[AstroDave]

As this article explains, except for certain values (integer powers of two, and reciprocal integer powers - such as 1, 2, 4, 8, 1/2, 1/4, 1/8 seconds), the shutter speeds which we are familiar with are not quite correct.

The author refers specifically to Nikon, noting for instance, that a 30 second exposure is really 32 seconds.

With a really simple experiment - taking a "30" second exposure with my K-1 and timing it with my wrist watch - I get a similar result: a so-called 30 second exposure really is quite close to 32 seconds.

My simple search of the forum does not pop up any discussion of this, specifically with respect to Pentax.

Does anybody have good evidence to support these claims - or evidence to the contrary (i.e. real data from Pentax/Ricoh or elsewhere) for our Pentax cameras?

I am in the process of designing a digital circuit which will allow me to measure my K-1 and K-3 shutter speeds to about 1% accuracy (I already can do this to 2%) and will report results here when I have them. My system counts optical flashes from LEDs, recorded during the exposure, not the sometimes-hyped acoustic schemes using your computer audio input.

These differences are something to keep in mind if doing interval shooting with intervals of 15 or 30 seconds - not only do you have to allow for camera overhead such as writing to the SD card before starting the next exposure, but you should also keep in mind that the actual exposure is a second or two longer than you think!

[photoptimist]

Interesting!

And how does this affect half-stop and one-third stop settings?

Is 1.5 sec really 1.414 sec?

Is 20 sec really 15*√2=21.2 sec or is it 16*√2 = 22.6 sec?

I look forward to your results!

[ekaphoto]

Here is an article on the subject

Shutter-Speed-Tester for your iPhone! · Lomography

As we are all analog photographers, we have one thing in common: We all work mostly with old, second-hand cameras. But did you ever ask yourself, whether those cameras still expose your film right, or not ?
The most critical component in every camera is the shutter, because the shutter-speed determines how much light the film gets. Most cameras provide multiple shutter-speeds like 1/30s, 1/60s, 1/125s, 1/250s etc. On most older cameras, those shutter speeds are created and regulated with complicated spring mechanisms. This can be a problem, because after several years, the springs inside the camera can lose their tension, which results in wrong shutter speeds: The camera exposes too long, and your film gets overexposed.
But how can you know if your camera is exposing correctly ? How can you know whether your 1/125 is still a 1/125 and not more likely a 1/80 ?
With this app, you will know. It allows you to measure the actual shutter-speed of your camera and gives you the deviation from the target value.
How does this work ?
The App uses an acoustic measuring method. When you release the shutter of your camera, the shutter generates the well-known shutter-release-sound. But this sound is not just a sound, it contains a lot of information about the shutter ! With the App, you can record this sound. Then, the waveform of the sound is displayed on your iPhone. In this waveform, you will recognize two Peaks:.......................cont

[stevebrot]

I am in the process of designing a digital circuit which will allow me to measure my K-1 and K-3 shutter speeds to about 1% accuracy (I already can do this to 2%) and will report results here when I have them. My system counts optical flashes from LEDs, recorded during the exposure, not the sometimes-hyped acoustic schemes using your computer audio input.

At some point you might want to post how this is done.

A few things for all of us to consider for comments on this thread:

Exposure time:

Moment of leading curtain start until moment of trailing curtain start

Elapsed time for exposure sequence:

Shutter initiation latency +
Mirror up time +
Curtain movement time (not the same as exposure time) +
Mirror down/Shutter reset/Sensor data flush-reset (assumed to be near simultaneous)

I timed my K-3 for 30s and 15s and got 32s and 16s respectively. However, what I timed was mirror up sound to mirror down sound minus my reflex lag to start plus my reflex lag to stop. In regards to actual exposure time, I am taking on faith that the clock chip for the shutter is proper and properly invoked. In regards to timing interval shooting, the extra time is something that should be accounted for in calculating the interval and an actual test might be appropriate to determine interval adequacy. Of course, that is always a good idea.


