Nikon Z6/Z7 as a UV/IR Conversion: the Bad News

[Andrea B.]

The PDAF (phase detect auto-focus) banding issue in the Nikon Z6 or Z7 has been reported by Birna to be particularly bad in her Z6 UV/IR conversion when shooting reflected UV photos. I want to provide some background for this topic before Birna posts her examples.

 

First, let's note that labeling this banding as "PDAF banding" may be a misnomer because it occurs also when using Z6//Z7 manual focus or when using Z6/Z7 CDAF (contrast detect auto-focus). On the Nikon Z6/Z7 you can set pin-point AF to force CDAF. But note that other mirrorless cameras which use only CDAF (example: Panasonic S1R) do not have any striping artifacts reported.

 

The PDAF banding in the Z6/Z7 is *not* a function of the sensor. It is rather due to the camera's data processing which attempts to fix the well-known problem of PDAF striping due to reflections from the PDAF pixels' metal masks. (The older Sony As had this PDAF striping, for example. Along with Sony lossy compression and shutter shock, it is why I dislike my Sony A7R. But nevermind that because I learned to use it anyway.)

 

The various well-known techies and bloggers such as Marianne Oelund, Jim Horshack, Jim Kasson and others (see links given below under the bar) who have been investigating the Z6/Z7 banding have observed that high dynamic range scenes and/or brightly illuminated, low ISO scenes can often induce the banding. And if the banding is there, then shadow lifts make it look worse.

 

Well, my dears, those are precisely the kind of conditions we have in reflected UV photography. UV photos typically have UV-dark and UV-bright areas which require a camera capable of wide dynamic range. And we like to use as low ISO as possible in UV work to avoid the noise typically associated with UV shots. And shadow lifts or exposure pushes are common during processing because UV shots are rather dark and we tend to underexpose them.

 

Please note that so far these folks have not been able to pin down precisely what causes the PDAF striping algorithm to engage. And there is no fix yet for Z6/Z7 PDAF banding. I'm not sure there will be anything soon because the commenters seem to think that the typical Z6/Z7 shooter is not going to encounter the PDAF banding very often. The fix is to simply use another camera for those situations where it might occur. Not much of a fix to my way of thinking. Especially when we cannot be sure precisely what triggers the data processing algorithm.

 

To mitigate Z6/Z7 banding:

• Use high ISOs because noise masks banding.
  
• Use the 12-bit setting because noise masks banding.
  
• Try the repair in the Raw Therapee converter. It's not perfect, but helps.
  

Frankly, I think that using noise to mask banding is not something I want to do. We have enough noise already in UV photos.

 

I had really, really liked the Z6 conversion which I pre-tested for Birna because I could easily focus at f/4 through the EVF or LCD under the BaaderU. And the Z6 has a nice short FFD. And Focus Peaking worked a treat in UV. But my use of high ISO for my test photos obscured just how bad the PDAF banding is.

 

We may have to unconvert Birna's camera. Geez!!

 

My earlier test link:

http://www.ultraviol...s-finally-here/

 

---

 

Here are some references on other forums for the topic of Z6/Z7 PDAF Banding.

A lengthy discussion was made on Dpreview in the Z forum:

• https://www.dpreview...s/post/61781820
  
• https://www.dpreview...s/post/61811583
  

Then Jim Kasson made several blog posts about the Z PDAF Banding:

• https://blog.kasson....f-banding-faqs/
  
• https://blog.kasson....7-pdaf-banding/
  
• https://blog.kasson....7-pdaf-banding/
  
• https://blog.kasson....ding-constancy/
  

[Andy Perrin]

Oh no! I know both of you were looking forward to that camera so much. :/

[otoien]

One thing that might make banding show up more easily in invisible spectrum photography is the often extreme white-balancing that has to be performed. So even if extreme stretching or exposure compensation in post is not performed, if one channel is very underexposed compared to the others, it is more likely to show up.

 

My experience is from the D7100, which has banding of different origin, but the same principle should apply, as the banding lurks down in the deep shades in both cases. During the recent lunar eclipse, I shot though light polluted clouds, and when the light pollution was subtracted, the banding showed up in a considerable amount of frames, likely originating in the underexposed blue channel. (This event contributed strongly to my decision to upgrade the D7100 to a D500.)

[dabateman]

Birna,

I am sorry that the conversion didn't work. I had high hopes for that camera after Kolari tested the Z7. Its too bad the sensor is misbehaving.

Please post some samples, I would like to get a feel for what your seeing and if there may be any possible work arounds.

 

[Andrea B.]

Again, it is not the Z6/Z7 sensor, per se, which causes the banding. It is a problem in the software algorithm which attempts to mitigate artifacts caused by PDAF. So by fixing one kind of "striping", we get intermittent "banding".

 

I've read that the new Canon mirrorless also has a banding problem, but I've not looked into that so I don't know if Canon banding is the typical mirrorless PDAF striping or due to some attempt to fix that like Nikon did.

