spent on single lenses. However, it works for me. I haul out the backpack with the dSLR and gear for "important" events; 99% of the time it stays at home and I use the camera on my phone.
posted by mrbill at 10:05 AM on September 13, 2013 [1 favorite]


The reason ISO is still relevant is because no matter what you're recording on, it's very useful to have an international standard measurement of how sensitive to light that medium is. 'ISO' is an acronym for International Standards Organization, which specifies the measurement process. It's just as relevant today as an f-stop. Not sure why you'd think it should have died.
posted by echo target at 10:57 AM on September 13, 2013 [2 favorites]


you can capture what looks like a totally black image in raw at ISO 80 then boost the exposure in PP and get back virtually the same image you'd have got if you shot at ISO 1600.

The under-exposed image will be noisier than the higher ISO one. Here are some test images I found that illustrate this.

I think basically what's happening is that there is that every step of the analog signal path up to and including the digitization introduces some amount of noise. If you boost the signal midway through the process (that's what increasing the ISO does) you only are boosting the earlier noise. But if you wait until it's all digitized and then multiply everything by 16 in photoshop, all of the noise you could possibly capture is fully amplified.
posted by aubilenon at 11:24 AM on September 13, 2013


I can't find them, but I did a bunch of tests on "shoot at ISO1600" vs. "shoot at ISO200 and boost in post". This was a Sony A700, but the results were practically identical. Obviously YMMV, but generally I'm much more concerned with the shutter speed being correct than the ISO being correct.

It is much more difficult (mostly impossible) to de-blur a picture in post.
posted by lattiboy at 11:35 AM on September 13, 2013


Not to mention, worrying about shooting high-ISO is kind of ridiculous if you have a good camera made in the last 3 years or so. ISO6400 on my NEX-6 is incredibly good. With an automated de-noise pre-set in Lightroom it's almost impossible to tell it from ISO200 at any reasonable size. Perhaps if you're viewing 1:1 or printing a 13x19 and looking at it from a foot away. Which doesn't happen a lot...
posted by lattiboy at 11:37 AM on September 13, 2013


In addition to noise, it's also generally true that your camera's sensor is capturing more and better information toward the right side of the histogram than the left. It's true you can recover detail from underexposured areas (apparent "blacks") that you cannot from blown-out highlights—and for sure, there are artistic reasons to underexpose sometimes. But I'd caution against developing a shooting technique that relies on underexposing.

Of course, as Lattiboy notes, mostly these cameras are awesome, Lightroom and Photoshop are awesome, Snapseed is awesome; and most of us are looking at not-awesome JPEGs on not-awesome websites. Nobody notices grain on Twitter.
posted by cribcage at 11:43 AM on September 13, 2013


3rd-ing mrbill: Of the few pictures I actually sold, two (1,2) were shot with a point and shoot while my $4000 Sony A900 setup was sitting at home.

I'd love to get the Nokia 1020, but it's AT&T only at the moment and still a WP8. The files I've played with from that thing are pretty astonishing considering it isn't a dedicated camera.

Also, I posted my breakdown of what camera most people should buy in AskMe, which might be relevant. Short answer: The RX100 (or the RX100 II) are just the best all-around camera(s) made right now. Nothing really comes close to the mix of features and size and built quality. Weird thing is, it's been out for over a year and no competitor has come up with a challenge to it.

I'm of the mind interchangeable lens cameras are kind of silly these days for most people. Even the smallest EVIL camera is still much larger than the likes of the RX100, and the difference in IQ in all but the most extreme scenarios is negligible. Not to mention, a kit lens that starts at f/1.8 isn't practical on a larger sensor camera.
posted by lattiboy at 11:53 AM on September 13, 2013 [2 favorites]


any portmanteau in a storm: "Chocolate Pickle, why isn't ISO still relevant? While I'm sure a new camera's ISO 800 has less noise then my d70's ISO 200 and may not have any benefit at a lower number, it is surely still useful going the other way for higher sensitivity (ie ISO 25600 on that new camera will give you a properly exposed picture, but it will be much noisier than if you were able to take it at ISO 800)."

