Olympus E-P1 Blur Anomaly

Olympus E-P1 Blur issue

by Shawn Barnett and Dave Etchells

Because the full story of our testing of the Olympus E-P1's blur anomaly takes so many words to describe, we didn't want to burden our overview/shooter's report with all the details behind it. We also had (and continue to have) some concern that reams of detail on this one issue might overshadow the E-P1's good points, of which there are many. This is particularly important, given that the 1.1 firmware update appears to have made a change that largely eliminates the IS-related part of the problem.

At the same time, we can't just ignore the issue of anomalous blur with the E-P1. We also believe that it's always important for a test organization to share full details of their testing, so readers can trust and properly evaluate the results. Hence this separate page in our review of the E-P1, for those who are interested in all the details.

The E-P1's anomalous blur occurs under two circumstances, both arising from the same underlying cause.

1. With the 14-42mm M.ZUIKO lens mounted, the shutter's vibration can produce sufficient vibration in the lens elements, especially at 35 and 42mm, to cause noticeable blurring from about 1/100 to 1/200 second.
2. With firmware version 1.0, and far less so with 1.1, turning on image stabilization (IS1) will result in increased blurring in about the same shutter speed range.

We think Olympus has addressed the image stabilization issue well in the version 1.1 firmware, even though Olympus has made no formal acknowledgement of the issue, and there's no mention of the 1.1 firmware doing anything other than improving continuous-AF performance in both the camera and the two kit lenses. The only solution we can come up with for the 14-42mm M.ZUIKO lens vibration is either to not use it, to use it on a tripod when shooting in the critical speed range of 1/100 - 1/200 second, or else limit your print sizes. The lens vibration affects image quality whether image stabilization is on or off, and you're likely to see evidence of it somewhere in the image at all shutter speeds between roughly 1/100 and 1/200 second; at 1/250 and above, the second shutter curtain covers the sensor before the vibration can affect the image. Below 1/100 (assuming you can even hand-hold the camera steady enough to be able to tell), the vibration dies out quickly enough that you won't notice.

As an aside, we were a little surprised not only to see no mention of this problem on any other review site, but when we looked for evidence of blur in the images of other writers, we found very few shots taken at 1/160 second, with most writers omitting that shutter speed altogether. This is not to say that anyone was trying to hide those images, just that they didn't choose them, possibly because they were slightly blurry. Had we not noticed the issue, we'd have also just left out blurry images that we perceived to be caused by errors on our part, so we wouldn't have had any hand-held images at 1/160 either.

But Shawn in particular shot with the Olympus E-P1 for about two months before sitting down to really take a close look at both the image quality and the shutter speeds at which they were captured. This anomaly showed up in a large percentage of his images, most often taking the form of image-doubling at shutter speeds where it should not occur: 1/80 to 1/160 second with image stabilization enabled (IS1 setting).

To reduce the likelihood that this anomaly was only present in the earlier camera bodies we received directly from Olympus, with serial numbers 00126 (Camera 1) and 02460 (Camera 2), we purchased a third body at retail, which had serial number 05532 (Camera 3). This body still showed both problems, but somewhat less so than the others, which is encouraging.

What we saw

Blur resulting from this issue had two characteristics that let us distinguish it from ordinary motion blur:

• Because it results from a momentary vertical movement of the camera body and lens, anomalous blur is always in a vertical or nearly-vertical orientation. Ordinary motion blur may be in other directions (although it isn't always), and focus blur doesn't have a directional component to it.
• Because it is caused by the first shutter curtain reaching the bottom of its travel, the bottom of the sensor will see more of the effects of this vibration than will the top of the sensor, which is covered by the second curtain of the shutter relatively soon after the vibration begins (at least within the critical shutter speed range we're talking about). Because the image is inverted on the sensor plane, this means that the top of the image will show the most blur. Accordingly, if we saw equal blur at the top and bottom of the image within the critical shutter speed range, that meant we could safely discount it as user-caused blur, rather than anomalous blur caused by the shutter mechanism.

Indeed, if the shake is induced by the impact from the first shutter curtain, there would have to be a shutter speed where part of the sensor is covered by the second shutter curtain, and therefore protected from the influence of any sensor or lens motion. In the crops below, taken from an image captured at 1/160 second, you can see that the top and center of this particular image are affected by the anomalous motion blur, but the bottom is not.

