One of the big headlines when the first Nikon Z6 and Z7 tests appeared online is that they exhibited a pattern of “banding,” or line pattern noise, when recovering shadows too much. Although most reviewers were quick to point out that 5-6 stops of shadow recovery is excessive, this issue nonetheless drew a lot of attention. But now that we have performed complete dynamic range tests on today’s mirrorless cameras at Photography Life, my question is… why?
Let’s start with an example. The following image has been underexposed by six stops at base ISO 100. I opened it in RawDigger, then exported as a TIFF and recovered the image with the “Exposure” slider in Capture One. As you can see, there are indeed visible lines of noise, although only in certain parts of the image. I’ve circled the most obvious patterns:
However, just because the Nikon Z6 has these patterns of noise doesn’t mean that…
Ah, sorry – it looks like the image above isn’t from the Nikon Z6, but from the Canon EOS R.
Okay, here’s the image I meant to show. Once again, I underexposed this photo by six stops, recovered it in RawDigger and Capture One, and then circled the offending noise. Note that this is a 100% crop:
This time, there is noticeably less line pattern noise, although it’s still visi—
Oh, that’s the Sony A7 III. My bad.
All right, I’m kidding around. But the two examples above go to show that line pattern noise isn’t a Nikon-only factor. The Canon EOS R has a decent amount, and the Sony A7 III has a bit as well. I’ll cut out the act now and show how the actual Z6 sample image compares:
No tricks, that’s the Z6 image. The most noticeable thing to me is that it appears greener than the other photos – but that’s something you can correct with a small white balance adjustment. In terms of line pattern noise, though, it has perhaps a slight bit more than the A7 III, and definitely less than the Canon EOS R. Certainly not what you would expect based on the articles online.
Just for good measure, here is a sample image from the camera that sparked this discussion, the Nikon Z7:
I’d say the Z7 is the best of the lot. This is largely due to the lower base ISO of 64; the other cameras here would likely show similar performance if they offered such a low native ISO. Although you can see a bit of line pattern noise in some of the gray color swatches on the right, it is far from objectionable.
Here are the same four images one more time so that you can click and compare without the circles covering them. Note that I downsampled the higher-resolution photos so that they all appear the same size for comparison:
I definitely don’t begrudge initial reviewers for writing about the Z6’s and Z7’s line pattern noise when the shadows are boosted too much. It might not be the most real-world of tests, but part of this job is figuring out how these cameras fundamentally work, and—
Cough, cough! Who underexposes images by six stops?? Ahem.
Sorry, something in my throat.
Anyway, the Z6 and Z7 certainly aren’t perfect in terms of line pattern noise. However, I can’t help but feel some surprise that this issue became so strongly associated with them, yet is hardly mentioned in discussions of other mirrorless options today. After all, the root cause of these banding issues is generally due to on-sensor phase detection pixels, which are part of the autofocus system in nearly all modern mirrorless cameras – making this more than just a single company’s problem.
So, how did it happen that the Z cameras took most of the heat for this type of noise? A large part of it is due to the trickle-down nature of camera reviews. One or two big websites publish a set of high-quality tests; several others copy the pro/con page and hammer out a watered-down version of the original review. You can see obvious examples on any website that lists “banding” as a negative for the Z6 or Z7, yet not in their review of the EOS R. It means the website most likely did not test these cameras side by side, but simply took existing information and repackaged it with a few unique sample images. It’s the quick, easy way to write a review, and I’m guessing you’ll see more and more of them in the coming years.
That doesn’t mean it is a bad idea to read about small differences like this, but you also shouldn’t take everything you hear about cameras today at face value. Look at sample images for yourself, and try to figure out whether you’ll see any of these issues in your own images. At least in the case of line pattern noise, you almost certainly won’t; we’ve taken thousands of photos with all these cameras and never once seen the issue in real-world shooting. In short, there are many more important things to care about when buying a camera today, like whether it was manufactured in Thailand or Japan, and what color the camera is.
I honestly believe that Sony is promoting the spread of misinformation like this through their marketing arm to boost their sales and minimize their competitors.
Amen. Well said. I grew tired of hearing this complaint. I’ve been shooting with a Z7 for nigh on 2 years and about 10,000 images. Never once have I noticed any banding. Then again, I never grossly under-expose images. But I do recover a lot of shadows that are considerably under-exposed. Again, no banding, and very little noise. The image quality from the Z7 is simply fantastic.