Because, as an older & wiser engineer taught me, since we can electrically amplify the signal as high as we want (by changes in the detector chip - I know "we" can't do this, but follow along), we "set" the ISO wherever we want on any camera. Except: it won't mean better pictures, because all the noise will be amplified as well. What we're really doing is "pushing the film" - only we're using op-amps instead of changing the time the film spends in the developer bath.

However, as every old-school film developer knew, this carries the price of increasingly noisy photos. In digital cameras, the highest "ISO" rating you can set (which is really the amplification level of the signal) is typically set where noise is just beginning to be annoying (otherwise, the camera gets reviews that refers to its upper ISO levels as "unacceptably noisy").

The real metric we're interested in maximizing is SNR (Signal-to-Noise Ratio) expressed at the limits of grayscale resolution. If the first grayscale bit value of the camera (at its current settings) is 15 eV (roughly 3-4 photons), and the background noise of the system has a practical typical minimum of about 3 eV, the SNR is 5.0. If the intrinsic system noise is improved by a factor of 2 (1.5 eV), the SNR doubles to 10.0.

An SNR of 5.0 corresponds to some value of ISO, but ISO has no intrinsic meaning, except by translation back into SNR. I don't know the conversion - it was incredibly painful for us to derive it back then - but suppose an SNR of 5 = ISO 200. Then for the second case above, SNR=10 yields ISO 400: twice the sensitivity above the noise. Not because we're "pushing" the image, but literally twice the ISO at the same noise level.

TL/DR: ISO is set by making the digital image lighter or darker, but the noise level changes right along with this process. We don't just want more signal; we want more signal without more noise. Then we can truly take a good photograph in lower light.
posted by IAmBroom at 12:08 PM on September 13, 2013 [2 favorites]


ISO is set by making the digital image lighter or darker

I was about to say "no no no, that's wrong", but then I did a bit more reading, it turns out that this varies from camera to camera. Canon's ISO adjustments are an analog process but Nikon's are not. Which is why CHDK's dual-iso trick does anything at all.
posted by aubilenon at 12:52 PM on September 13, 2013 [1 favorite]


What do you guys think about the mirrorless interchangeable lens cameras, or compact system cameras, or whatever people are calling 'em? I was looking pretty seriously at that category recently and wondering if it'll hit that sweet spot of a compact size with DSLR-like features and quality.
posted by naju at 1:27 PM on September 13, 2013


Of the few pictures I actually sold, two (1,2) were shot with a point and shoot while my $4000 Sony A900 setup was sitting at home.

Whenever I hear this, one of my first thoughts is always, "Why invest $4,000 to leave your camera at home?" Of course that's shallow and simplistic, and there are many, many answers, but still. The first thing I did after buying a full-frame DSLR was to buy a small camera bag for it. It's no bulkier than a woman's purse, and I can toss my wallet and cell phone inside. I carry it often. You hear people vaunt the photo they took while their DSLR was at home, and I always think, "Now imagine the photo you'd have taken if you'd brought it with you!"

What do you guys think about the mirrorless interchangeable lens cameras

I thought they'd replace the lower end of DSLRs. Mirrorless cameras are terrific, and mostly I see no reason to lug around a Canon Rebel if you can have a smaller body that takes faster shots of equal quality. I expected the entire mid field would be supplanted by mirrorless options (Sony NEX, Nikon 1, Fuji X100, etc), and reflex technology would be limited to the highest-end cameras. Recent sales figures indicate I'm wrong. Nikon is pulling back. Maybe this is due to unrealistic marketing strategy that was hoping to stop the mobile-photograpy bleeding, or maybe I'm just a big wrong dummy. I think mirrorless does hit that sweet spot, as you say, and if you're just looking for an option that will be perfect for a couple years, go mirrorless. Full disclosure, I don't plan to buy one.
posted by cribcage at 1:51 PM on September 13, 2013 [2 favorites]


