| Articles |
Secrets of the Sony A55
I was one of the early adopters of the A55 in November of 2010, it being the first Sony Alpha that I have found both affordable enough and interesting enough to own. I'm not going to review it, or even write much about the now well-known things that make it - and the other SLT models - so revolutionary. Instead, I want to write about several aspects of the camera that Sony has not publicized, and that are only being documented now by users like myself.
There's a lot to know about this camera, and every time I start to write something I realize there's more I need to explore. This page will definitely undergo revision as I have time...
A55 Bugs, Limitations, and Oversights
Let's jump right in with a list of A55 behaviors that are undesirable. These can be either bugs (undesirable things that Sony did not actually intend the camera to do); limitations (undesirable things that Sony probably knew about but could not overcome with present technology); or oversights (undesirable things that Sony certainly knew about and could have fixed, but failed to address). There could also be a much longer list of missing or poorly implemented features that Sony should have and easily could have corrected in hardware or firmware; but all cameras suffer from this and probably always will.
Bug 1: The very first thing you should do when you get your A55 is turn off Auto Review! Auto Review is unnecessary with SLTs because you're always seeing a simulated, adjusted view of your final shot. Leaving it turned on adds a very annoying delay to even the most casual shooting. I don't know why this is the case, and I don't know why Sony decided to make it default to on. Turn it off, and press the Play button
when you want to review something.
Bug 2: The second thing you should do is obtain and apply the firmware version 2.00 update if that has not already been done. With the original firmware, HSS (High Speed Sync) autoflash did not work correctly in wireless mode. This particular problem was corrected in the update, and some other nice features were added as well.
Limitation 1: The live view image shown in the EVF and on the LCD is always a fraction of a second behind reality. Depending on lighting conditions, the delay might be either imperceptible or as long as 1/10 second. This means that the user may need to anticipate action shots that require precise timing. It's not Sony's fault, though. At present, this type of delay exists in all electronic viewing systems and is unavoidable.
Limitation 2: There is an additional delay whenever a recognized dedicated flash is used (either the pop-up flash or a Minolta/Sony external flash). This delay occurs between the moment the shutter button is depressed and the moment the exposure begins. My own tests have shown that the delay is about 1/5 second. It's not clear why the delay exists, and a similar though shorter delay reportedly occurs in the A580. Delays of this length are not present in any previous Alpha DSLRs, so I don't think the issue can be blamed on simple pre-exposure autoflash metering time. Perhaps Sony could do something to eliminate the delay in newer cameras. There is no such delay when using non-recognized flash units fired with cables or radio triggers.
Oversight 1: When the shooting mode is set to Auto or Auto+, and the camera determines that flash is needed, it chooses a shutter speed, aperture, and ISO that are appropriate for mild underexposure of the ambient lighting. This is not always a bad thing (see the later note here about intentional Slow Sync), but when the scene is dark or dimly lit the shutter speed can become extremely slow, on the order of one second or more. Then the flash fires as well, but at reduced intensity because it assumes the ambient light has contributed something to the scene. The result is a shot that can be partially blurred, incorrectly exposed, and poorly color balanced, perhaps confusing users who generally shoot hand-held and probably know nothing about the mechanics of exposure and flash use. This behavior won't happen in 'P' or 'A' modes, where the camera does not allow the shutter speed to drop below 1/60 second when flash is active, so the two Auto modes should probably have been made to work the same way. Of course, advanced users who know what they're doing can always override the behavior in 'P' and 'A' modes as well by engaging the camera's Slow Sync flash mode. All slow shutter speeds then become available.
Oversight 2: When in 'M' mode, the camera attempts to show you the exact exposure you will obtain based on your settings and the ambient lighting - unless you are using a recognized dedicated flash, in which case the camera will automatically 'gain up' the display so you can see the scene. This is a good arrangement in many situations. But suppose you're shooting in a dark or dim environment (like a studio) while using external non-recognized flash units. The camera doesn't know about those, so it will continue to show you a very dark and sometimes unusable image. Sony should have included an option allowing you to force the display to gain up in such situations. The company is definitely aware of this oversight because it also exists in all previous Alphas that offer live view, but at least those cameras offer the option of using the optical viewfinder instead. The important difference with the SLT models is that these new cameras have no optical viewfinder! So, until this is corrected, here are some workarounds that will force the display to gain up...