At 1/160 second only the upper 2/3 of the image is soft while the lower 1/3 of the image is sharp. According to our analysis, this is because the second shutter curtain has already begun to close before the vibration of the camera or sensor begins. (The shutter closes in a downward motion relative to the camera body, but remember that the lens projects the image upside down and backwards onto the sensor, so it's the top of the sensor that captures a sharp image before the vibration can cause doubling.) The Olympus E-P1's X-sync is actually adjustable from 1/60 to 1/180 second, but we found no difference in behavior with the X-sync set to 1/60: It's possible that this setting only affects shutter operation when the flash is enabled, which is obviously not a condition we're testing under here.

As we said, the motion in the image above is quite vertical. This is characteristic of IS-induced blur, while with the IS off, as seen below, the blur offset suggests motion downward and slightly left. Both were quite consistent over hundreds of images.

Controlled testing before and after update

Once we saw the phenomenon in all three cameras, we decided to eliminate the first unit, Camera 1, because it was early production.

We had already updated Camera 2 with the autofocus update and saw a reduction in the overall number of significantly blurry images, but there was still a difference between the images with IS on and IS off. So we assigned values to the amount of blur per image, then shot six sets of 30 images at 1/160 second, taking the average of each to determine the amount of blur both with and without IS. Two of the sets were from Camera 2 and only include firmware 1.1. The last four sets were from Camera 3, shot first with firmware 1.0, then with firmware 1.1.

The images above are cropped from the middle of the test target, and show the amount of blurring that we assigned the various scores to. Note that few of the images at 1/160 second were really tack sharp, but also note that we're really pixel-peeping here. If you look at the top of the number two, you can see the clearest sign of motion blur. With IS on, the motion is more vertical, while with IS off, the motion is vertical and slightly left.

Tally all the results and here are the scores:

It's interesting that while Camera 2's images got noticeably better with IS on, Camera 3's images start out less blurry with IS off, but actually got very slightly more blurry with IS on, both before and after the firmware update. The good news is that the firmware update improved both cameras, but it did not entirely eliminate the problem, and thus leaves a risk of image doubling at random. The amount of blur in Camera 3's images, the one we purchased at retail, is far less noticeable, and though image stabilization does add a little bit of blur, it's not significant. (Note that Camera 2's firmware was updated before we began this particular test.)

Why the blur occurs at the shutter speeds it does

It seemed pretty unusual for a faster shutter speed to be blurred by such noticeable and repeatable motion blur. As we mentioned, we believe the same vibration is occurring with images shot at slower shutter speeds, it just doesn't make up as large a percentage of the exposure, and so isn't nearly as visible. But as the duration of the exposure gets shorter, the moment of vibration makes up a larger percentage of the exposure time. The graph below illustrates this point.

Note that this is not an actual plot of the vibration in question,
just an illustration of the concept.

In the illustration above, the hypothetical vibration is shaking the camera/lens for a majority of the exposure time at 1/160 second, so the resulting image shows the resulting blur pretty clearly. At 1/30 second, though, the vibration dies away quickly enough that the image is stationary on the surface of the sensor for most of the exposure time. The "fringe" around objects in the image resulting from the vibration will thus be all but invisible.

IS vs Lens

It's important to reiterate that the above images were made with the 14-42mm lens mounted and set to 42mm, the focal length that also produces the greatest IS-off blur. The amount of blur varies as you change the focal length, and it also varies from lens to lens. We took a 10-shot series at each focal length, taking care to keep the target the same size in the image frame, and scored them as above. The numbers below show our results (higher numbers indicate more anomalous motion blur).

Though at 42mm Lens 2 has higher levels of anomalous blur, moving to 25mm seemed to really tighten up the mechanism, resulting in fewer soft exposures. Looking at the incidence of blurred shots, though, there were three slightly soft shots at 25mm, and only one each at 18mm. Lens 1 also had only one blurred shot at 14mm, while Lens 2 had three (slightly) blurred ones. It may be that as the front lens barrels come back out at 14mm, the likelihood of vibration increases somewhat.

Just to finish out the analysis, at 35mm each lens had six out of ten shots with anomalous blur, and at 42mm Lens 1 had nine out of ten with noticeable anomalous blur, and Lens 2 had ten out of ten with noticeable anomalous blur.

