One of very, very few repeatable AF.C autofocus tracking tests

beholder3 · 02-25-2019, 12:29 AM - **5 Likes**

The German magazine" c't Digitale Fotografie 02/19" just published another AF.C autofocus test comparison.

While 99.9% of what is written about autofocus on the internet not even reaches elementary school scientific quality the testers here did set up some scenarios which could be reproduced technically 100%, so results are relevant and say a lot (within the boundaries tested).

They did set up contrasty Lego toy trains in good light and shot them as they were coming at the camera in various angles and with various items block sights.

I will not invest the effort to describe more details, just the results. Who really cares has to buy the magazine/article: Autofokus im Praxistest | c't Digitale Fotografie | Heise Magazine

Again the results are telling a lot about the credibility, skills and brains of users who claim "near 100% autofocus hit rates" on the net for their loved gear. They seem to shoot brick walls only.

It also shows that even within a group of relatively similar scenarios, the relatively best camera can change. So overarching general statements which gear is best are nonsense all the time.

Obviously all cameras struggled quite a bit.

Attached Images

**clackers** · 02-25-2019, 01:42 AM

The Fuji XT-3 has 425 PDAF points on its sensor for nearly 100% coverage.

Correlation with its performance on the test is shown by this data to be poor, especially compared to the DSLR.

**Rondec** · 02-25-2019, 03:37 AM

Originally posted by clackers

The Fuji XT-3 has 425 PDAF points on its sensor for nearly 100% coverage.

Correlation with its performance on the test is shown by this data to be poor, especially compared to the DSLR.

It does appear that there is something still to be said for having a dedicated auto focus sensor rather than having auto focus points on the sensor. The D500 does appear to outperform the mirrorless options in this test.

biz-engineer · 02-25-2019, 03:55 AM

Originally posted by clackers

Correlation with its performance on the test is shown by this data to be poor, especially compared to the DSLR.

On the DSLR, the PDAF sensor works while the mirror is down and the image sensor unload the image into memory.
On the mirrorless camera multiple image frames are acquired for PDAF prediction and also capturing images that are in focus for the camera user (time sharing of the image sensor resource).
But the MILC is cheaper to build, single sensor, no mechanical actuation, and more simple to operate than DSLR.

beholder3 · 02-25-2019, 06:04 AM

Originally posted by Rondec

The D500 does appear to outperform the mirrorless options in this test.

In three out of five cases yes. In the two others (#2 and #3) it underperforms.

#2: Training coming straight at the camera, distracting moving objects cross in front of it.
#3: no tracking really, just panning, following the train from right to left in front of distracting background

BrianR · 02-25-2019, 06:23 AM

Full marks for setting up a repeatable test, but I have to ask how relevant is this for common shooting situations? Lego trains are pretty small and I assume fairly close?

How did they classify shots as focus nailed, failed somewhat, and failed completely? I'm genuinely curious, I'd be interested in (hopefully free) software that can objectively assign some kind of sharpness value to a photo or parts of a photo (like after the fact focus peeking).

**ffking** · 02-25-2019, 06:46 AM

I find that fascinating insofar as it is abundantly ckear that no camerais best in all scenarios, and the D500 is brilliant atsome and rubbish at others - ie no one test gives a reliable picture od a cameras capability

normhead · 02-25-2019, 07:03 AM

Somehow that's what I always suspected. Tracking AF is still largely irrelevant to my shooting. That it is far less accurate than AF.s, which is well over 95% is one of the best kept secrets in photography. Bottom line, if you can get the shots you want with AF.s you are still ahead of the game, by a considerable margin.

beholder3 · 02-25-2019, 08:54 AM

Originally posted by BrianR

How did they classify shots as focus nailed, failed somewhat, and failed completely?

All scenarios had the requirement to focus on the tip of the locomotive leading the train with all waggons. "nailed" was focus correctly on tip of the train. failed somewhat means focus hit the locomotive but not the tip of it. failed completely means the lcocomotive was not in focus at all.

In the magazine there is a pretty detailed explanation with smaple images and sketches of the setup to explain what was done.

biz-engineer · 02-25-2019, 11:31 PM

Fast AF tracking performance is for gear focused photographers. If AF tracking was a must, there would be zero market for medium format. Evidence shows that the best professional photographers don't use AF tracking. And a lot of people who seek high AF tracking performance have money but no talent, unfortunately.

**UncleVanya** · 02-26-2019, 02:00 AM - **1 Like**

Originally posted by biz-engineer

Fast AF tracking performance is for gear focused photographers. If AF tracking was a must, there would be zero market for medium format. Evidence shows that the best professional photographers don't use AF tracking. And a lot of people who seek high AF tracking performance have money but no talent, unfortunately.

I suspect after a point this is true. A little bit of this does work wonders for a friend who has essential tremors and grandchildren. The gx85 Panasonic has good tracking which has made him more able to capture special moments. He is no pro that is true, but he is very very happy that this camera does this reasonably well. But after a point the differences in ability seems like "measurbating".

biz-engineer · 02-26-2019, 03:21 AM

Originally posted by UncleVanya

A little bit of this does work wonders for a friend who has essential tremors and grandchildren.

For tremors I would tend to believe that SR help more than af tracking.

**Rondec** · 02-26-2019, 04:06 AM

Originally posted by beholder3

In three out of five cases yes. In the two others (#2 and #3) it underperforms.

#2: Training coming straight at the camera, distracting moving objects cross in front of it.
#3: no tracking really, just panning, following the train from right to left in front of distracting background

I guess I'm more thinking about the D500 versus the XT-3. The other two have smaller sensors and therefore have a bit more depth of field to play with in test situations. My i phone focuses pretty fast, but that's because it doesn't really have to focus...

beholder3 · 02-26-2019, 08:51 AM

Originally posted by Rondec

I guess I'm more thinking about the D500 versus the XT-3. The other two have smaller sensors and therefore have a bit more depth of field to play with in test situations. My i phone focuses pretty fast, but that's because it doesn't really have to focus...

I am just making assumptions, but I guess that even with 1 inch at F4 at these close distances it is pretty easy to judge if the AF hit or not. Yes, there will be more DoF, but it probably did not influence their judgement.

**UncleVanya** · 02-26-2019, 10:59 AM - **1 Like**

Originally posted by biz-engineer

For tremors I would tend to believe that SR help more than af tracking.

Both help, the inability to keep the camera pointed directly at the subject is where fancier tracking options helped him. Also the movie capture modes work wonders when kids are scrambling about - at least for him.

The tech advances are a boon to some people, but most of us really don't require them. I frequently feel like users are spoiled by the level of sophistication of modern gear.

02-25-2019, 12:29 AM - 5 Likes	#1
beholder3 Pentaxian Join Date: Dec 2011 Posts: 3,112	One of very, very few repeatable AF.C autofocus tracking tests The German magazine" c't Digitale Fotografie 02/19" just published another AF.C autofocus test comparison. While 99.9% of what is written about autofocus on the internet not even reaches elementary school scientific quality the testers here did set up some scenarios which could be reproduced technically 100%, so results are relevant and say a lot (within the boundaries tested). They did set up contrasty Lego toy trains in good light and shot them as they were coming at the camera in various angles and with various items block sights. I will not invest the effort to describe more details, just the results. Who really cares has to buy the magazine/article: Autofokus im Praxistest \| c't Digitale Fotografie \| Heise Magazine Again the results are telling a lot about the credibility, skills and brains of users who claim "near 100% autofocus hit rates" on the net for their loved gear. They seem to shoot brick walls only. It also shows that even within a group of relatively similar scenarios, the relatively best camera can change. So overarching general statements which gear is best are nonsense all the time. Obviously all cameras struggled quite a bit. Attached Images