Now that the LCD is used dominantly for the image framing, adjusting and capturing of photos on these new mirrorless cameras, we’re seeing a downside of immediate final image preview… now the user can crank their exposure compensation low (or higher) until they get the image they want to capture. As someone who shoots product photography, I often enjoy shooting -2.0 EV as that makes the blacks look the blackest. With immediate preview on the LCD, I’m encouraged even more to crank the exposure down further since I can actually see it getting closer to perfect in real life. I’ve seen this problem on my Z6 (with just shooting -2.0 EV and NOT PULLING IT LIGHTER, LEAVING IT DARK AS SHOT) and it’s true the banding is a disappointment. Even without “pushing” shadows it’s visible.
Topaz DeNoise AI makes this a non issue. It will clean up any noise or banding without degrading the sharpness or image quality one bit. I am not affiliated with Topaz Labs in any way other than I pay for and use their products. If you don’t have it you are missing out.
Sorry but I notice the banding in almost all shots even correctly exposed. With most cameras. But not all eyes are the same. Also those samples are worst than real world shoot some sunsets and take another look.
DPReview’s review of the Nikon D780 included: “With the Nikon Z6 and Z7, on-sensor phase detect brought about unwanted banding in the shadows of very darkest tones shot with these cameras: the D750, with no on-sensor phase detect, never had this issue. With the D780, Nikon has done a better job processing out its on-sensor PDAF points compared to its mirrorless siblings.”
It seems Nikon have taken note of the concerns raised about banding. Result!
…Nikon has taken note, Nikon is a singular noun. Is English a forgotten language?
Thank you for fact-checking this and the excellent humor. Now I feel more confident that I am sticking with Nikon and when my Nikon D750 dies I’ll move to a mirrorless Nikon :).
If you look hard enough you can find fault in just about anything these days if you try hard enough. You know what they say about people’s opinions right? LOL! I enjoyed this article and the light hearted humor. Cheers!
Thank you for pointing out that the Sony A7 III shows line pattern noise, and reminding us that Canon’s sensors for the last several years generally have inferior dynamic range at low ISOs than their competitors. But “Cough, cough! Who underexposes images by six stops?? Ahem.” seems to imply there is no benefit in high dynamic range sensors, which I can’t agree with. Several years ago in Yosemite we were on the side of a steep valley, shaded from the afternoon sun by the ridge above. The valley bottom was a forest of dark conifers, but in the valley mouth was a vista with a waterfall in full sunshine. Then using a Nikon D90, I took my usual bracketed set, noticed the highlight warning flashing on the monitor for the waterfall and snow, reduced the exposure compensation, saw the highlight warning was still flashing, etc, and eventually lost the warning with -4EV exposure compensation! In the JPEG, the forest formed a solid black mass. I was delighted Photoshop Elements’ Adobe Camera Raw was able to recover significant detail, but would have liked to have done much better. I guess trying to retain colour in a setting sun within a landscape could lead to even more extreme dynamic range.
As for Nasim’s “ if a scene has that much dynamic range, most of us would be doing bracketing for HDR or luminosity masking anyway”, those Yosemite shots led me to try various free HDR programs, expecting to find something that would do a better job, but none were able to align the hand-held Yosemite shots. No doubt a better HDR program would have got the result I wanted. I’m sure HDR is a great work-around when your camera lacks the dynamic range to record what you can see in an extreme high-contrast scene, but I’d far rather have a camera that could capture enough detail in a single exposure for me to lighten the shadows to my taste when processing the image.
Perhaps banding should be only a minor issue if deciding whether to buy a Z6 or Z7 or not, but they’d be better cameras if they offered the cleaner, more usable dynamic range of the D750 and D850 (although I admit that the conifer foliage in the dark areas of my Yosemite shot would probably have camouflaged banding effectively). I hope reviewers will continue to publicise weaknesses such as these, to influence the manufacturers to try to resolve them. If it isn’t possible to avoid banding from on-sensor phase detection pixels through improvements to the in-camera image processors, perhaps other manufacturers should follow Canon’s route of dual-pixel AF.
Honestly, the Nikon showed the least banding, I think the Sony had finer or more compressed lines, making them appear less obvious but they’re no better. It would be interesting to see what they would look like on a solid color background.