Strange "Confetti Noise" on Long Time Exposures with Sony A7RII - Any Idea?

[rrccad]

Hi there I need some help.

I`m a newbie in the Sony E-Mount Campus, just received my A7RII last Thursday. In the past I normally shoot with Nikon/Canon Gear, but wanted to give the Sony a try.
I did some 'test' night shoots with long time exposures between 6-30s exposure time, ISO fixed between 100 and 200ISO. After 1h of testing I opened the RAW in Adobe LR CC2015 with the newest RAW Converter Version from Adobe. When I adjust the exposure by just +1-2 stops I see very strong unusual 'confetti noise':



here another screenshot with exposure setting from LR:



I realized that when I push up the exposure around +1-2 stops a very strange 'confetti like color noise appears'. This is mainly the case with exposures longer than >6-8 seconds.

that's honestly a crazy amount of hot pixels.

whether or not it's correctable in post by filtering, I've seen people return cameras for 10 stuck pixels versus .. what's that .. 100,000 to a million maybe? something else must be wrong there. if the thermal design is that bad.. *shakes head*

what would be interesting is ..

is it noise and baked in RAW or actually stuck pixels? take two / three consecutive shots with lens cap on, and see if the spots "move" from shot to shot (ie: random noise from heat) or are they actually stuck on.

 

[Mlenny]

Hi there I need some help.

I`m a newbie in the Sony E-Mount Campus, just received my A7RII last Thursday. In the past I normally shoot with Nikon/Canon Gear, but wanted to give the Sony a try.
I did some 'test' night shoots with long time exposures between 6-30s exposure time, ISO fixed between 100 and 200ISO. After 1h of testing I opened the RAW in Adobe LR CC2015 with the newest RAW Converter Version from Adobe. When I adjust the exposure by just +1-2 stops I see very strong unusual 'confetti noise':



here another screenshot with exposure setting from LR:



I realized that when I push up the exposure around +1-2 stops a very strange 'confetti like color noise appears'. This is mainly the case with exposures longer than >6-8 seconds.

that's honestly a crazy amount of hot pixels.

whether or not it's correctable in post by filtering, I've seen people return cameras for 10 stuck pixels versus .. what's that .. 100,000 to a million maybe? something else must be wrong there. if the thermal design is that bad.. *shakes head*

what would be interesting is ..

is it noise and baked in RAW or actually stuck pixels? take two / three consecutive shots with lens ap on, and see if the spots "move" from shot to shot (ie: random noise from heat) or are they actually stuck on.

i guess it's a thermal thing. this evening i shot 30-40 different images wir different exposure times between 0.5s and 30s. the pixel are not fixed, so they change from image to image...so no dead pixels like on a display... it starts around 2s and gets worses until 30s... with long exposure nr turned off... what i normally do everywhere.. so always some color noise hot pixels.

if yout turn in camera long exposure noise rduction to ON then they are completly gone in the raw file, BUT it takes twice the time as the camera generates a black frame and corrects this issue.

i don't have to enable this long time nr neither on canon nor on my nikon in those exposure times... so... no better solution yet as to wait twice the amount of time if i expose longer than a second, OR use c1 as a raw converter which filters those pixel out,,,

it's not my cam... i could reproduce that with a friend who just got one on saturday with the same effect, so i guess its not a defect of my individual cam...

not good to wait twice the time, especially not for timelapses where this is sometimes just impossible.



--
www.mlenny.com

Loading…

www.facebook.com

 

[GotoDengo]

not good to wait twice the time, especially not for timelapses where this is sometimes just impossible.

Futhermore, LENR also drops the precision to 12-bit. This is somewhat troubling.

 

[J A C S]

i guess it's a thermal thing. this evening i shot 30-40 different images wir different exposure times between 0.5s and 30s. the pixel are not fixed, so they change from image to image...so no dead pixels like on a display... it starts around 2s and gets worses until 30s... with long exposure nr turned off... what i normally do everywhere.. so always some color noise hot pixels.

if yout turn in camera long exposure noise rduction to ON then they are completly gone in the raw file, BUT it takes twice the time as the camera generates a black frame and corrects this issue.

That is strange. If they are really random (form shot to shot), LENR would not do anything. On the cameras I used, hot pixels get stuck for the whole shooting session, and then LENR helps. They disappear after that.

The RAW files shows that those pixels do not have maximal RAW values as typical hot pixels - they are like 7 stops below saturation or so, with different values. I would call them warm pixels.