1. Set up the camera as you normally would, but rotate the mode dial one notch to 10fps mode for focusing and framing, then back to M for the shot. It's reasonably easy for many situations.
2. Rather than using a specific shutter speed, use the 'Bulb' setting, which always invokes automatic gain. Press and release quickly, and you can simulate fairly fast shutter speeds.
3. Lift the pop-up flash and then mount a hot shoe adapter. This is possible because the SLTs have a unique new flash hinge design. Then you can attach a radio trigger or cable for firing your remote flash. You can block the pop-up flash with something if you don't want it contributing any light to the scene.
Purple Fringe and Sony's SLTs
Reports have recently appeared describing a purple fringing problem with the SLT cameras. The best known example so far is Kurt Munger's article here. (Although Mr. Munger's site is a great resource and is very informative overall, that particular comparison page is questionable at best. Not only are the purple fringe observations incomplete, but the high ISO comparison must have been shot in some parallel universe where anything can happen. Please go instead to sites like DPReview and Imaging Resource for example images comparing SLTs against other cameras. They show results consistent with the real life experiences of most users.)
What is purple fringing, or PF? Rightly defined, it is the appearance in a digital photograph of distinct purple coloring where a very bright area meets a dark area at a sharp boundary, and which is not present in the original scene. It has been observed and documented for years, but its causes are still not completely understood. Some people feel it involves unwanted reflection and/or refraction of light at or near the surface of a camera's sensor, probably related to the particular design of the sensor's microlenses and perhaps the conductors that run between them. Others say it is the result of bright/dark boundaries being incorrectly rendered by the Bayer interpolation used in most digital camera sensors; and since the predominant Bayer color is green, this theory says the resulting error appears as purple, the complementary color. Another contributing cause could be overloading of a sensor's light wells causing spillage into adjacent wells - also known as blooming.
What is NOT purple fringing? Well, any sort of similar phenomenon that is caused only by a lens, and that would also appear when that lens is used on a film camera. Traditionally understood and defined chromatic aberrations (CA) are NOT purple fringing; but purple fringing does often appear to be related to the presence of CA in a lens. So a lens with inherent CA may also be likely to produce purple fringing under the right (wrong?) conditions with a digital camera.
Unfortunately, many people erroneously think that simple axial or lateral chromatic aberrations are 'purple fringing' and report them as such. These distinctions are important, and they must not be mixed up. An obvious giveaway is that when you see two unexpected colors appearing along boundaries in an image (generally magenta/green but also sometimes blue/yellow), you are seeing CA.
Purple Fringe: My Personal Findings
At first I was baffled by such reports because no shots I had taken with my A55 showed evidence of unusual purple fringing; and I have had little success finding it with my wide or medium lenses stopped down, which is the way I shoot the vast majority of the time. Only now do I know the best recipe for finding it:
Use a long lens (greater than 50mm, but NOT a mirror lens)
Use a fast lens (f/2.8 or greater)
Shoot wide open
Include small intense highlights adjacent to dark areas
So, yes - following that recipe with my A55 will produce some visible purple fringing. But wait... doing so also produces purple fringing in the same portions of an image of the same scene with the same lens and the same parameters when shot with my Konica Minolta 5D. That hasn't caused me any regrets or stopped me from using it for the last five years. Purple fringing, when it exists, is just a fact of life with digital photography; and it can be largely or completely corrected in post processing. Thus the real questions are:
1) How bad is the purple fringing produced by the A55 (or the other SLTs, which have different sensors and might therefore behave differently) compared to other cameras?
2) Is purple fringing with the SLTs due to the sensor design only (as it presumably is in all previous DSLRs), or is the SLT mirror also a contributing factor?
3) Will Sony find a way to modify the sensor and/or the SLT mirror and/or something else in these or future SLTs to improve the situation?
Answer #1: The purple fringing produced by the A55 is in fact more intense than that produced by the KM5D. Fortunately, I have to pixel peep at high magnification to see the difference; and as with the 5D, it only shows up under certain conditions described above. The lens that shows the effect most noticeably in both cameras is my Minolta AF 135mm f/2.8. It's well known that that lens, though highly regarded, also shows considerable axial CA with all cameras, so I suspect the two effects are related. (My Minolta AF 500mm mirror lens is free of CA and apparently free of purple fringing, as expected.)