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Results from a really repeatable AF.C test (no Pentax inlcuded) as reference	beholder3	Pentax K-1 & K-1 II	14	03-13-2019 05:24 PM
What does autofocus tracking really mean?	maxxxx	Pentax DSLR Discussion	22	03-08-2014 02:06 PM
How good is the K-3's AF-C tracking AF?	Jeff Charles	Pentax K-3 & K-3 II	6	12-09-2013 03:41 PM
K-5 vs K-r (autofocus and tracking objects)	WoRaS	Pentax DSLR Discussion	13	02-26-2011 05:22 AM
K10d Autofocus Tracking	Workingdog	Pentax DSLR Discussion	28	12-11-2007 05:40 AM

02-25-2019, 01:42 AM	#2
clackers Otis Memorial Pentaxian Join Date: Jul 2013 Location: Melbourne Photos: Albums Posts: 16,397	The Fuji XT-3 has 425 PDAF points on its sensor for nearly 100% coverage. Correlation with its performance on the test is shown by this data to be poor, especially compared to the DSLR.

02-25-2019, 03:37 AM	#3
Rondec Loyal Site Supporter Join Date: Mar 2009 Location: Gladys, Virginia Photos: Gallery Posts: 27,653	Originally posted by clackers The Fuji XT-3 has 425 PDAF points on its sensor for nearly 100% coverage. Correlation with its performance on the test is shown by this data to be poor, especially compared to the DSLR. It does appear that there is something still to be said for having a dedicated auto focus sensor rather than having auto focus points on the sensor. The D500 does appear to outperform the mirrorless options in this test.

02-25-2019, 03:55 AM	#4
biz-engineer Pentaxian Join Date: Feb 2015 Photos: Gallery Posts: 12,231	Originally posted by clackers Correlation with its performance on the test is shown by this data to be poor, especially compared to the DSLR. On the DSLR, the PDAF sensor works while the mirror is down and the image sensor unload the image into memory. On the mirrorless camera multiple image frames are acquired for PDAF prediction and also capturing images that are in focus for the camera user (time sharing of the image sensor resource). But the MILC is cheaper to build, single sensor, no mechanical actuation, and more simple to operate than DSLR.

02-25-2019, 06:04 AM	#5
beholder3 Pentaxian Join Date: Dec 2011 Posts: 3,112 Original Poster	Originally posted by Rondec The D500 does appear to outperform the mirrorless options in this test. In three out of five cases yes. In the two others (#2 and #3) it underperforms. #2: Training coming straight at the camera, distracting moving objects cross in front of it. #3: no tracking really, just panning, following the train from right to left in front of distracting background

02-25-2019, 06:23 AM	#6
BrianR Veteran Member Join Date: Dec 2010 Location: Ontario Photos: Gallery Posts: 3,332	Full marks for setting up a repeatable test, but I have to ask how relevant is this for common shooting situations? Lego trains are pretty small and I assume fairly close? How did they classify shots as focus nailed, failed somewhat, and failed completely? I'm genuinely curious, I'd be interested in (hopefully free) software that can objectively assign some kind of sharpness value to a photo or parts of a photo (like after the fact focus peeking).

02-25-2019, 06:46 AM	#7
ffking Site Supporter Join Date: Nov 2015 Location: Old South Wales Posts: 6,038	I find that fascinating insofar as it is abundantly ckear that no camerais best in all scenarios, and the D500 is brilliant atsome and rubbish at others - ie no one test gives a reliable picture od a cameras capability

02-25-2019, 07:03 AM	#8
normhead Pentaxian Join Date: Jun 2007 Location: Near Algonquin Park Photos: Gallery \| Albums Posts: 40,451	Somehow that's what I always suspected. Tracking AF is still largely irrelevant to my shooting. That it is far less accurate than AF.s, which is well over 95% is one of the best kept secrets in photography. Bottom line, if you can get the shots you want with AF.s you are still ahead of the game, by a considerable margin.

