5D IV underexposure banding

Someone else, somewhere in this thread, talking about astrophotography, did a black lens cap test and pushed it. I can't find it now but you could check it out.

GaryJP said:

Someone else, somewhere in this thread, talking about astrophotography, did a black lens cap test and pushed it. I can't find it now but you could check it out.

Click to expand...

I've just tried the same test. Here's a shot I just took at ISO 100, 15s, with the camera body cap on (i.e. no lens attached). This JPEG has been pushed 5 stops in Lightroom, everything else was set to zero (no sharpening, NR etc). There's some banding visible but not much, plus a few hot pixels. Note that up until about a 3.5 stop push I can barely see the banding at all, so looks to me like there's something wrong with the OP's camera.

Thanks for your replies!

i am leaning toward the same conclusion: my 5D and some others have a defect that causes banding not seen in other new 5D's

does the shutter speed really need to be 1/10,000 for this shot? Some rather poor samples from dpreview...

jakdaniel975 said:

As canon is stingy and does not put the spot metering linked to the af points if you make a reportage and speed you can not use manual mode and aperture priority uses such as the picture against happens the erring exposure ... then in these cases need not retrieve a shot underexposed in laboratory tests !! then look at all the other cases where it serves high dynamic range in the examples I have posted ... you expose for highlights and then raise the shadows ... you saw what banding is out and big yellow stripes? ... I say that everyone has praised recoveries 5 stop when not needed, but in situations where it is needed can not be used for banding ... then I say that it seems strange that dpreview did not notice this !!! and I would understand if all mk4 5d have this defect.

Click to expand...

--
A7R2 & 5D3

jakdaniel975 said:

dpreview gallery

Click to expand...

The transitions on the yellow area in the original image look bad ...

This is my one week old 5D4. Exposed for 15 sec with viewfinder cap on, lens cap on, shot at f4 with 16-35mm ISO 100, AWB. Shot RAW, processed in Capture One with Exposure +4, Shadow pushed at 100%. Should I return it or is this acceptable and expected at such extremes. I would not expect such correction necessary in any real life shooting situations. These three shots were taken one after the other with exactly the same settings, the only difference was the the zoom. One was shots at 16mm and the other two at 35mm. None of this is visible when shooting JPEGs pushed the same way. It simply might be not enough shadow info in JPEGs for this banding to manifest.

Maybe you are taking things too far. I haven't got Capture One. Therefore I could only emulate your test with ACR and my 5ds with the same exposure settings. Pushed by +5 and shadows maxed out in ACR I get some fine red banding. Much finer then in your images but then the 5ds has 50mpx. Someone else has to comment on the stronger red bands you are seeing but they are probably quite normal too at such extremes and vary in position from image to image. If they occurred in the same position every time I'd be more concerned.

You could of course go one step further and leave the camera out in the rain for 24 hours and see what sort of banding you'd get then. :-D

--
Machines were mice and men were lions once upon a time. But now that it's the opposite it's twice upon a time.

marshwader said:

Maybe you are taking things too far. I haven't got Capture One. Therefore I could only emulate your test with ACR and my 5ds with the same exposure settings. Pushed by +5 and shadows maxed out in ACR I get some fine red banding. Much finer then in your images but then the 5ds has 50mpx. Someone else has to comment on the stronger red bands you are seeing but they are probably quite normal too at such extremes and vary in position from image to image. If they occurred in the same position every time I'd be more concerned.

You could of course go one step further and leave the camera out in the rain for 24 hours and see what sort of banding you'd get then. :-D

--
Machines were mice and men were lions once upon a time. But now that it's the opposite it's twice upon a time.

Click to expand...

I agree. This is not a real life scenario, in fact I stated that in my post. It is purely a demonstration of what the sensor is doing when pushed beyond its capabilities. Probably totally useless but in line with what some other 5D4 owners have reported in this thread. I guess it just confirms that my camera is not the only one doing this and therefore I dare to say it is typical for 5D4. Reason for alarm or concern?? Probably not☺. For the short few days I've had it, I can only attest that this camera is performing truly admirably in a real life shooting situations.

Some of the tests may feel like extreme limits of the camera, and I agree. But, don't lose sight of the fact that I started this thread because I, and some others, have noticed our cameras exhibit banding with a 2.5 - 3 stop push. That's problematic, and isn't so extreme. Moreover, some folks aren't seeing their new 5D's exhibit banding with a 2.5 - 3 stop push. Therein lies the issue.

