April 19, 2014
Sony A7R Shutter Shake, Part Three: The Results of a Refined Solution
This page: http://www.josephholmes.com/news-sonya7rshuttershake-part3.html
Future URL of this page: http://www.josephholmes.com/news/2014/04/19/sony-a7r-shutter-shake-part-3
My prior two articles published here, on my site, about this topic (Part I, Part II) detailed the nature of the issue with this most promising and capable camera.
Essentially, the shutter shakes, perhaps somewhat more than most shutters, thanks to the live view without electronic first curtain nature of the camera, and the camera is much, much lighter than most, the net effect being roughly triple the shutter shake of a typical full-frame SLR. When tripod mounted, this shows up most when: 1) The camera is mounted vertically, 2) The lens is long, 3) The camera is not connected directly to the tripod head, and 4) The shutter speed happens to be one which coincides fairly well with the greatest motion of the sensor, as it's moved by the shock of the shutter's sharp motions.
The deadweight solution which I have devised is not used when the camera is clamped directly to the tripod head in horizontal position, nor is it needed in such cases. For all the other five connection categories (vertical with direct connection, horizontal and vertical with adapter foot connection, and horizontal and vertical with collared lens foot connection) it would be used, except where an adapter foot comes too close to the camera’s L-plate to allow the clamp of the deadweight to fit. I am guessing that the Metabones and Sony LA-EA3 and -4 adapter feet are too close, but a Mirex foot or an RJCamera foot are no problem. It might be possible to use a Metabones foot with an extender together with a deadweight clamped to the underside of the camera I don’t know. Obviously any collared lens foot would easily clear the deadweight clamp as they all sit further forward of the camera body than the feet of adapters like the Metabones or RJCamera.com smart EOS to NEX/FE adapters.
The deadweight also turns out to reduce hand-held camera shake a lot. Another approach to avoiding shake problems is to know which shutter speeds are problematic for a given setup and to use an 8X neutral density filter to slow down the exposure by three stops and thus avoid the danger zone in question. It would be best if you would make a long series of test exposures at speeds between perhaps 1/200th and 1/2, using a target which will show subtle movement in the direction of the shutter (left-right for a vertical), to find the danger zone for your setups.
Although it's taken a few months to get the necessary parts made to order, then subjected to additional machining, I now have a second-generation, better-specified deadweight fashioned for the camera, which clamps to the bottom of the L-plate, so that it can be used as described above.
The first generation deadweight was a block of 316 stainless steel, formed to fit against the base and sides of an Arca-Swiss type screw-knob clamp (it is pictured in the second article). The clamp very firmly attaches the mass to the underside of the L-plate on the camera, which is itself firmly connected to the underside of the camera. The total connected mass being tightly connected to the metal case of the camera allows the mass to inertially dampen the shell of the camera, and in turn the sensor, against the tiny, high-G movements caused by the shutter. Any sticker on the base of the camera which protrudes and gets into the way of the metal-on-painted metal contact between the L-plate and the camera must be removed and any residual adhesive removed as well.
My earlier calculations showed that motions in the range of a few ten-thousandths of an inch were very significant as sources of distinct blurring in test images. Even a single ten-thousandth should theoretically show more than 1 pixel of blur with a collar-mounted 200 mm lens at certain shutter speeds. A 50mm lens would be no problem with a single ten-thousandth. A sheet of ordinary laser paper is 40 ten-thousandths of an inch thick. The little UL sticker on the underside of A7R’s for the North American market is similarly over 4 mils thick (over 0.004”, i.e. over 40 ten-thousandths) and includes a soft, springy adhesive layer. So if this tiny sticker compresses just 2% of its own thickness, that alone could actually make a just-visible difference in pictures made with a 200 mm lens. Multiply thousandths of an inch by 25.4 to get microns. Remember, the camera’s pixels are only about 4.9 microns each. Smaller than the short dimension of an E. coli, rod-shaped bacterium.
