Sony DSLR-A100 Exposure Variations

Exposure Variations

We don't normally go into this much detail on exposure accuracy for the cameras we test, but in the case of the Sony A100, we found a few situations where the exposure varied in ways we didn't expect. We first noticed it when Luke was checking the exposure for our Multi Target shot. After shooting an exposure bracket, he checked the images on the computer, then went back to the camera to shoot a resolution series at the proper exposure (EV boost) setting. To his surprise, this second group of images were underexposed compared to what he saw when bracketing.

After some experimenting, we consistently found two different kinds of exposure variation in the A100, one sort that reproduced the issues Luke encountered, another that didn't seem to correlate with anything in particular, but that was also somewhat less significant in its impact. The worst-case variations we saw were on the order of 0.5 EV. Try as we might though, we never found the ~2 EV jumps in exposure that at least one other reviewer reported. (And we've now seen and tested four separate A100 bodies.)

"Type I Variations"

The variations that Luke encountered were associated with a very specific camera setup, namely the combination of Wide-Area Autofocus and Multi-Segment metering. When we switched to center-weighted or spot metering, or manually selected an AF point, these variations largely disappeared. We refer to these as "Type I" variations, since they were the first we encountered. What appeared to be happening was that the AF system would select a primary subject, and the focus would be locked on that single point, but the other potential AF zone candidates were still contributing to the exposure determination, via a weighted average that could change significantly from one shot to the next as the AF system dithered between selecting one or the other AF point. This is pure supposition, but is about the only plausible mechanism we could come up with for such non-deterministic exposure behavior. It would also explain why these variations went away when the metering mode was changed or when the AF point was manually selected. The shots below show an example of this sort of exposure issue.

The shots above illustrate the issue we encountered with the combination of wide-area AF and multi-segment metering. They're two of a group of nine shots captured within seconds of each other, under studio lighting (HMIs in softboxes), with the camera locked down on a very solid camera stand. (So the lighting and the framing of the subject were absolutely identical between shots.) Of the nine shots, the camera exposed two of them at 1/20s and f/5.6, and the other seven at 1/30s and f/5.6. When we switched to center-weighted metering, the exposure on this subject became absolutely rock-steady.

From our experimentation, it appears that our Multi target might represent something of a worst-case subject for triggering Type I exposure variations, as it has lots of contrasting elements under or near AF points. Switching to a less-contrasty target produced smaller Type I variations, generally 1/4 EV or less.

For many casual users, an exposure variation of +/- 1/4 EV (a total worst-case range of 1/2 EV) may not be that big an issue (many Happy Snappers would never notice it), but for people who really care about exposure, it would clearly be an issue. In our own work it was particularly vexing, as exposure levels in our comparison photos are very critical.

"Type II" Variations

While the Type I variations described above are only encountered when multi-segment metering is combined with wide-area AF, we also found variations that were more random, in that they didn't appear to correlate with any subject content, metering, or AF mode. In a couple of instances, we found exposure differences of a third of an EV or more with the camera locked down on a tripod and just pointed at the backdrop of the "Sunlit Portrait" test shot. (And yes, we did check that the lighting wasn't varying; both by monitoring it with a Sekonic L-508 light meter and by shooting the same subject with another DSLR, which produced perfectly consistent exposures.)

This really played havoc with some of our shooting, as we in some cases had to shoot exposure brackets three or four times over to get one series that actually stepped linearly through all the exposure values. The series below shows a typical example of what we encountered:

In the series above, there's no difference in the aperture and shutter speed reported in the file header between the first and second shots, even though the second one is clearly brighter. There's a normal 0.3 EV step between the second and the third, but then the aperture and shutter speed both change between the third and the fourth, producing an 0.7 EV jump. Finally, between the fourth and fifth images, the exposure once again increases by the proper 0.3 EV amount.

The series above was shot using multi-segment metering, but with a single AF point selected, centered over Marti's hair on the left side of her head as seen in the images. (This is also why this series is framed a little differently: I wanted to have an AF point right on Marti's hair as that's the most reliably deterministic focus point for this shot, and I didn't want to perturb the camera at all to reframe after locking focus, to avoid introducing any variables that could possibly contribute to exposure variation.) The roughly 0.3 EV variation seen here is fairly typical of the results we saw in this particular pattern of camera behavior.