What's the harm in replacing it and doing the test again?

mmarian said:

This is my one week old 5D4. Exposed for 15 sec with viewfinder cap on, lens cap on, shot at f4 with 16-35mm ISO 100, AWB. Shot RAW, processed in Capture One with Exposure +4, Shadow pushed at 100%. Should I return it or is this acceptable and expected at such extremes. I would not expect such correction necessary in any real life shooting situations. These three shots were taken one after the other with exactly the same settings, the only difference was the the zoom. One was shots at 16mm and the other two at 35mm. None of this is visible when shooting JPEGs pushed the same way. It simply might be not enough shadow info in JPEGs for this banding to manifest.

Click to expand...

--
Once you've done fifty, everything else is iffy.

marshwader said:

Maybe you are taking things too far. I haven't got Capture One. Therefore I could only emulate your test with ACR and my 5ds with the same exposure settings. Pushed by +5 and shadows maxed out in ACR I get some fine red banding. Much finer then in your images but then the 5ds has 50mpx. Someone else has to comment on the stronger red bands you are seeing but they are probably quite normal too at such extremes and vary in position from image to image.

Click to expand...

It may be "normal" in the sense that it is a real likelihood that a camera could have this, but it isn't necessary to have this. This is not something caused by pushing. It is a flaw, exposed by pushing. It is an easily correctable one, too, but don't hold your breath waiting for your favorite converter to do so in an update.

marshwader said:

If they occurred in the same position every time I'd be more concerned.

Click to expand...

Why? That gives one more way to correct it. It would go away with LENR enabled. Unfortunately, commercial converters generally do not allow precise corrections of known artifacts, even if they are quite simple.

marshwader said:

You could of course go one step further and leave the camera out in the rain for 24 hours and see what sort of banding you'd get then. :-D

Click to expand...

Hardy har har not.

There is no reason not to hope that a camera does not do things like this. How much noise are we seeing down there, otherwise? What is the character of the noise, otherwise? It is pretty good. These shadows would be usable, expanding the modes of operation of the camera. I don't know about you, but I am not happy just for digital cameras to replace film. I want them to defer choice of ISO exposure index until making the final image in processing. Yes, you get a bit more noise shooting an intended or discovered ISO 3200 from the "ISO 100" setting, but the increased noise in the common tonal range may be acceptable in light of the extra 5 stops of headroom you get, which can capture things like the color of lights that would otherwise blow out in the wrong color.

And don't forget, some camera settings automatically "under-expose". If you shoot a very fast lens wide open, the sensor gets less sensitive and the camera compensates by pushing the RAW levels under the hood. IOW, if you shoot in Av mode at f/8, and something white averages 8000 DN (RAW pixel value), but your lens is an f/0.9 and you open it all the way and take the same shot, the level of white might drop to 5000 instead of 8000. The camera thinks that the levels should be corrected, though, and multiplies the RAW values all by 1.6, losing 2/3 stops of DR, and raising the "noise floor", relative to metering, by 2/3 stop.

Also, if you use HTP (which I almost always do), that raises the "noise floor" by 1 stop, while giving a stop more highlight headroom, transparently with Canon JPEGs and converters.

Then, the 1/3-stop ISOs don't use 1/3-stop analog gain; they manipulate the RAW data just like the very fast f-numbers do, with scaling of RAW values, and loss of original digitized tonal range in the 125/250/500 series.

So, if we do all these things at the same time, f/0.9 and ISO 250 with HTP, the noise floor is a full two stops higher than it is at ISO 100 and f/8, relative to metering. "Plus", we have lost 1 stop of DR, so "exposing to the right" is not a solution, because one stop of the "right" has been discarded.

I did a test with my MKiv by underexposing a 100 iso shot by 3 stops and opened the raw file in ACR and increased the exposure by 3 stops. I saw no banding but if I increased to 4 stops then I noticed slight banding in the shadows and a lot more when pushed by 5 stops. I did the same test with my MKiii and the banding and noise were considerably worse when pushed 3 stops and detail was lost in the shadows. The MKiv is much, much better than the Mkiii.

