[jsherman999]

I'd be interested in seeing the results of a controlled test on that. I'm not sure the best way to set it up, but I'd want to see a test that both timed focus to shutter AND focus accuracy.

If you have a phone that can record snippits of sound, one way to do it would be to touch your phone the camera or lens, focus on something close (min distance,) then record the sound made from close-to-infinity focus, and close-to-midrange.

Then, load the sound samples up into one of the many free programs available that show the amplitude/time graph of the sound image, and you can see exactly how long each AF run takes. Someone did that with a couple lenses here a while back, and it was interesting to see the results they got.

[regken]

When the k20D first came out a lot of early adapters that had k10D's were saying the AF was better. It took a while before more hard nosed buyers appeared and said no, not much difference at all. Think I'll sit on the sidelines until this issue is thoroughly beat to death.

[Ratmagiclady]

Well, it IS a marketing release and has no specificity or data to back up the claims, so what does this statement really mean?

I think it means they added a new light source to the data range of phase-detection AF.

We know what the light source sensor does, and frankly I would be hard-pressed to believe that adding a sensor that has to be read and processed and then taking that data and using it to adjust the AF would make the whole process FASTER. After all, in the K20D no such sensor/calculation was included.

Actually, since they said 'data range' rather than 'detection range,' it's quite possible that they *didn't* add a new sensor, just the ability to receive and process more useful information from *a* sensor. Could even be the very same one, that just previously hadn't been fully utilized.

More accurate in certain lighting, yes, as that is what it is designed to do, but Faster, no. This part has nothing to do with speed improvements.

Could have. . If having more useful data makes a positive lock easier under whatever set of conditions, then, there's less dithering to do. It's kind of like radar: a bigger antenna, so to speak.

How do algorithms improve low-light focus speed?

An algorithm is a mathematical process that is the means by which something like AF data is turned from a bunch of light falling on a sensor, into commands for something like the AF or SR.

If there's a more-efficient way to get the same results out of fewer computer cycles, (and fewer times blipping the lens to get more data) yes, it happens faster. This is why computer speed is measured in Hertz.

I also do not see any mention of anything new in continuous AF or in predictive AF capabilities, although improving the speed of the overall system should also improve the ability to continuously focus even if it is not a true predictive system.

Agreed. But this is also all about the math the computer does and what information it has to process. There isn't like a separate 'predictive AF machine' someone sticks in a camera, ...just more math.

Speed improvements come from that being fast and accurate, and from the machinery being able to respond as quickly as possible to it. The machinery can't go any faster than the calculations, though. So improving the calculations ....either with more computer power, a broader range of data, or more efficient ways to process it, certainly pays dividends.

[Wheatfield]

Frankly, I think it's hard to compare: after all, AF speed seems an anecdotal matter and observations at the counter certainly don't seem much of a judge...

If you go out and use two different cameras, and one just seems to have snappier performance, then you've a qualified comparison that one camera is faster than the other. I have more interest in this sort of anecdotal approach than a purely quantifed approach that may or may not have anything to do with actual photography.
For example, a friend and I were out shooting, him with his D300, me with my K20. At the end of the day, we didn't need to quantify that his D300 ran rings around my K20, it was obvious to both of us that the D300 was significantly more responsive.

[Mister Guy]

If you have a phone that can record snippits of sound, one way to do it would be to touch your phone the camera or lens, focus on something close (min distance,) then record the sound made from close-to-infinity focus, and close-to-midrange.

Then, load the sound samples up into one of the many free programs available that show the amplitude/time graph of the sound image, and you can see exactly how long each AF run takes. Someone did that with a couple lenses here a while back, and it was interesting to see the results they got.

Timing it is the easy bit. The tricky bit is how to fairly and accurately determine which AF scheme is more ACCURATE. You'd need a difficult lighting scenario on a subject with a lot varying depths, and then a fair way of verifying that you didn't simply mis-aim the autofocus point. I'm thinking the best way may be a mannequin head and a heavily locked down tripod in dim lighting. You'd need to autofocus carefully in bright day lighting to make sure the point was placed properly, perhaps with a daylight balance studio continuous light, then turn the light way down, deliberately un-focus, and THEN AF and post the result of both the control shot and the lower light shot.

[Ratmagiclady]

If you go out and use two different cameras, and one just seems to have snappier performance, then you've a qualified comparison that one camera is faster than the other. I have more interest in this sort of anecdotal approach than a purely quantifed approach that may or may not have anything to do with actual photography.
For example, a friend and I were out shooting, him with his D300, me with my K20. At the end of the day, we didn't need to quantify that his D300 ran rings around my K20, it was obvious to both of us that the D300 was significantly more responsive.