 

There are no workarounds at this point. If you look through the referenced links, you will find that folks have tried about everything as a workaround to no avail. The only thing which can mask this banding when it occurs is using high ISO noise or 12-bit noise. Or to try the fix provided in the Raw Therapee converter.

 

It is beginning to look like mirrorless is not the way to go for reflected UV at this time. I'm very disappointed myself.

 

Øivind, that is a good point about the hit on data quality incurred by a large white balance adjustment on a UV file. I've seen that cause some maze-like artifacts over the years. Usually it helps to expose UV to the right and keep UV exposures as short as possible with really good UV illumination for a clean file which can handle large WB changes a bit better.

[Andy Perrin]

It is beginning to look like mirrorless is not the way to go for reflected UV at this time. I'm very disappointed myself.

This seems like a massive exaggeration, Andrea. I know you are disappointed but keep a sense of proportion. Many of us have been using mirrorless cameras for years without noticing these artifacts.

[Andrea B.]

I been trying mirrorless conversions since 2009.

But I've never yet gotten the image quality I seek from one of my mirrorless conversions. Noise, PDAF artifacts, shutter shock, lossy compression, flimsy mounts, poor dynamic range, EVF blackout, badly implemented focus peaking, poor Live View, banding/striping. Been through it all. Wondering when are they ever going to get it all right?

 

It is safe to say here that I am definitely too much of a "purist" on the topic of image quality. :D :D :D

[dabateman]

Andrea,

Buy a full spectrum converted Olympus Em1, the first one not mk2 with a Panasonic not Sony sensor from Kolari. I think they still even have new ones. Ask for the dust shaker to be removed. Except for the menu and odd Olympus naming convention, I think its the perfect UV camera. There is no AR coating on the sensor coverglass, and UV transmission is excellent.

 

As for Birna's camera. I would like to see images of the problem. Ideally can 4 image be taken after each other of the same subject? If the banding is fixed it would be a problem. But if there is enough random behavior, you could use the median stack feature in Photo Afinity to clear it away.

[DaveO]

I stopped using the Nikon D7100 as I kept getting banding in UV images when I tried to open up the shadows. Perhaps I was being a bit too piccy but I got banding in the dark areas where I had never seen it with the Pentax K5 conversion.

 

Dave

[Cadmium]

I have no experience with converting a camera myself, so I have a limited understanding of all that might be involved or could undermine the quality of the conversion,

but I am just wondering if something may be amiss in the conversion process.

Can anyone think of any aspects of the conversion that would create such a problem?

[otoien]

I stopped using the Nikon D7100 as I kept getting banding in UV images when I tried to open up the shadows. Perhaps I was being a bit too piccy but I got banding in the dark areas where I had never seen it with the Pentax K5 conversion.

 

Dave

 

The D7100 are among the worst bodies out there with respect to pattern noise, and the banding is much uglier and stronger that what we have seen from the Z-bodies. But it is lurking in the deep shadow, not in the lights, so for many types of regular daylight photography it is a quite good body for most purposes. Also at ISO 1600 for astro photo at low temperature in a light polluted red zone it was not much of an issue. It is just that I lately started to feel that I could not trust it for the reasons mentioned above. When the discussion started on banding in the Z bodies, I had a little déjà vu of the discussion that happened after D7100 was released, in both directions, some bashing it, on the other side bagatellization. The truth was in my experience somewhere in between.

 

It is worth to note that Nikon pretty much solved this problem in the next iteration, the D7200.

 

I have forgot where I found it, but there was a recent comparison of PDAF related banding in the different recent mirrorless bodies. They all showed it, and the Z-bodies were not the worst. It is just that the Z bodies got a lot more attention than others.

[nfoto]

I'm swamped with social obligations at present thus not much time for testing the behaviour of the modified Z6 right now. However, playing around with it showing the banding is present at all ISO settings albeit less pronounced at the higher ISO values. For the low ISO banding is clearly visible even in the in-camera jpg and preview. Various software handle the NEFs and banding slightly differently. Apparently, Capture NX-D suppresses banding the best whilst Photo Ninja, Silkypix and RawTherapee are much less forgiving. RawTherapee ever so often chokes on these NEFs when they are done in UV and simply gives up making sense of them. Very strange as this program behaved very well with anything thrown at it before.

 

The banding occurs for shadows as well as highlights meaning it is well-nigh impossible to mask out properly.

 

An example (with NX-D, 1600 ISO) to show a striped sky in UV. Photo Ninja did 10 times worse.

 

 

The UV captures themselves are strange as in having very odd colour casts and I'm far from convinced that LifePixel did remove the internal shaker?

[Andrea B.]

Birna, thank you for this striking example of banding in UV from the converted Z6. You didn't mention your lens or filter, but I know from PM conversations that you were testing with the UV-Nikkor and the BaaderU.

 

I also note in your example that using the high ISO 1600 does not provide enough masking noise to cover that banding! I think we can dispense with that "solution".

 

I will check with Life Pixel about removal of the shaker, or not. These days shakers can be left in if also modified. Life Pixel is too professional to have made a rookie error with a shaker. But we will still follow up with them.