Here's one of the clearest examples I've been able to get using the 135mm wide open. It's a 1:1 crop of some details on my car's hood, taken from near the center of the frame. First the KM5D version, which I first upsized to match the magnification of the A55:
Now the A55 version (yes, the focus point and exposure were both slightly different... sorry). You can see that the resolution is finer, and the purple fringing is more obvious:
A difference is visible in this test and in other similar ones I tried with the 135mm at f/2.8. I'm not going to bother showing results at smaller apertures because the purple fringing in both cameras becomes less obvious as the lens is stopped down, until it is hardly apparent at all with apertures smaller than f/11. And as I said, with most of my other lenses the effect is even harder to see. In other words, the examples above are about as bad as it's ever going to get in practice.
Answer #2: Purple fringing in the A55 has nothing to do with the SLT mirror. It is strictly an interaction between particular lenses and that particular sensor. See my test shots and analysis below.
Answer #3. Although there's nothing about the SLT mirror that needs correction in terms of purple fringing, in the time since I began these tests, it has been shown conclusively that the mirror in the A77 (and presumably in its contemporary and subsequent models) does eliminate the known 'ghosting' effect of the earlier mirror... and there might well have been other advances too, which I hope to explore at some point.
The SLT Mirror: Effects on Image Quality
A great deal of bandwidth has been expended on discussions about the effects of the SLT mirror on the A55's image quality. I have shot several tests with and without the A55's mirror in place to learn the answers for myself. My most recent (and probably last) test is shown below. This scene was shot two times using the A55 and Minolta 135mm f/2.8 lens at maximum aperture and manually focused at the infinity stop. I chose that lens and aperture mostly because I wanted to be sure to invoke some purple fringing. I shot RAW+JPG with the mirror removed at ISO 200 and 1/4000s, then took the camera off the tripod, replaced the mirror, put the camera back on the tripod, and took the second shot at 1/3200s. The sun moved a bit in the three minutes between the two shots, so some of the reflective highlights appear slightly different. Everything else remained the same, and I did my best to frame the scene exactly the same way each time. Here is the overall scene shot both ways (non-mirror on left, mirror on right)...
Here is a 1:1 pixel crop of an area near the center of the scene shot both ways (non-mirror above, mirror below)...
Here is a 1:1 pixel crop of an area near the edge of the scene shot both ways (non-mirror above, mirror below)...
My findings from this and previous tests are...
1. The mirror reduces the amount of light reaching the sensor by about 1/2 stop. This is what we've been told all along, and my tests confirm it. Although it's probably hard to see in the images above, the shot with the mirror is actually slightly underexposed compared to the one without the mirror, even though the shutter speed in the second shot was 1/3 stop slower. If the A55 provided manual shutter speed increments of 1/2 stop, I believe a setting of 1/2800s would have produced exactly the same overall exposure as the first shot.
2. The mirror reduces image detail. It's logical to expect this... but again, many discussions have been posted about exactly how much image degradation occurs. The 1:1 pixel crops that I've included help to show how much. Another clue is provided through the relative file sizes of the two versions of the scene. A55 RAW and JPG files are compressed when saved from the camera. Both files from the non-mirror version are slightly larger in size than the files from the mirror version by a few percent. This indicates that the non-mirror version must contain more detail and/or more noise. Since noise should be essentially the same in both shots, we can attribute the larger size to more detail.
I am going to add something important here. Remember that in this recent test I used an aperture of f/2.8 in bright daylight to obtain the worst possible purple fringe and image degradation effects. For one thing, are you really going to do a lot of shooting at maximum aperture in bright daylight where such issues will be observable? Another consideration is that setting focus at the infinity stop of the lens might have been a bad idea for such testing. Why? Because the mirror, although extremely thin, might very slightly change the focal point of the lens when it's in place (this is known to happen with flat glass rear filters in mirror lenses, for example), meaning that the exact same position of the focusing mechanism should not be used for both shots. This has been hard for me to determine or control in practice, but I acknowledge that it might have affected the results above. The fact is that I have shot similar tests at smaller apertures, and in most of those tests the results without the mirror are indistinguishable in terms of detail rendition. Below are two examples of the same scene as above, shot six months earlier, using the same lens, but at f/11...