Steve

* Intervals are typically timed to start the exposure sequence with some form of fail-over should the camera not be ready. While there is potential for missed interval, this approach safeguards against cumulative lag. Some Pentax bodies support the option to start at end of previous sequence. This results in equal intervals between exposures, but with uncertain absolute timing for the start of each exposure.

[AstroDave]

At some point you might want to post how this is done.

Imagine a row of, say, 110 LEDs that turn on one after the other at a rate of such that there should be 100 of them illuminated during the expected time that the shutter is open. When the count gets to the last LED, the sequence starts over again, so the row of lights cycles endlessly at this rate. You take a picture of the LED array. If your shutter speed is correct, you should see 100 of the LEDs lit up. It doesn't matter where they begin - just count the lights. If there is an extra light or two, then your shutter is open LONGER than you expected. A few less than 100, your speed is faster.

If the lights are indeed in a row (my current system has an 8x8 matrix of LEDs), across the frame, this works even for high shutter speeds where the aperture is a moving slit between the top and bottom curtains. Conversely, if you are far enough away from the matrix, you will still see all the relevant lights within the slit.

Conceptually simple, but getting all those LEDs to light up in sequence takes a rats nest of wires!! I've got the design ready, just need to wire it up.

Another somewhat easy way to do it is to hang a single flashing LED on a long enough rope and swing it across the field of view. Takes only one LED, but you need a longish rope. (Fortunately, I have a place inside my house where I can hang a 15' pendulum - maybe I'll try this way first. I did something similar along time ago and had forgotten about that technique.)

In this way, you really do measure how long the shutter exposes the CCD.

I'll post my results (and circuit(s)) when I'm done.

[slartibartfast01]

Interesting!

And how does this affect half-stop and one-third stop settings?

Is 1.5 sec really 1.414 sec?

Is 20 sec really 15*√2=21.2 sec or is it 16*√2 = 22.6 sec?

I look forward to your results!

I think your 1.414 multiplier is only relevant to aperture values. 1.5 seconds will be 1*1.5=1.5 and 20 seconds will be 16*1.3333=21.3333.

After further thought I realise I am talking BS. You are correct. 1.5 sec is really 1.414 sec although 20 sec is really 16 multiplied by the cube root of 2 giving 20.16 sec.

[photoptimist]

I did a quickie experiment by shooting a "30" second picture of my iPad's sweep second hand and the exposure time is most definitely 32 seconds on both the K-1 and K-5.

Here's a crop from the image from the K-1:

Attached Images

 

[alpheios]

Thanks AstroDave and photoptimist for your findings. I finally have an answer to why I had to set the interval on my K-1 to 33 seconds when shooting 100 thirty second frames in interval mode for a startrail composit. I do write RAW to both cards but without any in camera corrections and the cards are fast. I always wondered why the camera needed three seconds to be ready for the next picture.

[photoptimist]

If we can get accurate true exposure time measurements, we will probably find the following discrepancies between the "dial" setting and the "true" exposure time:

Controls set for 1/2 stop (or 1.41421 multiplier) changes:

Dial...True
_8......_8
10......11.3
15......16
20......22.6
30......32

Controls set for 1/3 stop (or 1.25992 multiplier) changes:

Dial...True
_8......_8
10......10.1
13......12.7
15......16
20......20.2
25......25.4
30......32


Of these, the most interesting may be the "20" dial setting which may either have a 22.6 second exposure or a 20.2 second exposure depending on whether the camera is set to measure & control exposures in 1/2 stop or 1/3 stop increments.

[Bob 256]

Very interesting. The issue probably relates to how the shutter timing system was engineered. It may be that it's based on powers of 2, so 32 seconds falls out of the equation more easily. The more important question is, does this variance cause an error in actual exposure or is it compensated for using aperture (and/or ISO)? If one were to use assumed exposure information for a qualitative analysis of motion (e.g. judging the speed of a car from the blur length it produced in a photo), it would definitely be an issue (police analysis take note), but if the camera is designed for the correct exposure (compensating for slightly odd shutter times) it's much less an issue (not that it isn't interesting).

