Why does a 15-second exposure on my camera actually last 16 seconds?

I was doing a test and wanted the shutter to stay open for 15 seconds -- no problem, I have a 15-second exposure setting in my camera. It seemed to stay open for about an extra second, though, so I started timing it. The 15-second setting takes about 16.1 seconds (+/- about 0.2 seconds or whatever my measurement accuracy is) between the first and second set of sounds the mirror makes. I also timed the 4-second setting, and it takes exactly 4.0 seconds (+/- my measurement accuracy); I measured the 30-second setting at 31.9 seconds.

For the sake of this question, I'm assuming that the 15-second exposure is taking exactly 16 seconds and the 30-second exposure is taking exactly 32 seconds, while the 4-second exposure is taking exactly 4 seconds.

I understand why 16 seconds would be more desirable (and probably easier to implement in software) than 15 (more-precisely one stop longer than 8 seconds), and I also realize that for a 15-second exposure, 1 extra second is a fraction of a third of a stop, and I probably wouldn't be able to see the difference between two otherwise-identical pictures taken at 15 and 16 seconds. But why is the setting called "15" when it's actually 16 (and it's relatively easy to measure that delta)?

Is this common among cameras, or unique to my camera's brand (Canon) or model (30D)?

Is there a spec somewhere that calls this out?

Should I be trying to measure more shutter speeds (in the fractions-of-a-second range) to see if my camera is mis-behaving?

If this is intentional, do the shorter shutter speeds that aren't exact halves/doubles also have different shutter speeds than they claim (e.g. 1/60 - 1/125)?

-
In the spirit of a baker's dozen, I propose calling 64 seconds a "photographer's minute." – coneslayer Mar 25 '12 at 17:57
@coneslayer - and 976,562.5(000...) nanoseconds* a photographers millisecond :-). [Also a computer engineer's]. (* 976,562,500 picoseconds exactly) – Russell McMahon Mar 25 '12 at 18:37
+1 for a fascinating question - well spotted! – Mark Whitaker Mar 26 '12 at 7:29

In "classic" cameras systems exposures varied by a factor of two between adjustment steps and by a factor of 2 with standard aperture number changes. The aperture f numbers usually provided vary by a factor of square root of 2 as the aperture is proportional to the square of the diameter and stop numbers relate to the diameter.
ie aperture is an area measure which is proportional to diameter squared. Yes, that hopefully makes sense if you read it slowly a few times :-).

The important thing is that a "classic" range of exposure times would be, starting at 1 second:

• 1 2 4 8 16 32 64 128 ...

The question is more why it would be labelled 15.

They may well be wanting to match the times to fractions of a minute so eg

• 1 2 4 8 15 30 60 120 ...

BUT leaving theactual values in a true power of 2 progression from 1 second.

However, going to faster steps starting at 1 second we also "run into problems".

1 1/2 1/4 1/8 1/16 1/32 1/64 1/128 is OK in fractional form BUT when written as decimal we get

1 0.5 0.25 0.125 0.0625 0.03125 0.015... ... you get the idea. Gets messy.

SO along the way people cheated slightly. The series may end up slightly like

1 1/2 1/4 1/8 1/16 1/30 1/60 1/125 1/250 1/500 1/1000 1/2000 1/4000 1/8000 ...

The decimal values get untidy around 1/16th to 1/125th but get tidy again below 1.250th and people largely don't use these.

The slight errors introduced are inside the margin of error in all but the most tightly controlled lighting situations and well inside what any eye-brain can guarantee they have seen (even if the eye can actually resolve it).

The first few differences are:

1/32 : 1/30 = 6.7% longer 1/64 : 1/60 = +6.6% longer 1/128 : 1/125 = +2.4% 1/256 : 1/250 = +2.4% at all settings above here

Your 15 seconds shown, 16 seconds actual is 6.7% longer than shown.