I honestly cannot claim to see a difference in these two shots. But again, the relative file sizes tell me that slightly more detail was captured in the version without the mirror. (Notice also that purple fringing is absent in both shots, although there is still some lateral CA.)
So how significant is the reduction of image quality in real terms? I can't be scientifically precise, and my tests are subject to other influences as I have described...but I'm going to say that the mirror might produce up to about a 5% loss of detail, depending on the aperture, and perhaps on other factors as well. This figure comes from comparing the file sizes in the worst case examples as mentioned earlier, and also from doing separate experiments not shown here (downsampling the non-mirror shot, then upsampling it again to its original size, then comparing that result to the mirror shot and also to reduced resolution JPEGs of the same and similar scenes). So until I learn of a more scientific way to do this evaluation, I will say the mirror can be expected to retain at least 95% of the image detail of a non-mirror shot most of the time. In my judgment, any such difference is probably inconsequential for most real-world purposes.
3. Purple fringing is not influenced in any way by the SLT mirror. This is immediately obvious by looking at the 1:1 pixel crops.
4. I have nothing to add regarding the 'ghosting' effect of the mirror, which has been shown to cause tiny duplicates of point source lights in night shots. I believe that has been adequately documented.
5. I have nothing to add regarding David Kilpatrick's discovery of a special type of internal flare caused by the mirror in very unusual circumstances. I have been able to duplicate that effect under similar conditions.
The SLT Mirror: Construction, Durability, and Cleaning
In order to gather the information above, I have removed the A55's mirror several times. It's not something I plan to do frequently, but it isn't really difficult. In fact, it's easier overall than removing and replacing the focusing screen assembly found in ordinary Alpha DSLRs, and lots of people decide (sometimes unwisely) to take that task on. I'm not going to describe exactly how to remove the mirror, because I don't want to be seen as recommending that; but instructions are out there if you look for them.
The mirror frame appears to be made of a rigid gray polycarbonate or ceramic material. The mirror itself is definitely not glass, as some have stated, but rather a thin and flexible plastic sheet. The shot below shows the edge of the sheet, which extends beyond the frame, with a ballpoint pen tip and a human hair in the scene. As you can see, it's really thin. I've measured it with a micrometer at less than 0.004", the same as a sheet of standard copier paper.
To show that it's flexible, here's another shot where I'm bending one of the corners, which returned to its position when released. By the way, both of these shots were made with the same A55 while the mirror was out - in manual focus mode, of course. Removing the mirror makes autofocusing impossible, but everything else works fine.
Here are some more interesting tidbits... Although handling the frame while it's out of the camera is easy, I had a bit of a fumble the first time because I was also trying to photograph it. As a result of juggling too many things, I ended up with a 1/4" square fingerprint on the front-facing side of the mirror. It looked like a fingerprint on any lens element. We are warned in the user manual that we must never touch the mirror, and it has been said that getting any substance at all on it will require a return to Sony for replacement (see the section on dust reduction about halfway down this page in the Imaging Resource A55 review), and I think the cost these days is about $100. Well, I cleaned the fingerprint myself. I did it the same way I have sometimes cleaned little dust particles off the surface of the mirror: I used a clean cotton swab and wiped very gently and carefully, occasionally breathing on the mirror to provide a little moisture. I also took the opportunity to rest the swab lightly against it and confirm that the mirror does flex a bit in response to pressure. Do not actually press on the mirror material because it can permanently deform.
I shot some test scenes while the mirror was removed and repeated them after cleaning and replacing it. Looking at both versions afterwards, I saw no ill effects from the fingerprint and cleaning episode and have seen none after shooting thousands of frames afterwards. In the years since these tests I have also easily removed visible dust particles from the mirror by carefully lifting them off with a tiny artist's brush on several occasions. So... I would say SLT owners can rest assured that although the mirror is not glass, it is not nearly as fragile as Sony would have us think.
Text and images © 2011-2014 Ray Lemieux