The first deadweight assembly totaled 16.8 ounces, not counting the L-plate, which was about four ounces less than I had intended. Many experiments showed it to help considerably, but that more mass would do better. My new target mass for the metal block, the clamp and the one stainless screw which holds the block to the clamp was 23.4 ounces, give or take a few tenths. The L-plate weighs an additional 2.8 ounces, according to the manufacturer. This time I was more careful about calculating the mass which would result from a given set of dimensions and took every machining detail into account, and sure enough, the finished part is just a hair over 23.4 ounces.
This block is 3.00" long, 1.25" front to back, and 0.63" tall. It has a groove milled into the top which is 1.50" wide (the width of the clamp), and 0.100" deep. This time the screw hole with a counterbore to accommodate the recessed allen head of the stainless 1/4-20 cap head screw is centered exactly in the block, so the center of gravity is directly under the tripod screw of the camera, instead of being beneath the shutter (experiments showed this worked better).
There are two viable choices for making such a deadweight, one employing fairly dense, readily available low-cost metals (stainless, brass, or bronze, in one of many alloys of each, or even simply pure copper) and the other employing a specialty metal often used where dense weights are needed (commercial aircraft weight trimming, race car center of gravity lowering, flywheel balancing, etc.). That metal is one of a group of sintered tungsten alloys. Pure tungsten has a specific gravity of 19, almost identical to that of gold, whereas the other three metals are only around 8 or 9. The alloy I like best, is only 17.0, but that still puts it at double the other metals. I decided to go for the more difficult but otherwise superior solution. Lead is 11.3, but quite poisonous. Silver is 10.5 but much too expensive and available in the wrong sizes. Gold and platinum are astronomically expensive and depleted uranium is… well, depleted uranium (radioactive, etc.). There are a number of exotic metals which are wonderfully dense (iridium and osmium, for example), but very expensive owing to their rarity or the difficulty of purifying them from ore and not readily machinable. So, beyond the basic metals mentioned first, the tungsten alloys are the standout option if you’re willing to go whole-hog. Some tungsten suppliers charge huge prices for a custom part, some don’t, but it’s still expensive. The bulk price of tungsten is much less than I figured it would be, only a dollar-something per ounce. It’s a lot more than that when formed into a useful part.
A block of 22.9 oz. made from pure copper could have these dimensions: 3.20" long by 1.40" front to back by 0.99" tall, with the groove cut for the clamp to a depth of 0.200" instead of 0.100", improving the location of the center of gravity compared to a groove of 0.100”, by moving the block that much closer to the camera. This part could be made from a 1 x 1.5 x 3.4" flat bar of Copper C110 H02 (1/2 Hard) from OnlineMetals.com: $22.78
To get up to about 23.4 oz., I would shoot for 3.20" long by 1.40" front to back and 1.012" tall, with the same 0.200 deep groove, and this part could be made from a 1.5 x 1.5 x 3.4" Copper C110 H02 (1/2 Hard) square stock $37.54, with more waste to cut and mill it down. Depending on your equipment, you can make it shorter front to back and taller, etc., to keep the mass about right.
Pure copper has a specific gravity of 8.91. Bronze and brass are mostly copper and come in various alloys from OnlineMetals. Bronze is more brownish and brass is more reddish golden. Type 316 stainless is only 8.0, so the volume of the block would need to be another 12% greater or so. Be careful that the block's width won't interfere with the battery door of the A7R. I figured 3.20" was more or less the largest acceptable width dimension. The easiest way to get more mass is probably to make the block taller. So stainless would be 3.20 x 1.40 x 1.13", roughly. Double-check how the deadweight will sit on your setup, so as not to be impossible to attach in a given circumstance, should you choose to make one for yourself.