BUT many modern cameras blow all this away with shutter speeds in 0.3 EV or other steps that seem good to them. Worse, auto ISO or shutter speed or aperture systems may choose semi random settings so you MAY see eg 1/325th of a second. Very hard on classical system sensibilities ;-).

-
To put this in perspective, remember that one stop is 100% longer. The little differences just don't have a practical impact in most non-technical use. – mattdm Mar 25 '12 at 15:02
For this same reason, things like "ISO 25,600" are silly. Just call it "ISO 25k" and be done. – mattdm Mar 25 '12 at 15:03
Thanks for doing the math for me :) – drewbenn Mar 28 '12 at 5:43

The shutter speeds should be....

1s, 2s, 4s, 8s, 16s, 32s

Not sure why 8s and 32s are "rounded." If they're not even multiples of two, they would cause variations in exposure.

I timed 15s and 30s settings on a camera (Canon 1D4), and they are indeed, 16s and 32s.

I have 1/3 stop settings, too. They're labeled as 20s and 25s.

16s * 1.26 = 20.1 s

16s * 1.26 * 1.26 = 25.4 s

16s * 1.26 * 1.26 * 12.6 = 32s

My 20s exposure is right around 20.1s to the nearest accuracy I can get with a hand timer. :) My 25s exposure is right at 25.4s, too.

Going the other way, I would assume the actual shutter speeds are also powers of 2:

1, 1/2, 1/4, 1/8, 1/16, 1/32, 1/64, 1/128, 1/256, 1/512, 1/1024, 1/2048, 1/4096

From a pragmatic standpoint, I can see why people said "one five hundredth" instead of "one five hundred twelfth." Besides, for earlier cameras, it could have been anywhere from 1/400 to 1/600, and that would have been considered to be extremely accurate.

-

Keep in mind lag times. Every camera has lag times for mirror blackout, metering, af, and shutter activation. Your timing between mirror flip to mirror flip needs to account for those lag times, which will mostly occur before the shutter opens.

In the case of the 30D, here are your lag times:

Mirror Blackout: 110ms

Metering/AF: 250ms (shutter lag; full AF with full shutter button press)
Metering/Prefocus: 68ms (shutter lag; prefocus AF with partial shutter button press)
Metering/AIServo: 105ms (shutter lag; continuous AF with AI Servo between frames)

Cycle time: 270ms
Cycle time w/ continuous hi: 200ms
Cycle time w/ continuous lo: 330ms

Between the mirror blackout time, normal metering/AF, and shutter cycle time, you have @400ms of lag time for every frame. That is assuming AF does not need to hunt for any lengthy period of time...if it did, total lag time could be as long as a few seconds. Metering performance is best in good light, like AF, and both will drop in performance as light levels drop.

You can reduce that by prefocusing, however that is a manual action, and you still have to fully depress the shutter button to actually take the shot, so human reaction time will usually cause pre-focus lag time to be worse. If you make use of AI Servo continuous AF mode, and hold down the shutter button, lag time is reduced as the AF system is constantly active and predicting where your subject will be next.

Finally, if you throw flash into the mix, your total lag time is limited by the flash cycle time, which can be as high as several seconds.

On average, I would say there is anywhere from half a second to two thirds of a second of lag time for any given shot in single-shot AF mode. Add to that some 200ms margin of error, and you have anywhere from 700ms to 1s of difference between total actuation time and actual exposure time, depending on the circumstances. If you use AI Servo (which reduces the precision of inter-frame metering and precomputes AF at all times), your initial activation time would be less, around 310ms, and inter-frame time would have to be around 200ms to allow for the maximum 5fps frame rate.

-
Thanks for listing all those delays. At least in this case, I was using a manual focus lens (so no AF delays) and I think I was in Manual, so hopefully very little time spent metering. But when I'm out in the field those are numbers I've never thought about, but should because they do affect me when I'm trying to get an action shot. – drewbenn Mar 28 '12 at 5:48