 

[barrakj]

Guys forgive my ignorance but I don't shoot long exposure shots however I wanted to test this. All I did was set the shutter to 30 and took these shots. Again if this is not what this topic is about please correct me. I do not see anything wrong with the photos and here are links to the RAW files on drop box.

Both were taken with shutter speed on "30.

Loading…

www.dropbox.com

Loading…

www.dropbox.com

 

[ChuckTa]

Capture one has a higher default NR, slide the value zero and the image should look more similar.

Sliders = 0 for both luminance and color has no effect using C1, has nothing to do with default NR. Remaining spots can be automatically removed with the "single spot NR" function in C1 (default = 0)

C1 is a more advanced RAW converter then LR, thats the reason that I switched from LR tot C1 a few years ago.

Any way doesn't mean that it is normal that less advanced converters are showing this hot spots. Sony should look into it.

 

[Tristimulus]

See the fix post. It appears to be simply a case of the bad pixel map not having been updated. Its easy to update by changing the date 2 months forward, then turning the camera off and then back on and correct the date. Its an auto feature to update the bad/hot/dead pixel map that gets subtracted in images not a trait of the a7Rii.

I would not be so quick to decide. A modded 6D would definitely be nice but a7Rii still holds a lot of promise (but a modded camera will always outperform a non modded camera for Milky Way type shots or astrophotography.

Greg.

Well - then there is this star eating thing (heavy spatial filtering in bulb mode - good for a lot of low light photography but very bothersome in the astronomy domain, and no cure).

Several astro photographers here have said it's not really bothersome and pulls out noise that would be accentuated like stars, anyway. The consensus was the 'real star eating' attenuation was fairly minimal. And that's bulb - you can always do multiple 30" or less exposures.

No - multiple 30 second exposures do NOT record the faint stuff I am after. Stacking lots of images does not help if the data is not there. Bulb is my only option.

The so called star eating issue is also reducing detail in astronomical objects. There is no way to reclaim the lost information. Did wait for the A7RII in hope for this issue to be corrected.

Did sit on the fence for the A7RII - but fell off. Just ordered the astromodded 6D and still more than happy with my Sony cameras for anything but the night sky.

Why post here if you're going with a 6D? Best of luck with that sensor.

Just telling that the A7RII is not a very good choice if more serious astrophotography is to be done. Nothing wrong with that I guess.

Also see lots of comments about DSLRs beeing better for fast moving sports and wildlife than the slower focusing mirrorless cameras - but have not seen any complaints about beeing informed about that...

Am using Sony cameras for everything but astrophotography. Guess I still can stay within this forum despite having and using two different camera brands ... and also note that I do NOT use the other camera brand for ordinary photography. That is where the Sonys shine!

 

[J A C S]

Guys forgive my ignorance but I don't shoot long exposure shots however I wanted to test this. All I did was set the shutter to 30 and took these shots. Again if this is not what this topic is about please correct me. I do not see anything wrong with the photos and here are links to the RAW files on drop box.

Both were taken with shutter speed on "30.

https://www.dropbox.com/s/wifrmp9xb4vifi8/DSC00178.ARW?dl=0

https://www.dropbox.com/s/nhng3gbqd26bb1a/DSC00180.ARW?dl=0

I do not see anything similar in your files. This type of noise appears in really dark areas, and you have very few such areas - but they look clean to me.

Did you have LENR enabled?

 

[wictred]

It`s of course a camera issue. what i meant is that with capture one you might 'cure' the spots easier oder better as of today than with acr. The hot pixels ARE there 4 sure due to the camera long exposure without NR... Anyway C1 ist as of today at least a possibility to fix it. A real solution shas to come from sony, if it`s possible to fix that through firmware.

ACR has a separate colour noise reduction, could it be that it's not really working properly for the A7R-II? (Does increasing it reduce the pixels? Or does decreasing it on other A7x models show more pixels?)

 

[Mlenny]

It`s of course a camera issue. what i meant is that with capture one you might 'cure' the spots easier oder better as of today than with acr. The hot pixels ARE there 4 sure due to the camera long exposure without NR... Anyway C1 ist as of today at least a possibility to fix it. A real solution shas to come from sony, if it`s possible to fix that through firmware.

ACR has a separate colour noise reduction, could it be that it's not really working properly for the A7R-II? (Does increasing it reduce the pixels? Or does decreasing it on other A7x models show more pixels?)