The best price I could find for tungsten was about $340 for three pieces as described above, but without the groove and screw hole, and 6 to 8 weeks turnaround time. $300 minimum order for one part (go in with two friends to make it more reasonable). Midwest Tungsten Supply. The blocks will have machined and brushed, square surfaces. A local machine shop then cuts the groove, drills the two-diameter hole, cuts the screw to length, and eases the edges and corners, or Midwest Tungsten can do the work on the block instead, but you must give them complete dimensions in a mechanical drawing, rather than giving them the clamp and screw which they would then be sure to match as needed for a perfect fit. Expect to pay over $100 for the local machining. The tungsten is fairly hard and requires carbide bits (cutting advice available from Midwest Tungsten in a PDF). I preferred the MT-17C alloy: 90% W, 6% Ni, 4% Cu. Specific gravity 17.0 plus or minus 0.1 (it does vary). All Midwest Tungsten machined block dimensional tolerances are plus or minus 0.005". The groove should have a tighter width tolerance than that for a nice, snug fit, which is an argument in favor of the local machine shop, though Midwest’s tolerance for a cut within a block might well be better than for the outer dimensions. I don’t know.
The clamp again is a Really Right Stuff B2-FAB-F, $60. I see that this flat-backed version of their B2-FAB clamp has just been discontinued. I like this clamp the best for this use, but many others are possible. One could get the B2-FAB and have the machine shop grind off the two tabs on the back side. The cap-head 1/4-20 screw is either a 3/4" or 1” from RRS, and currently not easy to find on their new web site, which is missing a lot of content for some reason.
Tungsten has the highest melting point of any metal, which is why it's long been used for making filaments in incandescent lamps otherwise known as heaters which rather incidentally give off some light. Owing to this ultra-high melting point, 6192° F, it is frequently alloyed by a process called sintering, where a powdered mixture of the various metals is finely mixed, then poured into a temporary custom mold. The mixture is heated enough that the nickel and copper melt, but not the tungsten, forming a rather ordinary-looking metal, which is quite resistant to corrosion when such a non-ferrous alloy is chosen, remarkably heavy for its size, sufficiently sturdy, yet machinable. The process of making a custom block to order is considerably more expensive than making a block of copper or bronze or brass or stainless from extruded bar stock, but the block need only be half as big, and that was an attractive proposition. I think my second choice would have been bronze, though I like the appearance of the brushed 316 stainless the best, because the bronze is a little denser than the stainless (being mostly copper). Bronze does fit OK though, tripod head permitting, judging by my friend Mike Schultz's experience.
Here is the finished piece on the camera:
The Canon 24-70 f/4L IS is shown on the A7r. I find this lens to be far superior to the outwardly lovely Zeiss 24-70 f/4 OSS, for which I had high hopes. I tried three copies of each. All three of the Zeiss lenses were useless for corner to corner sharpness below something like 40 mm. The three copies of the Canon lens also varied a fair amount, but the best among them was downright decent, even quite good, and vastly better at the short end than any of the Sony/Zeiss lenses. I also found the by-wire focussing on the Zeiss to work poorly, and the immense distortion (only sometimes auto-corrected) of the Zeiss to be quite disturbing. I have the older 24-105 f/4L IS, but, despite the 24-70’s shorter range, the quality improvements are, for me, well worth moving to the newer lens.
And off the camera:
The bottom line: It works quite well and just as expected.
My five ways of testing it so far were as follows:
1) Repeat my primary, long-lens test, where a series of all shutter speeds from 1/200th through 1/2 second were shot (varying the ISO with constant, indoor lighting), with a Canon 70-200 f/4L lens set to f/8, camera vertical on my first-rate tripod setup, and my custom long lens support used to stiffen and support the weight of the lens. The camera's L-plate is clamped to the tripod head, for best steadying of the camera against shutter shake, and the L-plate is stiffened by clamping the camera's strap lug triangle between the plate and the camera under force. (The part of the RRS L-plate which goes up the left side of the camera is infinitely adjustable in its in-out position, relative to the other part of the L-plate, the part which sits entirely under the camera.)
The result of adding the deadweight was essentially total elimination of the shake at every speed between 1/200th and 1/2, where 1/25th and 1/20th are normally the worst with this lens, with only a faint trace here or there not enough that a fussy photographer would likely see an issue and suspect the shutter.