I've got an Olympus EM-5 mirrorless that I like. It hits the sweet spot you're talking about for me, though of course everyone's idea of that spot is a little different. It's as good as a mid-level DSLR and beats the entry-level ones. The Micro Four Thirds system has the best selection of lenses right now, and since it's an open standard you can get the body from either Panasonic or Olympus. Sony's NEX system has good bodies and is starting to catch up with lenses. Fuji's X system doesn't have many lenses, but the ones they do have are high-quality and expensive. The Fuji X-Pro is probably the best mirrorless camera out there. The Nikon and Canon systems are basically me-too. Oh, and Samsung's got a system too; don't know much about it.

There are a few areas that no mirrorless camera can cover well. If you do a lot of sports or fast-moving wildlife, you're going to need a big DSLR with a billion phase-detect focus points and a predictive processor. If you do weddings and need soft creamy dreamy out-of-focus areas, there's nothing like the Canon 85 1.2 on a full-frame body. If you make gigantic detailed prints of landscapes or fashion models, get a Nikon D800e or bite the bullet and go to medium format.

For pretty much any non-professional use, and possibly for some professional uses, mirrorless is the way to go, I think. I got tired of lugging around a DSLR. Note that I use the term 'professional' to mean not 'high quality,' but 'this is my job, my only job, and I don't mind putting 10 grand or more into equipment.'
posted by echo target at 1:54 PM on September 13, 2013 [3 favorites]


When you are done comparing gear and pixel-peeping, there's this I made for you a while back.
posted by pjern at 2:06 PM on September 13, 2013 [1 favorite]


And then with a few lessons under your belt, spend awhile with the photography teacher nobody wants.

Was expecting a link to Ken Rockwell, but got a thoughtful essay instead. This post and thread are full of good stuff.
posted by TedW at 3:37 PM on September 13, 2013 [2 favorites]


What do you guys think about the mirrorless interchangeable lens cameras, or compact system cameras, or whatever people are calling 'em? I was looking pretty seriously at that category recently and wondering if it'll hit that sweet spot of a compact size with DSLR-like features and quality.

Roger Cicala (lensrentals.com blog) seems to think the 4/3 system is on the way out. Thom Hogan observed that mirrorless systems appear to have failed in the U.S. market and are only doing OK in the Asian market - that may be a marketing/pricing/positioning issue that could be overcome, but who knows.

I like the system. I recently bought the m4/3 Panasonic GH3 for video, but I've been playing around with the photo side of it, and it's a pretty darn good stills camera. One thing that's nice about the m4/3 is that it will take almost any lens with an adapter. Native m4/3 lenses are pretty decent, including zooms (I have the panny 12-35 f/2.8 and 35-100), though somewhat pricy, but you can use Nikon, Canon, PL whatever mount lenses with the appropriate adapter so that really expands your options tremendously. Now, with adapters you may not always have AF or stabilization or aperture control etc., so it'll be pretty much a manual affair, but some of the more expensive adapters do have electronic controls so some of these functions will work on those. The dumb adapters can be quite cheap - $15-$30 range.

An issue with some of these - like m4/3 - is that it's harder to get a good selection of wider lenses (basically it doubles the FF number), but there's a very exciting development in the Metabones Speed Booster adapters, which turn full frame, APS-C and potentially any larger than m4/3 lenses into wider than you'd otherwise get with just a straight adapter, plus, you can gain a stop or so into the bargain!

So I feel like this is a tremendous system at the moment. However, keep in mind that it may not survive going forward.
posted by VikingSword at 4:50 PM on September 13, 2013


An SNR of 5.0 corresponds to some value of ISO, but ISO has no intrinsic meaning, except by translation back into SNR.

That's not true, at least per sensor - with my first DSLR I had a good idea of what ISO I was going to need to get the exposure I wanted with particular aperture and shutter settings, ie, it does correlate to exposure. If ISO became SNR you'd (at least partially) normalize noise and sharpness across the camera world at the expense of losing the same (again partial) normalization we currently have with how ISO relates to exposure.