How much will this affect my prints?

It's important to note, of course, that we're pixel-peeping here. This is a 12-megapixel camera, so the blur is much more apparent above than it would be in all but the largest print sizes. When we printed out the images shown here, "level 3" blur was quite evident in 13 x 19-inch prints, but much less so at 8 x 10. With the right subject, discriminating users would be able to see Level 3 blur at 8 x 10 inches, but we believe the majority of users with the majority of subjects would not. "Level 2 blur was visible in 13 x 19 inch prints, but didn't leap out at us: The average user might not be aware of it, and certainly not at 8 x 10 inches.

Using the printed results as a guideline, and keeping in mind that the numbers in the table above are averages, we believe discriminating users printing at sizes equivalent to 13 x 19 inches full frame (in other words, printing at that size, or cropping into a portion of the frame and enlarging to that level) would find some of their shots taken with version 1.0 firmware, the IS enabled, the kit lens within the focal length range, and the shutter speed within the critical range to be unacceptably blurred. With firmware version 1.1, we believe that such users would find a few of their images unacceptably blurred, under those same circumstances. At sizes of 8 x 10 and below, even fairly critical users wouldn't notice the issue with version 1.1 firmware. Consumer-level users might not notice the blurring with version 1.1 firmware, even at 13 x 19 inches.

Bottom line, with the version 1.1 firmware loaded, we don't see this as a crippling deficiency for the E-P1. But we definitely advise updating your camera to version 1.1, if you haven't already done so.

Why Bother?

Some may question the amount of effort we put into testing the E-P1, particularly for what may seem to them like a minor issue in the final analysis. When we first encountered the issue, though, it was something we hadn't seen before. Until we could fully characterize just when and how it happened, no one could know whether it was minor or not. We don't want to slam a camera based on partial information, and just as importantly, we don't want to simply sweep a potential problem under the figurative rug. Likewise, different people may have different interpretations of what constitutes "minor:" We might call something minor that some readers would consider critical. On the other side, others will look at something like this and say "so what?" (As noted in the review, there's some difference of opinion even between IR's editors as to how we each feel about the issue.) That's why the approach at IR is to characterize exactly what's going on, when & how it happens, and show what the results look like -- Then our readers can make up their own minds whether it's important to them. We think any E-P1 owner should apply the version 1.1 firmware, as there's no downside, and blur is noticeably reduced in the critical shutter speed range with IS enabled. Beyond that, it will be a matter of the type of shooting you do and how large you print. For readers routinely printing 11 x 14 inches and larger and also shooting a lot of hand-held material in the range of 1/80 - 1/200 second, it might still be a problem. More casual snap-shooters printing at 8 x 10 and below, or enthusiasts shooting from a tripod should have no concern. The key is that we've rooted out all the details so you can make up your own mind about it. It's a huge amount of work fully characterizing something like this, but this is the kind of truly in-depth reporting our readers have come to expect from us, and that's available nowhere else.

Summary

We really didn't want to have to go into such detail on this, but as we kept working to well and truly nail down the problem, it proved elusive and more complex than we could cover with just a quick note. When purchasing a third unit showed that the problem existed in later units, we knew we had to more thoroughly characterize and document it. We could go into even more detail here (this investigation consumed many dozens of hours), but we're happier to move on and say that half of the problem has been addressed with Olympus's firmware update, and the other half is better solved by either avoiding the 14-42mm M.ZUIKO lens, or by just using a tripod for critical shots within the affected shutter speed range. You could also just decide to accept a little motion blur in your handheld shots at 35 and 42mm, and confine your prints to 8 x 10, where it'll have relatively little impact.

See the Overview page for our full conclusion, but it bears repeating that the Olympus E-P1 is an excellent little camera when we eliminate this issue by shooting with the 17mm f/2.8 lens and turn off image stabilization. It's easy enough to turn on image stabilization when we know our shutter speeds will be slow, as it does seem to help more than hurt when shooting below 1/80 second. The trick is remembering to turn it back off when shooting at intermediate shutter speeds. Thankfully, Olympus's firmware update gets pretty close to eliminating that worry.

We went through a period of disappointment with the Olympus E-P1 when we first discovered this issue, but we're glad we took the time to pinpoint why and when the blur occurs. We can now resume liking and using the E-P1 as much as before.