2) A series of hand-held tests done at 1/6th, 1/20th, 1/50th and 1/100th, with and without the weight, Canon 24-70 f/4L IS lens, IS turned off, set to 70 mm. This is the most hand-held testing I've done yet, with and without a weight.
The result is a major reduction in average blur in each series. I can hear the shutter being dampened by the weight as well as see the result in the images. The camera sounds more solid. At 1/20th and 1/50th, the overall camera movement was cut by more than half, on average, and the full range of results was shifted for the better. At 1/100th and at 1/6th the improvement was a little less but still quite substantial. Overall, it's quite clear even from just making a few dozen test exposures that the deadweight is a tremendously useful thing for hand holding the camera. It certainly won't match a SteadyCam, but it is definitely a big help to simply have a pound and a half effectively bolted to the base of the camera. The camera is just too light for this shutter at several shutter speeds (save for when the lenses are short, e.g. under 60 to 100 mm or so, depending on several variables).
3) A sensory test: The best way I've been able to feel the shake with my own sense of touch is to press the top of the hot shoe against my cheek bone and trip the shutter. I find this to be much more sensitive than using my fingers, as the shock is transmitted right into my skull. Without supplementary mass, the shutter produces a very distinct and obvious, small but sharp impact. With the new deadweight connected, the shake is simply gone. I can't feel a thing. I can still sense a bit of movement in other directions with my fingers, but the big motion is pretty much stopped, regardless of shutter speed.
Obviously it can't be perfect, but for all practical purposes, this compact mass of the optimal size and placement, connected to the camera with no give in the connection (beyond the elasticity of the metals and the modest compressibility of the paint on the body of the camera) essentially totally stops the camera from moving. My sense from simply feeling the camera with the weight on it is that it just won't shake enough for there to be any noticeable issue with any lens setup I might want to use, out to 200 mm vertical, camera’s L-plate connected directly to the tripod. The toughest case would be a 70-200 lens set to 200 mm, camera vertical, mounted by a collared tripod foot, because when the camera is trying to move itself, and the camera is not directly clamped to the tripod head, a given amount of shake will obviously move it more, and the further the collared foot is from the plane of the shutter, the more readily the camera can move, either up-down or left-right.
I don't have the parts to run that collar-mounted test, so I can't see how well the weight works for that situation. It will certainly help, but likely not solve over 90% of the problem.
4) I also ran a test with the Canon 1.4X tele-extender added to the 70-200 f/4L, lens set to 200, net focal length 280 mm. At 1/20th, the shake reduction with the weight was only about 60 to 70%, compared to no weight, with the camera vertical. Not good enough for critical work. With the weight removed and the camera horizontal, clamped directly to the tripod, lens supported by my long lens support rig, I see zero shutter blur at 280 mm and 1/20th. So I’m assuming that 200 mm is a safe maximum for verticals mounted as described, with the weight. I don’t know the upper limit for horizontals connected directly.
5) Two other lenses, a 150 mm fixed and a zoom at 160 mm (heavier and physically longer than the 150), both on a Mirex adapter foot, camera in vertical position. The small shake issues seen without a weight were eliminated entirely.
Since writing my earlier two articles, I have come to suspect that the double motion of the shutter is a smaller factor than I initially suspected, and that the major factor is simply the light weight of the camera. Given the incredibly fine tolerances required for motion blur during an exposure to be kept to a modest fraction of a pixel, we really need the benefits of electronic first curtain technology for a light weight and demanding camera like this with longer lenses, and to a lesser extent for long lenses with focal plane shutters in general. I’m guessing that it wasn’t included in the A7r simply because the required processing was more than could be handled by the camera’s circuitry with this sensor’s high pixel count (a 50% bigger job than in the case of the A7). We may never know. Fingers crossed for the A7r’s successor having EFC. Then we’re back to only having to think of wind, clumsy fingers and stampeding elephants messing with our longer lens captures.