[stevebrot]

The various online discussions I've seen all suggest that small values occur in bright sun and large values occur in dim light.

Not my experience. See comment above.


Steve

[stevebrot]

The AF module uses a photodiode array and in dim light (or with lenses with narrower wide-open apertures), those photodiodes take a little longer to gather a sufficient signal. You mentioned that the 1/2000 second shutter speed was not fast enough to stop motion of the dogs and that same blurring issue would affect the AF module.

Interesting...If correct, AF on black dogs should require very long AFIntegrationTime, regardless of LV.


Steeve

[mkymic]

Hey MightyMike Have you used the K1? Would you be able to compare it to the K1II for auto focus?
I have not sent mine in yet for upgrade ,it would be interesting to hear your opinion .

And after just getting back from a International Summit I am glad not to be such a
"professional" with a megaFPS camera ( make that several strapped to their bodies ).
You can't hear a thing as 40 or so professionals spray and prey for the money shot.

[MightyMike]

Hey MightyMike Have you used the K1? Would you be able to compare it to the K1II for auto focus?
I have not sent mine in yet for upgrade ,it would be interesting to hear your opinion .

And after just getting back from a International Summit I am glad not to be such a
"professional" with a megaFPS camera ( make that several strapped to their bodies ).
You can't hear a thing as 40 or so professionals spray and prey for the money shot.

My K-1II is an upgraded K-1... Here was my first report K-1 II AF-C Real World Report - PentaxForums.com and the gallery that followed K-1 II My First Airshow Gallery - PentaxForums.com

Some subjects require a spray and pray especially if people are being payed, however for most others smarter shooting, when to burst, when to wait, what settings to use, shooting with the rest of your senses (think ears), and anticipation or instinct play a big roll in successful result and higher keeper ratios.

[rawr]

I always thought that was a metric related specifically to how long it took from the camera being told to focus to when it acquired focus

Same here. The responsiveness of the AF operation.

But the idea, as the EXIFTOOL source code comment reveals,that it has something to do with how long (in ms) the AF sensor has been exposed to light in the lead up to an image being captured may be totally related to the responsiveness of the AF.

Time to post this K-3/K-1 AF infographic I did a while ago, just to remind us of what the hardware is:
:


Not in the diagram, however, is the additional SAFOX 'light source' sensor positioned next to the penta-prism of the K-1 and probably the K-3 too:
'to optimize autofocusing accuracy by compensating for focusing errors caused by specific artificial light sources' - probably tungsten light.

Pic related:

[stevebrot]

From this i would think manual lenses don't need center point only for focus confirmation? Just tried on my k-s2 sigma 70 on manual has all points but A 50 is limited to spot.

I don't know which bodies support what, but I was able to confirm on my K-3 that focus confirm allows the point to be set with AF lenses, but not with native manual focus lenses. This is all quite embarrassing since CIF and Focus Confirmation at other than center point has been a frequent question on this forum and the conventional answer has been center point only.*


Steve

* There is even a feature request thread for the K-1.

[MightyMike]

The various online discussions I've seen all suggest that small values occur in bright sun and large values occur in dim light.

The AF module uses a photodiode array and in dim light (or with lenses with narrower wide-open apertures), those photodiodes take a little longer to gather a sufficient signal. You mentioned that the 1/2000 second shutter speed was not fast enough to stop motion of the dogs and that same blurring issue would affect the AF module.

Interesting...If correct, AF on black dogs should require very long AFIntegrationTime, regardless of LV.


Steeve

I agree less light equals harder/longer time for AF to lock onto its target, the 1/2000th sec shutter speed was effective most of the time, note that the ISOs were between 400 and 800 at an aperture of F4.0, we don't know the sensitivity of the AF sensor however I figure its a great deal more sensitive than the imaging sensor. Until someone can define the AF modules sensitivity I don't think we can come to any conclusions as to what amount of movement it can handle and when that movement becomes too blurry to lock onto. This is one of the reasons I advocate for one of the benefits of in lens stabilization, when a person sees a stabilized image they are also therefore able to hold a telephoto lens more steady and the AF system likely has much more improved chance to lock onto a subject. This would suggest a person with steady hands will likely get more AF-C keepers over someone who isn't so steady. One solution I've suggested over the years is to make a stabilized OVF, it really shouldn't be too hard to take a set of those very small light weight elements and build a stabilization unit around them, it may make the camera a little larger but Pentax has creative engineers so I'm sure they could figure that out, and while they're doing that they could also do the same in the optical path to the AF module that would give Pentax all the benefits of a stabilized lens system while still having IBIS it should also improve AF acquisition time and tracking.

Steeve, it may be hard for the camera to lock onto a black dog, however it may be easier for the camera to see the contrasty edge between the dog and its background

[photoptimist]

Interesting. Today is the first time I have read of that value being the setting of a shutter of sorts. It had been my understanding that it was a report of the measured time to attain focus after the AF point was determined (the gap between red dot and green hexagon) rather than the settings value for an AF time window.