Yes, there are separate setting for Color or Luminance Noise in ACR. I played with both of those settings. (I`m using Adobe LR since Version 1 in Mid Summer 2007). Both don`t really help as of today. Those hot pixels will NOT be removed in ACR, some minor spots go away, but far from a solution to clean up a file.

 

[polarth]

I will disagree on the C1 more advance point. As I have recently download the trial version of C1 to compare with LR. The advantage of C1 is it does a lot more default processing than LR, and C1 default is different for each camera (I tried Sony, Fuji Xtran, Samsung NX), so sometime C1 default output is good enough, LR default is same for all camera, need a bit of processing to get good result. But for me, I can't say I could get a better output from C1 over LR. Sometime the image from C1 looks better, sometimes LR, the difference is not large. But I found I can actually extract more details from LR than C1, but C1 seems to sharpens the edges a little better but with more NR required.

So between the two, I would say different rather than better.

C1's camera profiles ( ICC profiles) are better than the one used by LR resulting in more realistic colors and contrast. Very hard to get same result in LR via manual controls.

C1 has mare advanced control tools (color balance, dynamic range recovery etc.) and the local adjustment tools (mask and layers) offers more option then LR mask functions.

There is a difference between handling noise en noise reduction, getting an image with less noise with the same or even more amount of detail that's what counts for me.

--
http://www.polartha.com

 

[wictred]

C1's camera profiles ( ICC profiles) are better than the one used by LR resulting in more realistic colors and contrast. Very hard to get same result in LR via manual controls.

C1 has mare advanced control tools (color balance, dynamic range recovery etc.) and the local adjustment tools (mask and layers) offers more option then LR mask functions.

There is a difference between handling noise en noise reduction, getting an image with less noise with the same or even more amount of detail that's what counts for me.

I completely agree.

Got the insanely cheap Sony Pro version when the Sony/Capture One deal was announced and I had it installed and played around for a while now but never really used it.

Just 2 days ago I edited a whole shoot in LR after updating my AMD graphics driver -> constant crashes until I found out the only option or now is turning off hardware acceleration.
When I was like 90% done I got so annoyed that I gave C1 a try: better results, better performance, much better controls and doing the whole edit again from scratch probably didn't take me longer than just finishing the rest in LR.

 

[barrakj]

Guys forgive my ignorance but I don't shoot long exposure shots however I wanted to test this. All I did was set the shutter to 30 and took these shots. Again if this is not what this topic is about please correct me. I do not see anything wrong with the photos and here are links to the RAW files on drop box.

Both were taken with shutter speed on "30.

https://www.dropbox.com/s/wifrmp9xb4vifi8/DSC00178.ARW?dl=0

https://www.dropbox.com/s/nhng3gbqd26bb1a/DSC00180.ARW?dl=0

I do not see anything similar in your files. This type of noise appears in really dark areas, and you have very few such areas - but they look clean to me.

Did you have LENR enabled?

Yes I had LENR enabled, with it disabled I saw the noise everyone is mentioning here on this topic. Hopefully it will get resolved I really wanted to mess around with astrophotography with this camera. I'll give it a try with LENR and see how the results fair.

 

[Astrophotographer 10]

Did it work?

Another poster implied it didn't fully handle it.

Sensors do develop hot pixels over time as they degrade slightly from things like cosmic ray hits.

But its a slow process. No doubt that's why its a monthly update to catch these new failed pixels.

Greg

 

[J A C S]

Did it work?

Another poster implied it didn't fully handle it.

Sensors do develop hot pixels over time as they degrade slightly from things like cosmic ray hits.

But its a slow process. No doubt that's why its a monthly update to catch these new failed pixels.

Greg

Those are not hot pixels. They are several stops below saturation.

 

[User1929865251]

Hey there,

After seeing your images I was concerned as this is a great camera for so many features. I borrowed one from work this evening and shot these (rather boring) test shots in the back yard. I cannot duplicate the possible 'hot pixel' noise you experienced at all!


30Sec,f/5.6,ISO 100, LENR OFF, back yard light on.


30Sec,f/5.6,ISO 100, LENR OFF, and back yard light off.




121Sec,f4.0, ISO100, LENR OFF




30sec,f4.0, ISO100, LENR OFF, +3.00 EV applied in Lightroom

All of these shot on A7R MKii with LENR OFF, 16-35mmf/4.0 Carl Zeiss @35mm. All shot RAW and opened in Lightroom CC with no other editing. The final image with +3 EV applied was the last in a 70 photo time-lapse series of 30 second exposures, using timer remote, (camera on BULB as some have suggested there may be differences in RAW output in BULB) to see if I could get some heat generated.

