Phase Detect Auto-Focus (PDAF) pixels visible on T6s Sensor

Started May 1, 2015 | Discussions
bclaff Veteran Member • Posts: 8,427
Phase Detect Auto-Focus (PDAF) pixels visible on T6s Sensor

This photo:

Canon T6 Sensor Issue (lensrental.com/blog)

which is the subject of a separate thread regarding the spots also clearly shows horizontal lines where the PDAF pixels are located.

Not a defect. Just kinda cool.

-- hide signature --
Canon EOS Rebel T6s (EOS 760D / EOS 8000D)
If you believe there are incorrect tags, please send us this post using our feedback form.
msowsun
msowsun Contributing Member • Posts: 655
Re: Phase Detect Auto-Focus (PDAF) pixels visible on T6s Sensor

bclaff wrote:

This photo:

which is the subject of a separate thread regarding the spots also clearly shows horizontal lines where the PDAF pixels are located.

Not a defect. Just kinda cool.

"Lensrentals writes that the issue affects 4 T6s sensors out of 10, and 2 T6i sensors out of 10"

It is not cool when they show up as little bulls-eye spots in the photos

If it is not a defect, then why does it only appear on some cameras?

 msowsun's gear list:msowsun's gear list
Canon PowerShot S110 Canon EOS 5D Mark III Canon EOS Rebel SL1 Canon EOS 80D Canon EF 70-200mm F2.8L IS II USM +23 more
OP bclaff Veteran Member • Posts: 8,427
Re: Phase Detect Auto-Focus (PDAF) pixels visible on T6s Sensor
2

msowsun wrote:

bclaff wrote:

This photo:

which is the subject of a separate thread regarding the spots also clearly shows horizontal lines where the PDAF pixels are located.

Not a defect. Just kinda cool.

"Lensrentals writes that the issue affects 4 T6s sensors out of 10, and 2 T6i sensors out of 10"

It is not cool when they show up as little bulls-eye spots in the photos

If it is not a defect, then why does it only appear on some cameras?

You are posting in the wrong thread.
That is a defect, the visibility of the PDAF is not.

-- hide signature --
msowsun
msowsun Contributing Member • Posts: 655
Re: Phase Detect Auto-Focus (PDAF) pixels visible on T6s Sensor

bclaff wrote:

msowsun wrote:

bclaff wrote:

This photo:

which is the subject of a separate thread regarding the spots also clearly shows horizontal lines where the PDAF pixels are located.

Not a defect. Just kinda cool.

"Lensrentals writes that the issue affects 4 T6s sensors out of 10, and 2 T6i sensors out of 10"

It is not cool when they show up as little bulls-eye spots in the photos

If it is not a defect, then why does it only appear on some cameras?

You are posting in the wrong thread.
That is a defect, the visibility of the PDAF is not.

I can see the faint horizontal lines but I don't think they are PDAF pixels. The info from Canon doesn't mention horizontal lines.

What makes you think the horizontal lines are PDAF pixels?

 msowsun's gear list:msowsun's gear list
Canon PowerShot S110 Canon EOS 5D Mark III Canon EOS Rebel SL1 Canon EOS 80D Canon EF 70-200mm F2.8L IS II USM +23 more
snofox Regular Member • Posts: 334
Re: Phase Detect Auto-Focus (PDAF) pixels visible on T6s Sensor

The T6s doesn't have PDAF, does it?

 snofox's gear list:snofox's gear list
Canon PowerShot G1 X Canon PowerShot G12 Canon EOS 60D Canon EOS 700D Canon EOS 80D +6 more
msowsun
msowsun Contributing Member • Posts: 655
Re: Phase Detect Auto-Focus (PDAF) pixels visible on T6s Sensor

snofox wrote:

The T6s doesn't have PDAF, does it?

The T6s and T6i both have Hybrid CMOS AF III which is a form of PDAF built into the sensor. The 70D and 7D Mk II have Dual Pixel CMOS AF.

 msowsun's gear list:msowsun's gear list
Canon PowerShot S110 Canon EOS 5D Mark III Canon EOS Rebel SL1 Canon EOS 80D Canon EF 70-200mm F2.8L IS II USM +23 more
OP bclaff Veteran Member • Posts: 8,427
Re: Phase Detect Auto-Focus (PDAF) pixels visible on T6s Sensor

msowsun wrote:

I can see the faint horizontal lines but I don't think they are PDAF pixels. The info from Canon doesn't mention horizontal lines.

What makes you think the horizontal lines are PDAF pixels?

My experience with stacking frames to do Fixed Pattern Noise (FPN) analysis tells me they are almost certainly PDAF.

The T6s (also known as (AKA) 760D "feature a new Hybrid CMOS AF III system for drastically improved AF speed and accuracy. By increasing the density of pixels dedicated to phase-difference AF towards the horizontal direction" (from Enhanced AF Performance for Easy Capture of Fast-Moving Subjects , only one of several references I could cite )

-- hide signature --
iso rivolta Contributing Member • Posts: 550
Re: Phase Detect Auto-Focus (PDAF) pixels visible on T6s Sensor

bclaff wrote:

msowsun wrote:

I can see the faint horizontal lines but I don't think they are PDAF pixels. The info from Canon doesn't mention horizontal lines.

What makes you think the horizontal lines are PDAF pixels?

My experience with stacking frames to do Fixed Pattern Noise (FPN) analysis tells me they are almost certainly PDAF.

The T6s (also known as (AKA) 760D "feature a new Hybrid CMOS AF III system for drastically improved AF speed and accuracy. By increasing the density of pixels dedicated to phase-difference AF towards the horizontal direction" (from Enhanced AF Performance for Easy Capture of Fast-Moving Subjects , only one of several references I could cite )

We were just discussing this issue on the EOS M forum, here . There is no need to stack frames for fixed pattern noise analysis. You can just use a RAW file with close to overexposed areas and bring the levels down using a RAW editor that doesn't correct the AF pixels. For example this CR2 file from imaging-resource (the T6s gallery) can be opened with IrfanView (with the latest DCRaw-based plugin) and then use Gamma correction. You'll get something like this (downscaled crop showing two double lines):

OP bclaff Veteran Member • Posts: 8,427
Re: Phase Detect Auto-Focus (PDAF) pixels visible on T6s Sensor

iso rivolta wrote:

bclaff wrote:

msowsun wrote:

I can see the faint horizontal lines but I don't think they are PDAF pixels. The info from Canon doesn't mention horizontal lines.

What makes you think the horizontal lines are PDAF pixels?

My experience with stacking frames to do Fixed Pattern Noise (FPN) analysis tells me they are almost certainly PDAF.

The T6s (also known as (AKA) 760D "feature a new Hybrid CMOS AF III system for drastically improved AF speed and accuracy. By increasing the density of pixels dedicated to phase-difference AF towards the horizontal direction" (from Enhanced AF Performance for Easy Capture of Fast-Moving Subjects , only one of several references I could cite )

We were just discussing this issue on the EOS M forum, here . There is no need to stack frames for fixed pattern noise analysis. You can just use a RAW file with close to overexposed areas and bring the levels down using a RAW editor that doesn't correct the AF pixels. For example this CR2 file from imaging-resource (the T6s gallery) can be opened with IrfanView (with the latest DCRaw-based plugin) and then use Gamma correction. You'll get something like this (downscaled crop showing two double lines):

Did you realize that you're reviving a thread that's over one year old?

In any case, you are incorrect in stating that Fixed Pattern Noise (FPN) analysis can be accomplished using a single image.

By definition FPN is a pattern of noise that is fixed and you can only determine if the observed pattern is fixed by comparing multiple frames.

Furthermore one component of FPN is Dark Signal Non-Uniformity (DSNU).
You cannot observe DSNU from inspecting a single "lens cap" image.

The other major component of FPN is Photo Response Non-Uniformity (PRNU).

Computing DSNU and PRNU requires a statistical analysis of multiple frames.
I use a stacking technique that requires only 16 black and 16 illuminated frames, far fewer than the most common method.

My technique also produces interesting visualizations of FPN.

Respectfully,

-- hide signature --

Bill ( Your trusted source for independent sensor data at http://www.PhotonsToPhotos.net )

iso rivolta Contributing Member • Posts: 550
Re: Phase Detect Auto-Focus (PDAF) pixels visible on T6s Sensor

Sorry, I didn't suggest that FPN can be obtained from one image. My intention was to show how the PDAF pixels can sometimes be seen by simple processing of a single image. The thread here might be one year old, but was interesting and seemingly no conclusion was reached at the time. The cameras in question are still current models.

Keyboard shortcuts:
FForum MMy threads