It’s been clear to anyone testing the Sony alpha 7R that its shutter is a different animal from what DSLR shooters are used to. I’ve been reporting on what that shutter does to pictures of test targets, of oscilloscopes, and of real subject matter. But what do I know about what goes on inside the camera when you press the shutter release? Not much. Time for a different perspective.
Mike Collette is a self-taught electronics engineer who invented the large-format digital scanning back in 1992, and founded Better Light, Inc. (www.betterlight.com) to develop and manufacture these high resolution digital capture devices. In addition to being a fine engineer, Mike has an unusual ability to explain complex technical issues to people who think a logarithm is a line dance. He’s also an accomplished photographer and a nice guy. Check out his work at www.betterlight.com/gallery.
While I’ve been using an oscilloscope to look at the effects of the Sony a7R’s shutter slap on images made with the camera, Mike’s been doing some great detective work on the causes of that shutter slap – or, more properly, as you’ll see, those shutter slaps.
I’ll let Mike tell you about it himself. In order to fully appreciate what he’s done, keep in mind that a loudspeaker contains an electric motor: it turns current into linear motion. Almost all electrical motors can be run backwards, and then they turn motion into current. When they are doing that, they’re called generators.
“Today I cannibalized an old computer external speaker to make a pseudo-accelerometer so I could measure the external vibration when the Sony a7R shutter is fired. Attached is a photo of the simple test setup. On the other end of the probe is a fancy digital oscilloscope that lets me record and expand whatever the speaker “felt”, and even transfer screen shots to my computer via USB cable.
“My test setup utilizes a high-tech cylindrical fiber tube (a rolled-up Post-It note) to transfer vibration from the camera body to the speaker’s voice coil (about the same diameter as the tube). The speaker voice coil is only responsive to vertical motion (in this setup), but horizontal motion of the camera body might “rock” the top end of the tube and cause some vertical displacement of the bottom end anyway. Since the focal plane shutter travels vertically in the Sony, I figure vertical body motion would be more important, and it’s also easier to “mount” the camera on its flat bottom than on either side.”
Here’s Mike’s setup:
Here’s the ‘scope image with Mike’s annotations showing what’s probably happening during each material segment of the waveform:
The annotated waveform image above is way too small to see in the blog window. However, if you click here, you can see it at whatever resolution your browser supports, up to 3085 pixels wide. If that’s not big enough, you can right-click this link and save it on your computer (Firefox and Chrome: “Save link as…”, Internet Explorer: “Save target as…”), you can look at it with any image program that can handle PNG files. Photoshop is one such program.
Mike has some further comments:
“I don’t know exactly when Live View actually ends, but I’m guessing it stops before or as the shutter starts closing. I don’t know when Live View actually starts either, but I’m pretty sure that this won’t be before the shutter is open again [OK, Mike, we get it].
“I also don’t know exactly when the shutter rewind/cocking actually begins; there is some vibration before the starting points I marked that could be related to this action, too.
“What the waveform does indicate is that the shutter doesn’t “slap” closed initially, but rather gets rewound closed. Note that there is a slight wiggle after the second shutter rewind/cocking (before the first curtain opens again), so the rewind action is creating some disturbance that might still justify having a longer delay before the exposure begins. But most of the disturbance is coming from the first and second curtains. There is an initial “recoil” (the downward swoop) as the curtain is accelerated into motion, and then a “slap” (the upward spike) when the curtain abruptly decelerates. Both of these forces impart energy that can contribute to camera shake, although only the initial recoil of the second curtain matters, since the exposure is over before the second curtain slaps.
