The data in yesterday’s post was pretty noisy, but it seemed to indicate that somehow the silent shutter mode on the Sony alpha 7 R Mark II (a7RII) might produce images that were slightly less sharp than the electronic first curtain shutter (EFCS) mode. I foolishly speculated that the reduced bit depth of silent shutter mode might have something to do with it. I say foolishly because Horshack, on DPR, and Jack Hogan, in a comment on this blog, both pointed out an error in my thinking: at ISO 800, which I used for yesterday’s testing, the a7RII read noise would provide plenty of dither for an image with 12-bit precision.
The next step was to find out if the difference between EFCS and silent shutter was really there, or just a product of that sometimes wonderful and sometimes confounding human property of finding patterns whether they’re real or not.
Horshack helped be with the design of the experiment. In order to get the most sensitivity, we needed a sharp lens. I pulled out the sharpest one I have, the Zeiss Otus 85mm f/1.4 ZF.2. We wanted to minimize the read noise, which would provide dither. I set the ISO knob — well, virtual knob — to 100. We wanted a stable platform. I put the camera in landscape mode.
We wanted lots of images, for two reasons. First, so that we could compute the average results. Second, so that we could compute some statistics, and thus know if we’d used enough images. I picked 16 images per data point, which is a compromise between accuracy and avoiding terminal boredom on my part.
- The camera: the Sony a7RII.
- The lens: Zeiss Otus 85mm f/1.4 ZF.2
- No filter. I don’t have a variable ND filter big enough to fit the Otus 85.
- The lighting: a single Fotodiox LED-200WA-56 daylight balanced flood.
- ISO set to 100,
- Manual focusing wide open, then lens set to f/2.8
- Shutter set to 1/500.
- Drive set to single
- Self timer set to 5 seconds.
- Camera on an RRS TVC-43 with an Arca Swiss C1 head,
- RRS L bracket used for mounting the camera to the head.
- Make 16 exposures with the mechanical shutter
- Make 16 exposures with EFCS
- Make 16 exposures with silent shutter
- Set the lens to f/8 and the shutter to 1/60.
- Make another set of 48 exposures.
- Develop in Lightroom 5.7.1 with standard settings.
- Crop, export as TIFFs, analyze the horizontal edge. Measure MTF50 in Imatest .
- Export the results to Excel and graph.
The results for 1/500 second:
EFCS is on the left, the mechanical shutter is in the middle, and silent shutter is on the right. The blue bar goes to a value defined by the mean of all of the results less three standard deviations. The gray bar goes to a value defined by the mean of all of the results plus three standard deviations. If the distribution is Gaussian, only 3 out of 1000 values will lie outside that range.
The top of the orange bars represent the average values.
Note that the vertical axis, the units of which are cycles per picture height, are enlarged greatly so you can see the rather small differences.
Conclusion: silent shutter is no less sharp than EFCS.
Now, lets look at the 1/60 second results:
The first thing to notice is that, even with this very stable setup, the mechanical shutter results are much worse than the other two. Well, much worse when you’re graphing slanted edge MTF50 results. Most people would consider 1300 cycles/picture height to be a nice crisp image.
The second thing to see is that the EFCS and silent shutter results are virtually the same.
The third thing is that the MTF50 numbers are lower than at 1/500. This has nothing to do with shutter speed. The Otus isn’t nearly (again, “nearly” has to be understood in the context of a great lens tested under laboratory conditions) as sharp at f/8 as it is at f/2.8 (how many lenses can you say that about?).