That said with current signal processing tech ISO should be a close to obsolete concept and we shouldn't have to bother with setting ISO/SNR manually even in professional settings. I'd like to see exposure options as follows:

1) No clipping allowed at the low end, high end, or either end, or x% of the image can be clipped at the low end, high end, or both ends. This is easy to do - my OM-D already does a real time clipping map on screen if you want it, which I love, and I think a lot of cameras/phones have this now, so extending this to an exposure rule is pretty easy.

2) a minimum x% of the range has to be used, with some cleverness to deal with low contrast situations.

3) an ISO cap where ISO is otherwise automatic (already a common option).

4) Priority rules for what setting gets modified first when your settings won't work with the scene.

As for the mirrorless cameras, I love mine. The only downside is the big DOF which is pretty disappointing - I did a portrait shoot with a friend a mine a few weeks ago and even at f/1.8 backgrounds 100 feet away were not particularly blurry. On a DX or APS-C sensor the same background would be so blurry as to be almost undiscernable.
posted by MillMan at 4:57 PM on September 13, 2013


yes on canon cameras the ISO is an analog gain, up to a certain point. after that point the "extended" ISOs are just digital multiplication of the data.

if you spend any time doing astrophotography you'll learn a few things. one of them is that every sensor has a certain amount of "read noise" associated with it, which is really the noise that the readout amplifier injects into the signal. the analog gain of that readout amp is what sets the ISO in a canon camera. so if you shoot a very dim target at a low ISO, what you will find is that all of your signal was crammed way down into the low end of the histogram, and it was destroyed/overwritten by the readout process. if you pump the ISO a bit, the signal will be above the read noise and so the signal is not destroyed by the read noise of the amp. in return you lose dynamic range, but lower dynamic range is better than having no image at all.

you can try this yourself by taking short exposures of a night scene at increasing ISO. you'll find that at the low ISOs the images are unacceptably noisy when stretched in photoshop or whatever. as the ISO increases the images are still noisy, but there's an image there.

for astrophotography with a DSLR, we try to find the ISO where one electron = one data number, that is, if you read out 'b00_1000_0000_0000 with N electrons in the well, then you read out 'b00_1000_0000_0001 with N+1 electrons in the well. this is called "unity gain" and below that you are actually wasting electrons (photons). in practice the unity gain is sometimes a ISO that does not exist, so you just go to the closest one.

this unity gain ISO can be anywhere from ISO400 to ISO1600 depending on the particular sensor.

it is true that the readout amplifiers also amplify the "dark signal" (that is the signal caused by heating of the sensor) but most of the noise you see in a unity gain-ISO picture is photon shot noise, simply because the exposures are short. photon shot noise is an artifact of the rules of the universe we live in and there's not much we can do about it, other than averaging a lot of different exposures together.

of course everything that IAmBroom writes above is correct as well - the high ISO performance in modern cameras is due to advances in the noise floor/sensitivity of the readout amps in these sensors. it's a good thing, too, because pixels have been getting smaller and smaller, each holding fewer electrons and making the A to D conversion problem just that much harder. advances in the silicon design have let us get closer to having shot-noise limited high-iso images rather than amplifier-noise limited images.

so anyway, TL;DR having some user control over the readout amp gain (ISO) is useful for overcoming some of the inherent problems of CCD/CMOS sensors in low-light situations.
posted by joeblough at 5:58 PM on September 13, 2013 [2 favorites]


What do you guys think about the mirrorless interchangeable lens cameras

I use them because they're cheap and no one takes you seriously when you pull one out - doesn't fit the stereotype of the paparazzi - therefore you can have free rein to run around and snap what you want - however for some jobs you need to put on a bit of a show and pull the big guns out - if only to live up to that stereotype.