Google was no help, so I grabbed my K-3 and did a series of shots of similar low complexity subjects (light switch and edge of window molding) at both f/4 and f/11 at both LV 4 and LV 15. All four cases resulted in AFIntegrationTime = 4ms. I then looked at a photo taken yesterday of a high contrast subject (variegated iris bloom) at f/5.6 and LV 12 and found the AFIntegrationTime to be 0ms. Hmmmm...does not seem to be related of LV or aperture.

I downloaded the ExifTool Perl module for Pentax and found the source code comment:

"# effective exposure time for AF sensors in 2 ms increments

OK...that is clear as mud, though it does sound like some type of timing window. Went back to the camera and tried AF-C with various slower-focus AF lenses and still got values of either 0ms or 4ms.

Whatever...


Steve

(...AFIntegrationTime goes back at least to the K10D and maybe further...)

Interesting. Here's data on average AFIntegrationTime as a function of EV level posted to a DPR thread (Real World AF Speed Data. K20D and K10D: Pentax SLR Talk Forum: Digital Photography Review) by a "MightyMike" from Canada! The 2009 data set comes from 1176 images all made with a Sigma 24mm F1.8 that MightyMike said was one of his slowest-focusing lenses.

These were all from photos that had a near full or very high battery charge
EV0 = 250ms (average AF Integration Time)
EV1 = 197ms
EV2 = 181ms
EV3 = 194ms
EV4 = 164ms
EV5 = 103ms
EV6 = 63ms
EV7 = 27ms
EV8 = 28ms
EV9 = 11ms
EV10 = 17ms
EV11 = 16ms
EV12 = 13ms
EV13 = 0ms
EV14 = 0ms
EV15 = 250ms

these are from photos that had a low battery charge (much lower sampling
EV3 = 219ms (good charge = 194ms)
EV4 = 221ms (good charge = 164ms)
EV5 = 140ms (good charge = 103ms)
EV6 = 92ms (good charge = 63ms)
EV7 = 78ms (good charge = 27ms)
EV8 = 24ms (good charge = 28ms) not a typo
EV9 = 25ms (good charge = 11ms)

There's a strong relationship between EV and AFIntegrationTime but there's also scatter in the data which could reflect "black-dog in sunshine" effects or low-contrast scenes or who knows what.

[MightyMike]

Interesting. Here's data on average AFIntegrationTime as a function of EV level posted to a DPR thread (Real World AF Speed Data. K20D and K10D: Pentax SLR Talk Forum: Digital Photography Review) by a "MightyMike" from Canada! The 2009 data set comes from 1176 images all made with a Sigma 24mm F1.8 that MightyMike said was one of his slowest-focusing lenses.

These were all from photos that had a near full or very high battery charge
EV0 = 250ms (average AF Integration Time)
EV1 = 197ms
EV2 = 181ms
EV3 = 194ms
EV4 = 164ms
EV5 = 103ms
EV6 = 63ms
EV7 = 27ms
EV8 = 28ms
EV9 = 11ms
EV10 = 17ms
EV11 = 16ms
EV12 = 13ms
EV13 = 0ms
EV14 = 0ms
EV15 = 250ms

these are from photos that had a low battery charge (much lower sampling
EV3 = 219ms (good charge = 194ms)
EV4 = 221ms (good charge = 164ms)
EV5 = 140ms (good charge = 103ms)
EV6 = 92ms (good charge = 63ms)
EV7 = 78ms (good charge = 27ms)
EV8 = 24ms (good charge = 28ms) not a typo
EV9 = 25ms (good charge = 11ms)

There's a strong relationship between EV and AFIntegrationTime but there's also scatter in the data which could reflect "black-dog in sunshine" effects or low-contrast scenes or who knows what.

I remember doing something on that but totally forgot i had any data. That was some 9 years ago with many generation old tech... however the principal which I never disputed remains the same there is a correlation.

For comparison my photos were taken in the realm of EV12-13, however since I lifted the exposure a little in post lets say the realm of EV11-13.

[stevebrot]

Interesting. Here's data on average AFIntegrationTime as a function of EV level posted to a DPR thread (Real World AF Speed Data. K20D and K10D: Pentax SLR Talk Forum: Digital Photography Review) by a "MightyMike" from Canada! The 2009 data set comes from 1176 images all made with a Sigma 24mm F1.8 that MightyMike said was one of his slowest-focusing lenses.