“In fact, the one-two punch of the first curtain recoil (down) and slap (up) produces a nice little wiggle that could tickle any resonance in the camera (a mass) and its support (a spring, more often than we would like). The big disturbances in the waveform following each shutter motion are caused by the resonance of my test setup, which had the camera and 55mm FE lens sitting on big blocks of dense closed-cell foam — certainly not a typical mounting arrangement, but good for allowing the camera body to move slightly so the speaker could “feel” it better. At least the foam is fairly well-damped so the camera settles down within a few wiggles. These shutter-induced wiggles will be considerably different in a typical setup, especially with a long lens adding its mass to the camera, and perhaps some springiness as well if the lens mounting foot isn’t stiff enough. A dead weight as proposed by Joe Holmes also adds mass to the camera that can help de-tune any resonance in the support system, or exaggerate a resonance, if we’re not careful.
“At shorter exposure times, the second curtain recoil/slap moves closer to the first curtain recoil/slap, until at 1/160 second (the fastest flash sync speed for the a7R), the second curtain starts closing almost immediately after the first curtain finishes opening, leaving just enough time for a flash to fire while both curtains are completely open, superimposing the second curtain’s recoil (downward swoop) on the first curtain’s slap (upward spike). While this overlap might help reduce the effect of these overlapping opposing forces, the first curtain recoil is unaffected, and any shake from this action represents the majority of what the sensor sees during such “short” exposures. Even shorter exposures have the second curtain starting to close before the first curtain has finished opening, so the two curtains chase each other across the sensor leaving a variable-width slit that determines the effective exposure time. This also means that both curtain recoils start overlapping and perhaps exaggerating any camera shake as the exposure is ending.
“If the Sony has already opened its first curtain to allow Live View, the only way to close this would be to rewind/cock this curtain again (note that the second curtain is already open/cocked, so the first “rewind/cocking” interval only has to rewind/cock the first curtain). After the exposure completes, both curtains are rewound/cocked again, probably keeping the sensor in complete darkness during this operation. Finally, the first curtain is fired open to allow Live View again.
“I suspect that during multiple exposures, the second opening of the first curtain and the initial rewind/cocking are omitted between exposures, and the shutter only needs to be rewound/cocked during the image readout portion following each exposure.
“Rewinding/cocking these shutter curtains in just over a tenth of a second is probably being accomplished by a fast electric motor, and if the curtain actuating springs utilized bistable “flippers”, most of the rewinding/cocking time would be winding up the springs, and near the end of this interval the curtain(s) would “flip” to their cocked states and produce the high-frequency “chatter” that looks similar to the shutter slap signatures in the waveform.”
I have been agitating for a firmware change to provide delay between the closing of the shutter when the release is pressed, and its opening to take the picture. Based on what he sees with his test setup, Mike doesn’t think this is a fix for the camera’s shutter slap:
“I don’t think much will be gained by asking Sony to delay the start of exposure after the shutter closes, although this shouldn’t be detrimental in most applications, either. The best way to address this situation is by paying careful attention to the camera mounting/support system. Every adapter, screw, mounting plate, quick-release, clamp, bracket, rotary head, etc. needs to be considered as a potential source of instability/elasticity.”
He puts the a7R’s problems into perspective:
“This problem certainly isn’t new to digital photography, but as sensor resolution increases (and as camera bodies shrink), camera shake will continue to be a potential source of image degradation.
“As much as I prefer electronic solutions to just about everything, there isn’t much better than a simple torsional spring to store and release energy quickly and with minimal inertia. I’m also guessing that most focal plane shutters behave similarly — when rewound/cocked, the first curtain is closed and the second curtain is open, and when fired, the first curtain opens followed by the second curtain closing. Exposures shorter than the transit time of a curtain across the image plane are effected by releasing the second curtain before the first curtain has fully opened, so the two curtains form a variable-width slit that travels across the image plane. The fastest flash sync speed is determined by the transit time of a curtain across the image plane; both curtains have to be fully open when the flash fires, and then the second curtain can begin closing.”
Mike’s experiment, and more importantly, his interpretation of the results, have made what’s probably going on in the a7R’s shutter much clearer to me, and has suggested some further experiments on my part. Thanks, Mike.