I have cropped in 100% and scanned the entire images, i cannot see anything like the 'confetti' you got. Maybe try the pixel re-mapping trick changing the date forward and rebooting the camera.

It was a pretty cold night if anyone is wondering. About 11 Celsius.

So far I'm really happy with the results...It seems as good if not better than Nik D810

Cheers

 

[Astrophotographer 10]

The definition of a hot pixel is a pixel that is brighter than it should be. There is no requirement for it to be at saturation.

If the nearby pixels are all at say 75 and one is at 200 then its noise not signal.

In this case it is most likely thermal noise otherwise called dark current. Its common in sensors and is why high end ccds are cooled to reduce this noise. I see them all the time before my CCD cools off They usually mostly are all gone by about 0 degree C and for sure by -20C.

Greg.

 

[Astrophotographer 10]

I can't say I have noticed the star eating effect as it is called. Have you actually seen it do that or is it really a mild thing? I think it occurs only in bulb and bulb is also only in 12bit.

Greg.

 

[PWPhotography]

I know this is an old thread. But a well known Sony basher brought this up in Canon forum who also in this thread. First time ever saw this thread. Here is my opinion.

This photo was not taken at base ISO 100 but 200 so doesn't leverage the best DR that is true to all cameras not just A7r II.

Still you only added EC. But a correct way should lift shadow first before using EC. Here is my process: +100 shadow, -100 highlight and only need +1.70 EC (now increased to +2.0 EC) to move the histogram into pass the middle, means correctly even a bit over brighten up.

See, what a big difference. Where are hot pixels if you process properly? My old 5D II/III at the same scene, exposed the same and if processed at the same will completely plunge into Confetti garbage.


-100 highlight, +100 shadow, +2.0 EC, all others default in LR

I have tons of photos from A7r II exposed upto 30 sec, and no problem to lift shadows several stops without showing much noise. A7r II has fantastic DR and can take long exposure without much issues. But best leverage is taken at base ISO 100.

--
https://www.flickr.com/photos/55485085@N04/albums
http://pwphotography.zenfolio.com

 

[JimH123]

http://www.dpreview.com/forums/post/56284904

Actually for long exposures, this is not the fix. When bad pixels are mapped, a much shorter exposure time is used by the camera. When you go with longer and longer exposures, the sensor heats up and more and more pixels start to act like stuck pixels and will fill the area with red, blue and green sprinkles.

As an example, here is another manufacturer's camera. An Olympus EM5ii taken for 30 sec at ISO 1600 with the lens cap on. It was shot in RAW and I see that Lightroom immediately shows lots of hot pixels and then a second later, it is far fewer as the default noise reduction kicks in. But for this JPEG, I loaded it into FastStone Image Viewer and noticed that it did not reduce the number of stuck pixels, and then saved as a JPEG. For this image, magnify it, and you will see all the stuck pixels.

For astrophotography, we call this a dark frame and when used with light frames, i.e. the stuff we are trying to image, the stuck pixels are removed from the image.

But this phenomenon is actually very normal and a known behavior of sensors. And this is why the sensors in CCD cameras are chilled so as to reduce this effect.



By the way, I treated the RAW file to DxO Photolab, and it can remove most of the stuck pixels with a slider bar. Where it fails, I zoomed in and saw two to three pixels right next to each other, horizontally, vertically and even diagonally. If it is looking for lone, individual stuck pixels, these are going to pass that criteria.

Shown here is a 100% crop of some stars taken using an Olympus EM5ii with the 12-40mm f2.8 lens at 12mm and f2.8 for 8 sec at ISO 1600. To the left is the original from the RAW file showing noise and some stuck pixels. To the right is the effect of DxO Photolab where Prime Noise removal is used a tiny bit and the Dead Pixel Slider is dragged to the right a tiny bit. You will need to magnify to see the effects. There are plenty of stuck pixels on the left and they are gone on the right.



Adding a second example, This time a older 500mm f4.5 Pentax m42 type lens. Focus is not quite exact, and you can see a bit of a decentering problem. This is a 30 sec shot at ISO 1600 and on the left you can see stuck pixel and on the right, after DxO Photolab, the stuck pixels are gone -- except for one near the bottom center which happens to be two adjacent pixels (vertical), that are not detected as stuck pixels.