Well, as K20's go, well, we kind of talked about this as regarded the question of if the new ostensibly-just SR-related algorithms speeded up the AF, (I still think so, even if it's just cause the new SR freed up computer space for the trickier conditions I tend to like) I'm not gonna dispute a d300 is faster, (Particularly since I've never had more than a half a minute of sharing-time in a friendly fashion with some Nikonian who'd never apparently seen an F-1 with a 1.2 lens on there. ) There was like a big Tammie on there, I was like, 'Yep, it focuses.' )

I don't doubt people saying it's faster than a K20d: I definitely don't see the difference being a deal-killer, particularly since I got my K20d for a full grand less and I sure *wouldn't* have gotten a brand new 50 1.4 that'd AF at *all* without still *further* expense if I were near that league. Only people with much of a leg to complain on are richer than I, anyway.

AF's still a crutch to me, anyway: given a bigger finder for digital, I'd probably still be doing MF most times. Doesn't matter how fast the AF is, in a way: telling it what to do isn't always faster than just doing it myself. (Though it often is: eyes and reflexes aren't getting any younger.)

Just for S&G I tried shooting at some oncoming stuff out of a moving car with my K20d, when the tracking thing came up, trusting to AF-C.

Seemed to work fine. I dunno. Sometimes I think too much is made of all this. But, then, I've only got the one FA lens, for AF not counting the kit.


Oh, but, salient point being, I don't know if you can pre-bash the K-7 based on holding one that may or may not have even been a release model, in a shop.

[asdf]

I have more interest in this sort of anecdotal approach than a purely quantifed approach that may or may not have anything to do with actual photography.

You could've turned your outing into a laborious experiment, if you took an audio recorder or some other means of recording AF speed. Then you'd take the average and standard deviation. On the other hand, the difference between DA 55-300 and DA 40 focus speed is pretty apparent...

[falconeye]

the DA*300/f4 on my K20D bodies in finding the subject and following fast ones. [...] I have gone back to my FA* 300/F4.5 for wing shooting birds because of it.

So, you noticed this as well?
I bought a used FA*300/4.5 after the DA*300/4 came out and I am very pleased with the AF performance. Flying birds in AF.S

[Mike.P®]

So, you noticed this as well?
I bought a used FA*300/4.5 after the DA*300/4 came out and I am very pleased with the AF performance. Flying birds in AF.S

Same here but a F* 300mm f4.5.
I love that lens, small, sharp and excellent IQ .. even with the Pentax 1.7x Tc on.

[Ray Pulley]

Sorry, this is flat out incorrect. There is no speculation that the 2nd sensor detects light wavelength. This is established fact straight from the horse's mouth. The 2nd sensor compensates for lighting type (tungsten, fluorescent, etc.). If you haven't watched the K-7 intro video on Youtube already, definitely check it out.

But you're right that this should have no bearing on speed, but rather on accuracy. Speed improvements should come from whatever mojo they put into the K-m. According to people who have tested it side-by-side with other Pentax cams, the K-m is significantly faster at AF than older cameras, which has nothing to do with less AF points because it's also faster when the compared cameras are both just using 1 point.

No one said that the color sensor in the new AF system wasn't established fact. You missed the point, but the K7 does have 2 AF "systems" and it is easy to see how people can get confused between that and the second sensor inthe AF system for color balance AF adjustment purposes.

I have read the claims of KM AF improvements, but have also seen that the improvements are small enough that they do not even convince everyone that the improvemenst are significant, and certainly would not satisfy any of the complainers who say that they cannot shoot action with Pentax.

Do not get me wrong, I will happily take any and all AF improvements Pentax comes up with. I am just starting to get a (sinking?) feeling that we are headed into a new round of AF whining if this new and improved AF system doesn't live up to the competition/expectations, especially for action shooting.

I would be happy to be wrong in this case.

Ray

[dhalbrook]

Sorry, this is flat out incorrect. There is no speculation that the 2nd sensor detects light wavelength. This is established fact straight from the horse's mouth. The 2nd sensor compensates for lighting type (tungsten, fluorescent, etc.). If you haven't watched the K-7 intro video on Youtube already, definitely check it out.

