Auto-focus sensors: high precision vs. regular testing

Started Apr 4, 2010 | Discussions thread
Flat view
mmullen Veteran Member • Posts: 4,280
Auto-focus sensors: high precision vs. regular testing

My testing shows that the high precision sensors are not as adept at getting a reading. In otherwords, often the regular sensors get a focus lock but not the high precision sensor but the image is taken anyway (without the HP sensor). The standard precision AF sensor is much more reliable.

One of my AF tests is available in my "lens test gallery" here on DPR. The test was designed to measure whether a lens was able to use the HP AF sensor under moderately low light on a terrycloth bathtowel with a brightly colored geometric print pattern. Light was a north facing window with white translucent window shades proving a light level of approx. EV 6 (over six stops above Canon's minimum threshold specification for required light level to AF).

I used an f/4 lens (70-200 f4 IS) to test the regular AF sensors and an 85mm f1.8 prime lens to see if it could focus more consistently by utilizing the HP focus sensors. The camera was perpendicular to the terrycloth bathtowel at a distance of 5.5 feet and the bathtowel was large enough to fill the entire frame with a planar surface. I took ten exposures with each lens run, resetting the focus to infinity on 5 of the exposures and to MFD on the other 5 exposures. The 7D's center AF point was used on One Shot AF mode and both lenses could consistently lock focus. There were no instances of the lens pausing, hunting or completely missing focus. The camera was tripod mounted and a remote shutter release was used.

This is what the full-frame test target looked like:

Canon EOS 7D, 89 mm, ISO 100, 1/4 sec, F4.0, ±0.00 EV, Apr 3 2010 14:08:42

The 85mm f1.8 was shot at f4, 1/5 sec and ISO 100

The f/4 zoom was shot at f4, 1/4 sec and ISO 100 (EXIF reported a zoom setting of 89mm)

I also did a run of 10 images with the 85mm wide open (f/1.8) but did not compile statistics on that test.

The shots were taken in the largest/finest JPEG available and 1500 square pixel crops were taken from the center of the frame.

RESULTS: The f4 zoom could achieve more consistently spot on AF than the 85mm f1.8 by a good margin (even when they were both shot at f4). This was verified visually and by filesize comparisons. JPEG filesizes indicate the amount of detail an image has. More sharply focussed images will be larger than slightly blurry ones.

The following statistics are based upon the twenty center crops taken at f/4 (10 from each lens):

Average filesize:
zoom: 16656 KB
prime: 15672 KB

Standard Deviation of filesize:
zoom: 14.7 KB
prime: 84.4 KB

Largest filesize:
zoom: 1687 KB
prime: 1680 KB

CONCLUSIONS: The very similar largest filesizes indicate that the 70-200 f4 IS is about equal in resolution to the 85mm f1.8 prime lens when the prime is stopped down to f4 and the zoom is wide open at 85mm and both lenses are well focussed. This agrees with the test crops at

PRIMARY CONCLUSION: The large amount of deviation on filesizes indicates that the prime lens focus was all over the board. This tells me the high-precision AF sensors were not being utilized to increase the accuracy beyond what the regular precision AF sensors could achieve with the f4 zoom. The reason for this could be that the light level was too low or the subject did not offer enough contrast. Perhaps the terrycloth loops were too fine for the HP AF sensors to get a reading. Likewise, the small amount of filesize deviation for the f4 zoom tells me that the AF on this lens was very consistent. These conclusions are also supported by visually inspecting the center crops.

-- hide signature --

Mike Mullen

 mmullen's gear list:mmullen's gear list
Canon PowerShot S80 Canon PowerShot G11 Canon EOS 40D Canon EOS 7D Canon EF 85mm f/1.8 USM +8 more
Flat view
Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow