Are these PDAF lines, or something else? PDAF filter does nothing



My friend has a Sony 6500 and has these horrible lines across the image when he takes pictures at night. I know that RT v5.5 has a filter for PDAF, so I figured I’d try it out, Except it did nothing. But perhaps these are not PDAF lines but some other malady?


Just to add some more info - I did look at the image 1:1, and I did export to JPEG, just to make sure it wasn’t the preview that wasn’t showing the filter, and it looked the same in both.


You mean the vertical bands? Those aren’t PDAF. Please provide a raw file and pp3.


OK, the raw file is here, Yes the horizontal they’re quite visible in the lower part where there is snow on the street, and upper left. I uploaded the .RAW

NS_01701.ARW (23.8 MB)NS_01701-1.jpg.out.pp3 (10.9 KB)


what do you think the lines are then?


PDAF looks like this


The banding is there even in the raw file. I suggest you use wavelets or frequency separation to deal with it. In the image below, the banding is apparent in the “residual” and “s0” scales.

(Glenn Butcher) #7

Did a little searching, found this, look at the sixth post:

I downloaded your .arw and inspected it with exiftool; the shutter speed, 1/200sec, might be a contributor. I’m willing to bet you have some sort of LED lighting in the scene that is modulated at a frequency close to your shutter speed. LED lights can be modulated at a wide variety of frequencies for various purposes, not just 50/60Hz corresponding to utility AC power.

I think slower shutter speeds can mitigate it, but with the wide variety of situations you can encounter, you might have to experiment to make sure you’ve selected one that mitigates the frequency. For the pictures you already have, constructing a layer mask in GIMP to pull whichever band you choose to the other is feasible.


Wow, great sleuthing and could answer this question of the lines! Thanks for your insights everyone!

(Flössie) #9

When a colleague of mine had a similar issue, I found this to be the reason.


(Ilias Giarimis) #10

Probably the electronic shutter was used so the scan speed of the sensor is around 1/20sec … for a speed of 1/200 the 1/10 of the sensor is exposed to light

the bands you see come from the interaction of flickering light with the scan speed and width