02-25-2019, 08:54 AM	#9
beholder3 Pentaxian Join Date: Dec 2011 Posts: 3,112 Original Poster	Originally posted by BrianR How did they classify shots as focus nailed, failed somewhat, and failed completely? All scenarios had the requirement to focus on the tip of the locomotive leading the train with all waggons. "nailed" was focus correctly on tip of the train. failed somewhat means focus hit the locomotive but not the tip of it. failed completely means the lcocomotive was not in focus at all. In the magazine there is a pretty detailed explanation with smaple images and sketches of the setup to explain what was done.

02-25-2019, 11:31 PM	#10
biz-engineer Pentaxian Join Date: Feb 2015 Photos: Gallery Posts: 12,231	Fast AF tracking performance is for gear focused photographers. If AF tracking was a must, there would be zero market for medium format. Evidence shows that the best professional photographers don't use AF tracking. And a lot of people who seek high AF tracking performance have money but no talent, unfortunately.

02-26-2019, 02:00 AM - 1 Like	#11
UncleVanya Loyal Site Supporter Join Date: Jul 2014 Photos: Gallery \| Albums Posts: 28,401	Originally posted by biz-engineer Fast AF tracking performance is for gear focused photographers. If AF tracking was a must, there would be zero market for medium format. Evidence shows that the best professional photographers don't use AF tracking. And a lot of people who seek high AF tracking performance have money but no talent, unfortunately. I suspect after a point this is true. A little bit of this does work wonders for a friend who has essential tremors and grandchildren. The gx85 Panasonic has good tracking which has made him more able to capture special moments. He is no pro that is true, but he is very very happy that this camera does this reasonably well. But after a point the differences in ability seems like "measurbating".

02-26-2019, 03:21 AM	#12
biz-engineer Pentaxian Join Date: Feb 2015 Photos: Gallery Posts: 12,231	Originally posted by UncleVanya A little bit of this does work wonders for a friend who has essential tremors and grandchildren. For tremors I would tend to believe that SR help more than af tracking.

02-26-2019, 04:06 AM	#13
Rondec Loyal Site Supporter Join Date: Mar 2009 Location: Gladys, Virginia Photos: Gallery Posts: 27,653	Originally posted by beholder3 In three out of five cases yes. In the two others (#2 and #3) it underperforms. #2: Training coming straight at the camera, distracting moving objects cross in front of it. #3: no tracking really, just panning, following the train from right to left in front of distracting background I guess I'm more thinking about the D500 versus the XT-3. The other two have smaller sensors and therefore have a bit more depth of field to play with in test situations. My i phone focuses pretty fast, but that's because it doesn't really have to focus...

02-26-2019, 08:51 AM	#14
beholder3 Pentaxian Join Date: Dec 2011 Posts: 3,112 Original Poster	Originally posted by Rondec I guess I'm more thinking about the D500 versus the XT-3. The other two have smaller sensors and therefore have a bit more depth of field to play with in test situations. My i phone focuses pretty fast, but that's because it doesn't really have to focus... I am just making assumptions, but I guess that even with 1 inch at F4 at these close distances it is pretty easy to judge if the AF hit or not. Yes, there will be more DoF, but it probably did not influence their judgement.

02-26-2019, 10:59 AM - 1 Like	#15
UncleVanya Loyal Site Supporter Join Date: Jul 2014 Photos: Gallery \| Albums Posts: 28,401	Originally posted by biz-engineer For tremors I would tend to believe that SR help more than af tracking. Both help, the inability to keep the camera pointed directly at the subject is where fancier tracking options helped him. Also the movie capture modes work wonders when kids are scrambling about - at least for him. The tech advances are a boon to some people, but most of us really don't require them. I frequently feel like users are spoiled by the level of sophistication of modern gear.