When the K-30 was announced, one of the most attention-grabbing features on the specs sheet was the new SAFOX IXi+ autofocus system, which apparently includes a new light source detection sensor and diffraction lens. Sussing out what these improvements actually mean in a technical sense is perhaps best left up to better trained minds than ours, but what it sounded like to the average user was: better, hopefully faster and more accurate autofocus.
It's no secret that autofocus performance has been a stumbling block for Pentax for the entirety of the digital age. First, it was too slow. Then speed was improved, but accuracy became a more pressing concern. Then more annoying snafus started to appear—particularly AF that, at least on some cameras, didn't work properly under both natural and artificial light.
So, it's understandable that every new iteration of the SAFOX system brings with it the hope for an end to Pentax's AF troubles, and a new dawn of fast, reliable focusing under all conditions. Is SAFOX IXi+ the savior we've all been looking for? Read on!
Autofocus speed is a curious thing. No matter how fast it is, users always want it to be faster. But when AF times dip below a second, the truth is that it becomes very difficult for most people to discern when one camera is quicker than another. In general, we have found the K-5 to be acceptably fast when shooting using traditional phase-detect autofocus, in nearly all circumstances. Could it be faster? Yeah. Does it bother us in day-to-day use? Nope. What really bothers us is a lack of decisiveness, as well as focus inaccuracy. But that's a subject for another section.
Anyway, in this section we've performed a pair of autofocus speed tests to compare the K-30's PDAF and CDAF speeds to those of the K-5. The first test was performed indoors, using studio lights, with a target set five feet away and shot at four different EV levels. The second was conducted outdoors, under natural light, with a test target some 15 feet away. All test variations were measured multiple times and then averaged. Both cameras were equipped with the SMC Pentax-FA 43mm F1.9 limited lens.
For this test, we placed the cameras on a tripod and pointed them at a test chart taped to a wall five feet away, varying the ambient lighting. All four tests were conducted indoors, using studio lighting that was dimmed for each progressively lower EV setting. The FA 43mm Limited was set to an aperture of f/2.0. For each test, we set the lens to minimum focus, infinity focus, three-foot focus, and five-foot focus (aka in focus), and measured the time it took the K-30 and K-5 (each using both PDAF and CDAF) to lock focus.
The measurements were made using advanced audio recording software. We started the clock when the lens first started moving and ended it after the second AF confirmation beep. (Note that the two cameras have very slightly different beep sounds, so one may have a milisecond or two advantage.) For each variation of the test we obtained at least three samples and averaged the results, discarding obvious outliers. The results are tabulated below:
|Shooting from minimum distance|
|Shooting from 3ft|
|Shooting from 5ft (in focus)|
|Shooting from infinity|
*Focus-assist lamp fired
†When focus-assist lamp fired, time was 0.00; when the lamp did not fire, the time was 1.25.
All measurements are in seconds. Green = best time of the series, red = worst time of series.
As you can see, the K-30's CDAF implementation is a marked improvement over the K-5's in nearly all circumstances, with the K-30 taking 9 out of 12 contests. Strangely, it lost definitively to the K-5 when shooting starting from three-foot focus using CDAF, but won handily everywhere else.
In the PDAF race it's a closer contest. Overall, the K-30 won four matches and the K-5 eight, and they tied five times. When one lost to the other, it tended to be by either a large gap or a very small one, indicating that PDAF speed, at least in low light, is a bit of a lottery. (See, for example, the flip-flopping PDAF results for the EV 1 and EV 3 MFD tests.)
Perhaps it shouldn't be surprising that the company's flagship dSLR beat out a new mid-level contender, but we have to admit we were hoping for more. That said, the K-30 is by no means slow on the whole, and can at times be faster than the K-5.
Also worth noting: The K-30 used its AF assist lamp less in PDAF mode, but more in CDAF mode.
For this test, we positioned the cameras on a tripod and aimed at a contrasty bush about 15 feet away. We measured an EV value of 13 using our cameras' internal light meters. The FA 43mm Limited was set to an aperture of f/2.8.
|Shooting from minimum distance|
|Shooting from infinity|
All measurements are in seconds. Green = best time of the series, red = worst time of series.
This test again shows that the K-30's CDAF system blows away the K-5's, by quite a margin. The PDAF systems, however, are on perfectly equal footing in this situation.
So, the K-30's PDAF system doesn't provide a remarkable jump in speed over the K-5's, but in most situations it hangs in there quite well. On the whole, we're prepared to say it's comparable to the K-5 in everyday shooting. But raw speed is only one part of the equation; stay tuned for the Field Shooting Impressions section later on.
In terms of CDAF speed, the K-30 is obviously in a class ahead of the K-5, and appears to be pretty much on par with the K-01—a camera built entirely around contrast-detect autofocus. The only curiosity is the CDAF system's performance in the three-foot focusing test, where it lost every time to the K-5 (which, we have to say, did surprisingly well).
To test the autofocus accuracy of the K-30, we conducted a number of tests. Some were pretty straightforward: point the camera at a variety of subjects in a variety of light levels and light types, and see how often it hits the mark. Some were more complex, as you will see.
In these tests, we identified three targets in a corner of our office, chosen for their unique properties that would (hopefully) give the AF systems of the K-30 and K-5 a run for their money. Indeed, our goal was to make the AF fail.
We shot each target with each camera, using each AF type (PDAF and CDAF), in three different light levels, all of them quite dark. For each iteration of the test setup, we took three samples. (Yep, that means 108 exposures total... or it would if the cameras got lock every time.) We made note of when the cameras used their AF assist lights, whether they achieved focus lock, and if the desired subject was sharp in the resulting photo.
For each subject we recomposed and used the central AF point. All tests were made using the FA 43mm Limited lens at f/2.0, ISO 100. The lighting was primarily tungsten with a tiny bit of daylight coming in through closed blinds.
In all of the following tests, a "PASS" in the Lock? column means the camera obtained a focus lock and beeped to confirm focus; "FAIL" means it did not. In the Hit? column, "HIT" means the resulting photo was in acceptable focus, while "MISS" means it was not. ("N/A" indicates that there was no photo to evaluate due to a "FAIL" in the Lock? column.) An asterisk (*) indicates that the AF assist light fired.
The results of this test surprised us, mostly because the K-30 had been dead-on accurate in every bit of field shooting we'd done up to this point. Nevertheless, the camera's PDAF bit the big one here, albeit in a very particular way: every time the K-30's PDAF missed the target, it instead focused on the more contrasty edge of the printer's paper tray, slightly below and to the left of the Canon logo. So, the bad news is that the target we wanted wasn't in focus eight times out of nine; the good news is that something was in focus every single time. These results may indicate that the AF points on the K-30 are in fact quite large, or it might be a quirk of how the camera's AF algorithm analyzed the scene. On the CDAF front, the K-30 was nearly perfect, only failing to find a lock in one of the EV 1 samples.
The K-5 had a bit spottier record overall. Though it managed to hit the desired target more often when using PDAF, it still missed in one of the three samples at each EV level, and when it missed it didn't hit the tray edge, either. When using CDAF, it missed the first three times, "locking" focus when absolutely nothing at all was in focus and firing away anyway. Once it started using the AF assist beam, it was all gravy.
The high-contrast shoe company logo proved to be a juicy enough target, in this case. The K-30 nailed it every time, with the exception of one random CDAF focus locking failure (again). Accuracy was spot on, and the focus assist lamp was never employed in PDAF mode.
The K-5 had a rougher time when using PDAF. It misfocused rather badly on all three EV1 samples, not even coming close to the ECCO logo.
|EV 2||PASS|| |
Well, this one was pretty diabolical. Our own eyes could barely make out the edges of the drawer handle, so it's no surprise that both cameras failed miserably when using CDAF. The K-30 posted a perfect negative score on CDAF, while the K-5 managed to eke out two hits at EV1, somehow. Overall, though, there's no question that neither camera is really up to the task in these conditions.
When using PDAF, both did much better. The K-30 nearly posted a clean sheet but had one miss at EV1, while the K-5 had a bit more trouble, missing three in total at EV2 and EV1.