After a while of being completely perplexed by these Type II variations, we began to see some patterns that suggested that the fault might lie with the lenses we were using (the standard 18-70mm kit lens). We discovered that if we shot a given exposure series several times over, by the last series the exposures settled down a great deal. In fact, any sort of significant repetitive series of exposures seemed to figuratively clear the camera's sinuses, with exposure becoming more uniform following a period of rapid activity. Also, while we don't have the recorded statistics to confirm it, it seemed that both production-level camera/lens systems we were working with during this time were starting to show more reliable exposure characteristics as we continued to operate them.

We haven't had an opportunity to use the cameras in question with any Sony lenses other than the two "kit" lenses that shipped with our production samples, but the behavior outlined above suggests that sticky diaphragms in both lenses might have something to do with the variations we were experiencing. We also seemed to see slightly more variations with one of the kit lenses on both production-level bodies than with the other lens on either. Our theory is that frequent cycling of the lens iris during an exposure series would temporarily loosen it up and help it operate more consistently. And over time, as large numbers of exposure cycles accumulated, the apertures operated more freely, and the problem diminished. This is pure speculation at this point, as we'd have to have samples of several different lens types to conclusively demonstrate that the lenses were at fault, but the good news is that Type II exposure variations seem to be diminishing as we've continued to use the cameras.

The bottom line

Having now seen these exposure issues in a prototype camera, a near-production sample, and two full-production versions of the A100, it does appear that there's a genuine issue with exposure accuracy when using multi-segment metering and wide-area AF together to photograph subjects with high local contrast. (The scenario for our so-called Type I exposure variations.) But the very length of the problem description speaks to the narrowness of its impact. If you're a casual photographer primarily interested in family snapshots, it's quite possible that you'd never notice this issue. On the other hand, if you're a more experienced shooter who (like us) prefers to choose the AF point rather than letting the camera pick it for you, you also might never see it. (In fact, if you don't commonly shoot exposure brackets of your subjects, you might never be aware that it's happening at all.)

Basically, for the Type I exposure problem to appear, the subject apparently needs to be fairly flat (coplanar) so the AF system may be a little undecided about which AF point to pick, and there also need to be elements with very different brightness values near the AF point the camera ends up using. If there is a single strongly dominant subject (such that the AF system's algorithm settles on a single AF point very easily), or if there aren't any very strongly contrasting elements near the AF point selected, the exposure can be quite consistent. Likewise, if you manually pick an AF point (as noted, this is manner of shooting we personally prefer 90% of the time), or if you use center-weighted metering, the problem never appears.

This is by no means to dismiss the issue as trivial though: Accurate, repeatable exposure is fundamental to photography, and a thousand-dollar DSLR should at the very least be able to expose consistently and accurately, regardless of the AF or exposure mode selected.

The second type of exposure variation (that isn't correlated with metering or AF pattern) is more likely to be seen by the average user, as it can apparently crop up in any shots you're taking, and regardless of AF or metering mode. Its effect was generally more subtle though: We could see it when scanning quickly through captured images on the camera's LCD, but even when viewing images side by side on a computer monitor, it was surprisingly difficult to pick out. To be able to reliably pick out which image was brighter or darker, we often had to overlay them in Photoshop and switch quickly between them. - So that's pretty subtle indeed, and likely to be a complete non-issue for the average user. There's also some (slight) evidence that the Type II exposure problem diminishes with use, as the lens iris gets "broken in," for lack of a better word.

So at the end of the day, the Sony A100 is subject to some variability in its exposure metering, but it's of a magnitude that many users may not notice. One type of exposure variation seems to be inherent in the design of the A100's AF and metering systems, but is easily circumvented by using either center-weighted or spot metering, or by manually selecting an AF point. (It's also possible that this variability could be fixed at some point in the future via a firmware update.) A second type of exposure variation may be linked to lens operation, but is much more subtle in its effect, and it seems to diminish somewhat over time.

Ultimately, you'll have to decide for yourself what you think of the A100's exposure variations: For many users, we think they'll be a complete non-issue, for others they could be nothing less than deal-breakers. Hopefully the information we've assembled here will help you decide which camp you fall into.