Using full electronic shutter, my Canon R7 camera shows up to 130 milliseconds variation in lag time between cabled remote shutter pulse and start of exposure. The pattern of this delay varies strongly depending on circumstances. In some situations the delay varies almost uniformly across the whole range; in other situations the delay usually falls in a narrow subrange but occasionally is much larger or smaller. The range of variation affects the settings that can be used safely for "mid curtain sync" using externally synchronized flashes. Exposure time of 1/4 second, with flash triggered 1/4 second after the shutter pulse, would be reliable in all the tests I ran.
My experiments confirm that the full electronic shutter travel time is about 30 milliseconds, moving from top of image to bottom of image with a constant speed so that a moving vertical line is captured as a slanted straight line. Interestingly, the mechanical and electronic 1st curtain shutters move in the opposite direction, from bottom of image to top of image, and the speed is not constant so that a moving vertical line is captured as a slanted curved line.
With the Canon R7, unlike earlier models, a flash trigger signal is generated when using electronic 1st curtain shutter (versus full electronic shutter). Thus there is no need for external flash synchronization unless full electronic shutter is used.
DETAILS
I recently purchased a Canon R7 camera to use as my main camera for focus stacking. One of its nice features is a full electronic shutter "silent mode" that is totally free of mechanical vibration and shutter wear.
Like many other cameras that use full electronic shutter, the R7 does not provide a trigger signal for electronic flash.
The standard way to address this lack is to use "mid curtain sync", in which the flash is triggered by an externally generated pulse that is somehow synchronized with the start of exposure. Common practice is to use a delay time that is fixed with respect to the shutter pulse that triggers the exposure, with no input from the camera itself.
Ideally the camera would have perfectly constant delay from trigger pulse to actual exposure. In that case the minimum required exposure time would be only a little more than the time needed for sensor readout, with the flash sync pulse scheduled to occur essentially one readout time after the start of exposure, just after the first shutter has progressively cleared the whole sensor and before readout has begun.
However, previous experience with a Canon T1i in EFSC mode made me suspicious that the R7 might have significant variation in shutter lag time.
So, I hauled out some equipment and tested it.
Indeed, the R7 does have significant variation in shutter lag time.
Further, the pattern of variation depends on circumstances in mysterious and intriguing ways. In some situations the delay varies almost uniformly over a range from roughly 70 to 200 milliseconds at center of image. In other situations the delay usually falls in a narrow subset of that range, but occasionally and unpredictably hits the extremes.
The major results are summarized in this illustration:
The instrumentation used here is essentially the same setup that I used in earlier tests.
It consists of an old-fashioned analog oscilloscope set up to sweep at 50 ms/div when triggered by the start of exposure pulse that is provided to the camera. The camera photographs the oscilloscope screen. While the exposure is happening, the camera records the current position of the scope’s bright dot, along with a dim history of the dot’s previous positions that is made visible by long-lived weak glow of the scope’s phosphors.
The timing of the start/end of exposure is indicated by the horizontal limits of the bright line.
For added context, the vertical axis of the scope trace is set up to show the flash trigger signal that is produced by an external synchronizer. The timing of the flash trigger is controlled by a setting on the synchronizer, and the overall goal is to find some combination of flash trigger timing and exposure duration that reliably places the flash trigger somewhere inside the exposure interval.
As shown in the pair of images above, exposure at the location of the bright line can start anywhere from about 70 to 200 ms following the exposure pulse.
To be more sure that I actually had found the limits, I shot 1000 exposures and ran them through a program to analyze the images. In addition to providing the max/min delays, that program also provided a mass of data that allows to look at the distribution of delays.
Here is the first set of data that I got:
For the most part, this graph showed what I expected to see: a random distribution of delays that uniformly filled the entire interval of 70-200 ms.
I was surprised to see the obvious patterning for the first 300 exposures. That clearly means something, but I have no idea what.
Nonetheless, the uniformity of the distribution is compelling. Here is the same data, sorted by increasing delay time. Aside from a few minor jiggles, I see a nice straight line that indicates a uniform distribution.
This run is with the camera USB-tethered to a Mac mini M1 computer, transferring small coarse JPEG files to minimize transfer time and file space. I repeated the exercise with different file formats and did not trip over any changes in behavior that seemed very interesting.
However, the delay time between exposures turned out have a huge effect on the results. The Run #1 results shown above were with about 1.4 seconds between exposures. When I increased that delay to be 5 seconds, the pattern got wildly different:
Sorting that data by delay time produced this summary:
I have no idea what causes this distribution to be so very far from uniform, and wildly different from the earlier runs.
I expect that investigating that aspect would be a fascinating diversion, but I’m pretty sure that I have better things to do with my time so I’m trying to stay disciplined and not start chasing that squirrel.
It is enough for me to know that 250 milliseconds exposure time, combined with 250 milliseconds flash delay, seems to reliably place the flash comfortably inside the exposure interval.
Of course, this test only measures the exposure delay at the center of the image. The delay at top and bottom of image will be different, depending on the direction and speed of how the shutter sweeps across the image.
To explore the direction and speed of shutter sweep, I changed the setup so that the dot on the oscilloscope was continuously moving quickly between the top and bottom of the frame, while simultaneously moving sideways across the frame. From the standpoint of the camera, the oscilloscope was displaying a thin vertical line that swept across the screen.
Photographing that line, with settings that are suitably shortened to show both the duration and relative timing of the shutter sweep, produced this picture:
Note that this is 5 milliseconds per division. The start of exposure is seen to vary by just under 30 milliseconds from top to bottom, beginning at the top, and the edges of the bar are essentially straight implying constant vertical speed of the shutter.
Running the same experiment with full mechanical shutter and with electronic 1st curtain shutter gives quite a different picture:
In these cases, the shutter sweep is from bottom of image to top of image, and is significantly nonlinear. In retrospect, the nonlinearity makes perfect sense in terms of a mechanical curtain that has to accelerate from a standing start not far outside the frame. But I had not thought about that before running the experiment, and I was surprised to see the obvious curve in the results.
Back to the full electronic shutter, there is one other quirk that I have read about but have not explored. That quirk, apparently shared with many other cameras, is that in full electronic shutter mode, fewer significant bits are captured in the raw data. Raw data in other modes is reported to be 14 bits, but in full electronic mode this is reduced to 12 bits, probably to keep the readout time down to 30 milliseconds. There’s a discussion thread at https://www.dpreview.com/forums/thread/4662859 , that has comments and graphs by Bill Claff of PhotonsToPhotos.
It's not clear to me how much I will care about the reduction in bit depth. The difference should be easily detected as coarser gradation in very dark areas, but those are rare in the deep high magnification stacks where I would particularly care about using full electronic mode. Anyway this is a matter for future investigation and I’ll probably put it off until I have a suitable subject already mounted up because I care about it.
That’s enough for now. Of course I wrote this for my own purposes, but I hope that some of you find it interesting also.
--Rik