These were all from photos that had a near full or very high battery charge
EV0 = 250ms (average AF Integration Time)
EV1 = 197ms
EV2 = 181ms
EV3 = 194ms
EV4 = 164ms
EV5 = 103ms
EV6 = 63ms
EV7 = 27ms
EV8 = 28ms
EV9 = 11ms
EV10 = 17ms
EV11 = 16ms
EV12 = 13ms
EV13 = 0ms
EV14 = 0ms
EV15 = 250ms

these are from photos that had a low battery charge (much lower sampling
EV3 = 219ms (good charge = 194ms)
EV4 = 221ms (good charge = 164ms)
EV5 = 140ms (good charge = 103ms)
EV6 = 92ms (good charge = 63ms)
EV7 = 78ms (good charge = 27ms)
EV8 = 24ms (good charge = 28ms) not a typo
EV9 = 25ms (good charge = 11ms)

There's a strong relationship between EV and AFIntegrationTime but there's also scatter in the data which could reflect "black-dog in sunshine" effects or low-contrast scenes or who knows what.

I could go back and look at my K10D files. I suspect we may just be seeing improved AF system metrics.

I remember doing something on that but totally forgot i had any data. That was some 9 years ago with many generation old tech... however the principal which I never disputed remains the same there is a correlation.

This afternoon, I took my K-3 with 18-55/3.5-5.6 into a closet at close to LV zero and still got 4ms at the raw edge of the camera's spec'ed sensitivity.


Steve

[MightyMike]

That's nice but without comparison's from Nikon , Canon and Sony cameras, we have no idea what those numbers mean. I've heard Canon AF called fast but inaccurate, so, I wouldn't in any way concede that Pentax AF is poorer than anyone else's without direct comparisons.

Of course the review sites who are the only ones that could easily do that kind of test, don't. The Camera store Nikon shooter in one of their videos was impressed with Pentax AF. He didn't say it was better, but it was in the ball park. That after a day of shooting all three systems. We don't hear that often. So it's always surprising when a guy doing comparison tests says, "Pentax isn't so bad." Funny how we don't have any Pentax shooters saying that. I think in terms of AF, there's been a lot of hoodwinking going on.

Hey normhead, since we seem to be digging up old posts I figured this one from the early days of the original K-1 may show a clear indication of how the other guys are actually better than Pentax. Check the bottom of the charts for the result from an old Canon 5D and Canon 70-200mm F4.0 L (not IS).

Extensive K-1 AF-C Bike Test Results: Pentax SLR Talk Forum: Digital Photography Review

[normhead]

Hey normhead, since we seem to be digging up old posts I figured this one from the early days of the original K-1 may show a clear indication of how the other guys are actually better than Pentax. Check the bottom of the charts for the result from an old Canon 5D and Canon 70-200mm F4.0 L (not IS).

Extensive K-1 AF-C Bike Test Results: Pentax SLR Talk Forum: Digital Photography Review

Except for the part where they are only discussing AF-c when 95% of the time I shoot AF-s even for action shots. So there is at least one AF environment where Pentax has the best AF (or second behind a Sony A9.

There are also Imaging resources tests showing the K-1 and K-3 are pretty much the fastest focus confirming of all cameras ever. I don't see that reflected in your title.


SO whether it's faster or slowest depends on which stats you cherry pick.

[MightyMike]

Except for the part where they are only discussing AF-c when 95% of the time I shoot AF-s even for action shots. So there is at least one AF environment where Pentax has the best AF (or second behind a Sony A9.

There are also Imaging resources tests showing the K-1 and K-3 are pretty much the fastest focus confirming of all cameras ever. I don't see that reflected in your title.


SO whether it's faster or slowest depends on which stats you cherry pick.

Taking 1000 photos of a cyclist is certainly not cherry picking. just because a camera and lens combo can acquire focus quickly doesn't mean it can follow the subject with any kind of consistency. How quick are you on the shutter button? would you be able to get even 50% of the photos of the dog running towards you in AF-S? I'm not trying to suggest you're not capable and your style doesn't have merit, just pointing out there are reasons for certain features and tests of those features. I still use CIF with a manual focus lens for some sports photos, that doesn't mean the AF tracking shouldn't improved.

Last year I was at a figure 8 race with a friend, I has the Sigma 70-200mm F2.8 EX Pre-DG he had the Sigma 70-200mm F2.8 Macro II IIRC (or some similar HSM or SDM lens anyways). My screw drive Sigma could snap into focus with blazing speed compared with his, however his was slower but could follow subtle movement better than mine. My lens would be ideal for your shooting style but a slower focusing lens though able to track smoother may not do well with your style.

[DSims]

Last year I was at a figure 8 race with a friend, I has the Sigma 70-200mm F2.8 EX Pre-DG he had the Sigma 70-200mm F2.8 Macro II IIRC (or some similar HSM or SDM lens anyways). My screw drive Sigma could snap into focus with blazing speed compared with his, however his was slower but could follow subtle movement better than mine. My lens would be ideal for your shooting style but a slower focusing lens though able to track smoother may not do well with your style.

That’s an interesting thought, Mike. In the past I’ve flounder that some screw-drive Pentax primes focus the fastest (and the best) for sports, but I never considered that some slower ~SDM designs could actually track better once on target. I may have to tests this out in the future.

[ANGUS YOUNG]

Thanks for the info. It will Be interesting to see the score with poor light. With sunlight I'm personally happy with My k30