It's a bit like old zoom lenses which once focused, had to be spot-on throughout their zoom range so refocusing wasn't needed. With autofocus, that became less of a requirement since the lens is focused prior to each exposure. In this case, if the camera is making an overall correct exposure call (shutter speed/f-stop/ISO), it probably isn't important that the shutter speed is exactly 30 seconds so the closest, best engineered value is used, and everything else is adjusted to fix the variance.

It will be really interesting to see AstroDave's measurements. Wonder if this affects short exposures (1/30, 1/125, etc) as well? Another interesting question - is a dialed in value of f5.6 really being shot at exactly f5.6?

[Ontarian50]

On my shutter tester, a truly accurate shutter measures like this...
1s - 1000ms
1/2 - 500ms
1/4 - 250ms
1/8 - 125ms
1/15 - 62.5
1/30 - 31.2ms
1/60 - 15.6ms
1/125 - 7.8ms
1/250 - 3.9ms
1/500 - 1.9ms
1/1000 - 1.0ms

Only a handful of quartz timed focal plane shutters have hit all the numbers across the range. Many seem to be designed for a slightly longer exposure at their X-sync speed. Mechanical shutters can be excellent (within say 10-20%) or be truly awful. Lots of older cameras do poorly at 1/8th second, and many have top speeds that "sag", often 1/500th is closer to 1/400th, and 1/1000 will barely get past 1/500th on the timer, no matter how well you lubricate and adjust them.

I remember testing a Nikon D70 with the sensor out of it, and found the shutter speeds nice and accurate right up to 1/125th, after which it never went any faster - which explains why the D70 had a 1/500th flash sync - Nikon were using electronic shutter at speeds above 1/125th.

[Not a Number]

According to the Pentax K100D service manual - shutter speed calibration is done at 1/4000th sec. via the histograms using a standard calibrated light source, light box and master lens.

When properly adjusted: "the shutter speed should be within 0.227mS ~ 0.30 mS (1/4000)."

[stevebrot]

I did a quickie experiment by shooting a "30" second picture of my iPad's sweep second hand and the exposure time is most definitely 32 seconds on both the K-1 and K-5.

So, the big question would be why Pentax (and Nikon) chose to not put the maximum time at the full-stop increment (32s) actually used instead of switching inexplicably to a five-based count at 15s. I am not very concerned about the fractional parts, but a difference of +10% is somewhat worse than timing "B" with a stopwatch. This makes no sense.

Does "B" timing on the K-1 suffer from a similar anomaly? If not, camera-timed "B" for exposures 10s and longer combined with a wired intervalometer might be an option if accurate exposure is paramount.


Steve

[photoptimist]

So, the big question would be why Pentax (and Nikon) chose to not put the maximum time at the full-stop increment (32s) actually used instead of switching inexplicably to a five-based count at 15s. I am not very concerned about the fractional parts, but a difference of +10% is somewhat worse than timing "B" with a stopwatch. This makes no sense.

Does "B" timing on the K-1 suffer from a similar anomaly? If not, camera-timed "B" for exposures 10s and longer combined with a wired intervalometer might be an option if accurate exposure is paramount.


Steve

I was thinking the same thing. When this thread started, I was extremely skeptical that the "30" setting would be 32 seconds and almost said that.

The easiest ways to program all the shutter speeds would be to either: 1) define a table of round-number shutter speeds that are displayed and then implemented exactly as shown (i.e. "30" really is 30 seconds); 2) use math to compute the exact value with full, half, or one third stop ratios and then display a rounded-values (5 full stops longer than 1 sec is 32 sec and displayed as "32"). But what they seem to be doing is computing the accurate shutter time but then using a table of different values for the user interface and EXIF value. It's weird!

The K-1's B timer was an add-on so maybe it's logic is different. It would be weird if in B mode, 30 was 32 sec, 1 minute was 64 sec, and 2 minutes was 128 sec given that B mode lets you pick 2'10" as a choice.

[skierd]

Other than the “it’s nice to know” part. if the exposure comes out the way you expect it why does this matter? Just curious. It doesn’t particularly bother me if shutter speeds are really a nominal speed than actual speed so long as the steps between are repeatable and consistent so I know what I’m getting in to when I take photos.