(It should be noted that all of the PDAF misses in this scene were relatively close misses, for both of the cameras.)
This test is a modified version of Pentaxforums community member robwill's K-30 in Tungsten light test. For this test, we placed five Matryoshka dolls in a line, so that only the head of the doll behind was visible above the head of the doll in front. The scene was lit with a single tungsten bulb, placed to the right of the dolls, just out of frame, and centered on the middle doll. Because of this, some dolls' faces were in shadow more than the others, but generally the scene was lit at around EV 4. We shot each doll's face with the FA 77 Limited at f/2.0, with both PDAF and CDAF, looking for misfocusing and any apparent tungsten-related front-focusing.
Overall, both cameras performed quite well. The K-30 had a little trouble with Dolls #2 and #3, while the K-5 stumbled on Doll #4. Part of this may be down to the precise framing of the shot and alignment of the tripod for each doll's face. Since the K-5 and K-30 bodies differ slightly in terms of where the lens mount sits, we had to reposition the tripod every time we changed bodies. Therefore, the framing (though very, very close) is not exactly the same for both bodies. Combined with the fact that the visible portion of the middle dolls' faces was barely larger than the central AF point, these slight differences may have led to focusing errors. Anyhow, in the end there was not a significant difference in PDAF accuracy between the two bodies.
Once again, however, the K-5's CDAF shows its weakness. It never lit up its AF assist lamp and was unable to obtain a lock twice while attempting to shoot the Doll #4. The K-30's CDAF, in contrast, used its AF assist lamp the entire time and never missed a shot.
Our goal with this test was to see at what point the PDAF and CDAF systems of each camera would simply cease to function. As such, we placed the cameras with the FA 77mm Limited on a tripod and pointed it at a static, very dimly illuminated scene. With the AF Assist lamp enabled, we started the test at an ambient light level of EV 1. With the lamp disabled, we started at EV 2. Each result represents three samples. OK means that all three samples obtained a focus lock; MIXED means that one or more samples failed to obtain a lock; and NO means that none of the samples was able to obtain a lock.
|AF Lamp Enabled||PDAF||CDAF||PDAF||CDAF|
|AF Lamp Disabled||PDAF||CDAF||PDAF||CDAF|
First the good news: both cameras are quite capable of shooting in total darkness... provided the AF assist lamp is enabled, your subject is stationary, and you have a lot of time on your hands.
Though we weren't testing for focus accuracy here (only the ability to obtain a lock, correct or not), we did note that the K-5 reported false locks with some frequency at extremely low light levels, while the K-30 did not. The K-5 also had much greater difficulty metering properly in extremely low light; all of the K-30 exposures at 4" or longer were properly exposed, while many of the K-5 exposures came out with a severe warm cast or were very dim, or both.
The only real curiosity here is that the K-5 eked out a minor victory in the EV 1 lamp-disabled CDAF test, with one sample finding lock. It's a hollow victory, though: that particular sample was badly out of focus.
To check out continuous shooting performance, we tried three different tests. All of these tests were conducted using PDAF focusing only.
First, we tried to shoot a very young, very energetic dog as she chased and retrieved a stick, running directly toward the camera. The venue was our back yard, and while the cherry tree cast some shadow, it was generally quite bright out. We used the FA 77mm Limited at f/4, which is a challenging combination, to say the least. (Slowish AF speed paired with a long focus throw and a narrow field of view.) We used Continuous Lo and Continuous Hi, as well as Auto-5, Auto-11, and Spot (center point) autofocus. We can summarize our findings like so:
Yep, our K-5 bit the big one during this test, on its fifth or sixth sequence of Continuous Hi shooting. About halfway through the sequence the mirror started to make an odd double-clacking sound. When we stopped shooting, it clacked a few times on its own. Concerned, we put it away while we shot with the K-30, then came back to it. After a brief rest it seemed to be somewhat rejuvenated, though it still clacks spontaneously from time to time.
Anyway, we can report with some certainty that the K-30 cannot handle a dog running at full tilt toward the camera—at least not a young, athletic dog running toward a FA 77mm Limited. In about 10 burst sequences with the K-30, using various combinations of fps and focusing points—about 200 shots, all told—we came away with precisely 13 that were in sharp focus on the dog's face. Spot AF had the best luck, getting six of those.