Here is an example from my testing.

larger image normal exposure







larger image with +5 stops







100% crop normal







and 100% crop with +5 stops









I do see a little bit of horizontal banding, but I dont think it is as bad as the OPs. what do you guys think? I mean, this IS pretty extreme. I dont think i will ever do +5 stops. but it would be nice to have that insurance policy. heh.

mmarian said:

I agree. This is not a real life scenario,

Click to expand...

That's such a stagnant ideology.

mmarian said:

in fact I stated that in my post. It is purely a demonstration of what the sensor is doing when pushed beyond its capabilities.

Click to expand...

Actually, it is showing a flaw that the system has. A flaw that need not be there, perhaps, except that they needed to produce the camera in a time frame, and the only way they could avoid it at this point in time was to drop 1 fps off of the max burst speed, or some other issue or compromise. Or, they didn't even notice it. It is conceivable that the engineers work "blind" on the shadows, only measuring noise globally and mathematically, but never really looking closely at it (or they do, and their managers are not concerned). Regardless of the reason that it is there, there is no reason that a converter shouldn't be able to fix it, especially DPP, since it is Canon's camera.

I can fix it in software, and I have never taken a course in camera engineering, or programming, so the "big boys" are failing big by my standards.

mmarian said:

Probably totally useless but in line with what some other 5D4 owners have reported in this thread. I guess it just confirms that my camera is not the only one doing this and therefore I dare to say it is typical for 5D4.

Reason for alarm or concern?? Probably not☺. For the short few days I've had it, I can only attest that this camera is performing truly admirably in a real life shooting situations.

Click to expand...

Compared to what you are used to, this banding may not show until pushed to a point where an earlier 5D-series camera would have copious random noise. Canon has, in general, been moving away from banding noise, especially this simple type that should be easily corrected, so some people are kind of expecting to get full use of the improved DR of the new ADC design. The 5D4 takes some of that back, though; you can't go way down in the shadows and accept greater random noise as a cost for the exposure latitude or highlight headroom; you get strips of color offsets in the mix.

ivanku said:

Whether something is in focus or not doesn't affect exposure, or the possible need to lift it / lift shadows in an OOF area.

For example, what if you use a lens that vignettes heavily, such as the 24-70 II, and correct corner vignetting and want to lift background, out of focus, shadows by one stop? That sequence of edits will constitute a 3 stop exposure lift. It's not routine, but it's not at all crazy.

Click to expand...

By the way, I was not questioning your top level point. I agree whole heartedly about wanting more DR to lift shadows, etc. I try not to create the situations that require that edit but there are plenty of legitimate situations that benefit from the DR. I have a 5DSR so the challenge is more pronounced for me.

But depending on the lens, you can see a lot more color aberrations in OOF areas. Furthermore, there is plenty of color "blobbing" for lack of a better word when something is out of focus. I just think it would be better to check something that is in focus for a control test. And was pointing out that the studio tests done by dpreview are all in focus.

And the vignette issue is real. I have several lenses that vignette heavily too (11-24/L, 16-35/2.8LIII, Sigma 20/1.4 Art, 40STM, etc.). This compounds the problem a lot and can be up to 3 stops just by itself.

This is an excellent thread. I am interested to see if this is a general problem or related to your copy. The images I was commenting on where adjusted by more than 3 stops, though. The exposure was increased by 2.9 stops globally and the shadows where lifted by 57 out of 100 on top of that. Not sure if there was also lens correction for vignette-ing.

also, I wonder if this noise is more apparent with base ISO. i have some iso 200 shots, and when bumped +5 stops, i dont see any banding at all.



here they are





+5 stops

I'll check at ISO 200 later today. Could you perform the same test at ISO 100? I think the banding does not affect all new 5D's, and I bet your camera is one of the banding-free bodies

ilumo said:

and 100% crop with +5 stops



I do see a little bit of horizontal banding, but I dont think it is as bad as the OPs. what do you guys think? I mean, this IS pretty extreme. I dont think i will ever do +5 stops. but it would be nice to have that insurance policy. heh.

Click to expand...

OK. I'm convinced. It is much better than the images you first posted. Lots of noise but no patterns to speak of. I suppose anyone who buys a MK IV should do similar tests as soon as they receive their cameras to confirm proper function.

So ... at this moment it looks like only some cameras are affected?

Correct. We need to make sure that we are performing tests and comparing images taken at the same ISOs, but so far it's looking like only some 5D's are affected.

What may be causing this is over my head. A sensor issue? Noise further down the signal chain? Too much gain from the in-camera processing? I'm out of my depth.