But you're right that this should have no bearing on speed, but rather on accuracy. Speed improvements should come from whatever mojo they put into the K-m. According to people who have tested it side-by-side with other Pentax cams, the K-m is significantly faster at AF than older cameras, which has nothing to do with less AF points because it's also faster when the compared cameras are both just using 1 point.

Actually accuracy and speed have to ultimately be considered related, depending on how you measure speed. If an AF system is incredibly fast, but inaccurate to the point of needing to course correct (or accept a flat out inaccurate result), then in the end it may take a longer time to achieve accurate focus than an AF system that is slower but is accurate on the first go. Does that make any sense?

I'm hoping the color-sensor and AF-assist lamp help difficult focus situations, but I'm betting the AF motor is also stronger in this body as well, leading to an increase in speed. There's a decent chance that there's a few hardware tweaks as well, and verified that there the focus algorithms have been revised.

As others have said, the proof will be in the pudding. I'm seriously hoping AF is much improved, and it's going to be hard to wait as I very nearly already replaced by K10D more than once.

[Torphoto]

I think we have to take the Pentax Af systems performance in comparison with other similar cameras. Having played with the 40D and D300, I can tell you that AF tracking in AFs is as fast in good light. In low light the K20D can lock with out AF assist better than the 40D but both the 40D and D300 just kill the K20D in AFC ( also their AF assist especially on the D300 is better ), the K20D's AFC simply can not track subjects comming toward or going away from it, latteral tracking is no problem.

I further noticed that the lens I was mostly using was the 50-135mm, slowish AF vs the 40mm and 70mm limiteds. The limiteds faster AF made a huge difference as the faster AF from the lens enabled it to track better, much better than the 50-135 but still far behind the 40D or D300 with perpendicular subject movement.


From what I have gathered the AFC on the K7 is much improved vs the K20D, the question is how does it compare now vs the D90, 50D and D300.

[Ray Pulley]

I'd think that 'adding a light source to its data range' means that the AF is now sensitive to more wavelengths: Wavelengths being what 'phase-detection' *detects,* that's potentially a big deal. You get more data processed faster, (with new algorithms and processor speed, apparently) ...so, in theory, you get faster and more accurate AF.

No. The AF now has a sensor that can read the color temperature of the scene and apply an offset to the AF to compensate for the fact that the AF system is calibrated at a given color temperature (probably daylight) and therefore will be a bit off at other color temperatures.

Note all of the complaints in reviews and tests about bad auto white balance in all DSLR systems. This is because the shutter is closed and the camera doesn't read any data about the scene (other than metering) until the shutter is fired and the mirror flips up and the exposure is taken. The metering system is not color-sensitive in most DSLRs, so how can it ever know anything about color temperature and adjust the AF before the image is captured? Only by using another sensor that is looking at the scene and sending color temperature information to the AF system before the image is captured.

This is a nice feature, no question, but it is definitely addtional processing and work that has to happen that was not happening on the K10D or K20D.

We do now have the Prime II processor, so it could well be that the processing power has improved enough to incorporate this function with no speed penalty, or even while still allowing the system to be faster overall.

That's the claim anyway.

BTW, accuracy over speed is fine with me too, but I would like better low light AF speed.

Ray

[Ratmagiclady]

Do not get me wrong, I will happily take any and all AF improvements Pentax comes up with. I am just starting to get a (sinking?) feeling that we are headed into a new round of AF whining if this new and improved AF system doesn't live up to the competition/expectations, especially for action shooting.

I would be happy to be wrong in this case.

Ray

I'm kind of right with you, there, though I still think people complaining if they can't have the word 'Pentax' on a pricey lens and body to go play 'action shooter' with, then there are places to go.

I don't*have* any particular interest in another company playing on canon nd Nikon's terms. If there's yet another two thousand dollar bloated camera and yet another two thousand dollar big lens out there, I couldn't touch em if I wanted to.


As long as they keep improving things, I've got a place to be. 'faster AF' on a d90 does me no good if I'd spend far longer cause I couldn't set exposure with a dial and I can't afford the thing anyway. One thing I can't help but notice is that the doomsayers are all over the place looking for comparisons. Who's got what I need and can conceivably have? And workarounds when I *can't* have the latest nicest and fastest?

Guess who.

[Ratmagiclady]

"No. The AF now has a sensor that can read the color temperature of the scene and apply an offset to the AF to compensate for the fact that the AF system is calibrated at a given color temperature (probably daylight) and therefore will be a bit off at other color temperatures."


Ah, so what you're saying is they added the color of the light source as a *parameter* for AF at all. This will also help.