Performance with the K-5 (up until the point it pooped the bed) was quite similar, with Spot AF providing the best results of a bad bunch.
Next, we tried using the same AF drive and point modes to shoot the same dog running across the frame from side to side (panning, some might call it). Here, the K-30 performed much better. Of 137 shots, 96 (70%) were in acceptably sharp focus, and several more probably would have been if not for inept panning skills on our part.
Granted, our canine subject was a bit winded at this point and probably wasn't running as quickly as she could have, but we think it's safe to say that the K-30 deals with panning far better than subjects directly approaching the camera. In other words, its predictive AF isn't much to write home about, but given a slightly more predictable target it can perform quite well.
Spot AF once again provided the most consistent results, though all three modes did well in this test.
We also tried the K-30's new Expanded Area AF mode here, and found that it generally got results comparable to Spot AF. This makes sense, since in this mode the camera will use your chosen AF point unless there's nothing to focus on there. As long as you have good technique, Expanded Area AF will work exactly like Spot; if you slip up, it just might save your butt from time to time.
Finally, we asked our trusty assistant (human, not dog) to drive her truck at us in a non-life-threatening way at approximately 20 miles per hour. She repeated this six times (bless her patient soul) so that we could try Continuous Lo and Continuous Hi, as well as Auto-5, Auto-11, and Spot AF modes.
In each sequence we aimed at the grille of the truck and tabulated the number of focus hits and misses. We've collected these results into a chart below. Note that our assistant's truck is white, the grille is chrome, and the scene was very bright—all factors that could conceivably lead to a lower than normal number of focus hits.
|Lo / Spot||14||5||74%|
|Lo / 5-point||8||8||50%|
|Lo / 11-point||4||10||29%|
|Hi / Spot||20||6||77%|
|Hi / 5-point||13||10||57%|
|Hi / 11-point||5||16||24%|
Conceptually, this test is much like Dog Test the First, except that in this case the target is much larger, much more sharply defined, and moves at a more constant speed. As you can see, the results show a pretty clear pattern. Spot AF was once again definitively the way to go, while Auto-11 AF brought up the rear with a pretty sad hit percentage.
Auto-5 and Auto-11 AF modes simply seem to make the K-30 think too much, which slows down its reaction times and results in more out-of-focus shots. It's almost like the focusing system gets distracted and can't make up its mind. With just the single, central AF point to worry about, the K-30 thinks much more clearly and gets better results.
In the future, this is something we would (all) like to see Pentax improve on, but looking back on past models the K-30's behavior here is simply typical.
Ok, so we've put the K-30 through a battery of grueling focusing tests, desperately trying to make it misbehave. And we've had some success in that regard, though overall the K-30 performed on par with the K-5 in extreme focusing conditions. But what of the infamous problem of front-focusing under tungsten light?
Well, at the urging of forum-goer Class A, we gave it a shot. We mounted each and every one of our AF Pentax lenses on the K-30 and pointed it at Tim Jackson's famed focusing test chart. We tested each lens twice—once with five overhead bulbs (around EV 7), and once with a single tungsten bulb behind a shade, positioned to the right of the chart (around EV 3).
We're happy to report we found no evidence of misbehavior in either lighting condition. Each lens locked on quickly (all without the AF assist beam!) and accurately, and the results were perfectly in focus every single time.
We would have liked to try this on our K-5, but it was still misfiring in the wake of Dog Test the First. However, as we mentioned in the above-linked thread, our K-5 has never shown any difficulty with focusing under tungsten light, with any lens. As such, we suspect that the K-5 results would have been just as pinpoint accurate as the K-30's.
The most important thing to keep in mind here is that these tests were designed to test the K-30 under extreme conditions. We used lenses known to focus notoriously slowly. We used lighting situations in which most people wouldn't even hope to get a shot in focus. We intentionally tried to trick the K-30's autofocus system at every turn. We wanted to see where it would fail, and we found those places. But the truth is that the average K-30 will more often be used under good light and in normal shooting conditions.