Iso and noise and all that are fun to think about in the downtimes but I'm for accepting its part of the medium - but equally it depends on the job/project/idea - printed landscapes are always going to be a high end noise free area for example - i have a lovely sigma dp1 merrill i'm trying out just now and its very compact and high quality - as good as a nikon d800, but a pain in the arse to use.
posted by sgt.serenity at 2:57 PM on September 14, 2013 [1 favorite]


Since we're talking about ISO so much i will add that the new Kodak Portra 400 is remarkable pushed to 3200, it's fantastic for general purpose as well. So if you're still shooting film, check it out.
posted by lawrencium at 1:08 PM on September 15, 2013


I had a Canon S95, and it was a great camera, but I felt like I wanted more manual control and better build quality, so I ended up with a micro four thirds (mirrorless) Panasonic Gx1 and 14mm lens for about $250. The problem is that it's a little too big for my pocket, so I lug it around in a larger camera case, making it almost as awkward as a DSLR.

I'm also living in the Middle East and you just can't take out a camera and take a picture of a woman wearing traditional attire, on the street. Because women move about in most public settings it makes street photography a little tricky. But I'm learning to use it and it has a little more to offer than the Canon S95, which is a notch or two below the Sony RX100 II but still a great point and shoot and pretty cheap on the used market.

Find the camera you want, then go to Flickr and look at photos from that camera, which will all be amazing, so then you'll feel justified in getting it! It's a foolproof methodology.
posted by mecran01 at 11:22 AM on September 17, 2013 [1 favorite]


aubilenon: "ISO is set by making the digital image lighter or darker

I was about to say "no no no, that's wrong", but then I did a bit more reading, it turns out that this varies from camera to camera. Canon's ISO adjustments are an analog process but Nikon's are not. Which is why CHDK's dual-iso trick does anything at all.
"

No, it's true even if the image is set digitally before ISO conversion.

There are many detectors out there that produce 14- or 16-bit grayscale (2^14 grayscale levels from pitch black to the threshold of pure-white), but the camera operator is only able to access 10- or 12-bits of grayscale. They throw away the bottom few bits of signal, to avoid the bad-review problem I mentioned above.

In my limited experience, this was more common than actually using all available bits of information... which is pretty galling when you're desperate to squeeze out the teensiest signal possible, and can tolerate false positive (noise). Sometime this "wastage" occurred on-chip, so that even the camera producer's best technicians & engineers couldn't access the bits.

This wastage represents a multiplicative change to the least grayscale levels (2x, 4x, or 8x).
posted by IAmBroom at 1:47 PM on September 23, 2013


No, it's true even if the image is set digitally before ISO conversion

I don't think I'm parsing this sentence right because it doesn't make any sense to me. Would you mind re-wording/expanding it?
posted by aubilenon at 2:02 PM on September 23, 2013


Yeah, that was poorly worded.

My former company needed a camera able to see very faint targets at very high speeds; essentially, we expected the the target "dot" to be fainter than the darkest gray setting above pure black in a dark sky. Therefore, ordinary cameras would simply show a pure-black field.

What we needed was the most sensitive detector available in the world, except we couldn't use liquid-nitrogen cooled detectors.

We discovered that many vendors had detectors that were actually far more sensitive than the output images seemed (2x to 4x), so that although they wouldn't output anything but pure black, the original signal from the detector pixel where the target was imaged would actually have a grayscale value of (say) 1, 2, or 3 out of the full 0-to-2^14 range.

However, before the signal even left the detector chip, the bottom two bits of grayscale were thrown away (too noisy).

This meant that the chip intrinsically "saw" at one sensitivity, digitally, but reported the image 4x darker (dropped the bottom two bits of grayscale). A gray value of 3 on the detector D/A itself (0000 0000 0000 11) would be output as pure black (0000 0000 0000, and the bottom two bits aren't output).

Is that any clearer?
posted by IAmBroom at 10:24 AM on September 24, 2013


« Older PPP vs 538   |   Every single possible colour Newer »


This thread has been archived and is closed to new comments