[nfoto]

UV-Nikkor+BaaderU is correct. The NEF (1600ISO) processed straightforward in Capture NX-D latest version.

[Andrea B.]

Well, Birna, we need to figure out what to do. I think perhaps a refund & reconversion from Life Pixel is in order. But sending the camera back from Norway is going to be a pain. Maybe we should try for a refund and you can have the Z6 restored by your local Nikon shop (which I remember fondly! Nice guys there.)

 

There was really no way for us to know that PDAF banding would be so bad in UV given its sporadic occurence and mild appearance in unconverted Zs and other mirrorless with striping/banding problems.

 

I'm STILL embarassed that my testing, brief though it was, did not uncover the extent of the Z problem in UV. But this won't be the first time that an unpredictable phenomenon has been missed. Sigh.

---

 

I have no experience with converting a camera myself, so I have a limited understanding of all that might be involved or could undermine the quality of the conversion,

but I am just wondering if something may be amiss in the conversion process.

Can anyone think of any aspects of the conversion that would create such a problem?

 

The things which can go wrong in a conversion are not such that they would induce an electronic or software banding problem. Bad conversions are usually due to such things as:

• trapped dust (All too common in the past. Much better these days.)
  
• the wrong replacement filter or filter material being used (plastic - NO!)
  
• poorly manufactured replacement filters (ZBW - NO!)
  
• insufficient sealing around the replacement filter
  
• faulty re-adjustment of the focal plane (skewed or wrong distance)
  
• And, of course, there is simple physical breakage of a board or some internal piece.
  
• Or an electronic frying out of some board (which I've done twice!!!)
  

In all the conversions I've made or made for Birna over the years, I've seen one trapped dust problem, one loose screw problem, and one plastic filter replacement (Spencer's). The trapped dust and loose screw were in a D5200 conversion by Life Pixel a few years ago. Never seen that before or since from them.

 

The major "failure" by conversion shops is that they seem to fail to test a converted full spectrum camera model in all three wavebands. The IR-LED shutter monitor fiasco should never have happened. And I'm thinking that this Z6 fiasco should not have happened, but it is sporadic.

[Andy Perrin]

Yeah, that’s pretty bad. I have never seen that issue with my A7S in UV or anywhere else. I don’t doubt that it happens sometimes but it has not been a huge problem for me. The biggest image quality issue I have with A7S is amp glow at very high ISO.

[GaryR]

Disappointing news for sure. Lifepixel is well-known to remove the dust system in ALL conversions. I don't know if they ever changed that policy, but I had a previous conversion done at Spencers Camera for that very reason. Kolarivision are great, and always specify if the dust system can be saved. I also noticed that they do caution D700, D750, D3, D3S, D3X, D4, and Z7 owners, of the IR internal shutter monitor issue for UV shooting. My last conversion at maxmax did not go well, and probably should have gone with Kolarivision from the beginning...oh well, live and learn.

[Alaun]

Will the camera still be useable for IR?

[bobfriedman]

The banding occurs for shadows as well as highlights meaning it is well-nigh impossible to mask out properly.

 

An example (with NX-D, 1600 ISO) to show a striped sky in UV. Photo Ninja did 10 times worse.

 

 

The UV captures themselves are strange as in having very odd colour casts and I'm far from convinced that LifePixel did remove the internal shaker?

 

clearly this is present in an unmodified camera as well... in sensor PDAF's time hasn't come, I'm afraid.

[otoien]

Sorry to see this Birna, pretty bad, as this is not only banding lurking down in the dark as I described for D7100 above, assuming that the image was not underexposed and extremely compensated in post. Even the "bad" D7100 will not show the banding at ISO 1600 unless one goes to very extreme measures. Could this be some interaction of the PDAF mask with the UV wavelengths so that the firmware compensation works even less as intended? It certainly would be interesting to see the results with proper filtration in the IR and visible band.

 

One could also wonder if the sensor vs camera firmware's adjustment of the PDAF compensation was wrong to begin with, in other words a defect camera sample. Then reverting it to visible light only would of course not help.

 

This is also of interest for the astro photo community, as full spectrum bodies or hydrogen alpha modified bodies are fairly commonly used. I will keep my eye out for this regarding Z6 in the two dedicated fora I participate in.

[nfoto]

IR test photos appear much less impacted. UV @100 ISO is clearly the worst of all situations.

 

I have used Z6 and Z7 quite extensively for about 6 months now, under a wide range of shooting conditions, and never seen anything like the banding this new modified Z6 delivers. I simply was not prepared for the problem when it emerged in the new camera.

[nfoto]

Here is an example of the 100ISO banding. The Colorchecker Passport is deliberately slightly defocused to make the pattern stand out better.

 

(100ISO, UV-Nikkor f/8, 1.5 sec, Baader U, modified Z6)

 

 

The example shown has been run through Photo Ninja, but in this case even the Capture NX-d couldn't improve much.

[Dmitry]

For example, if a pattern repeats every 16 lines, can you just discard every affected line and slightly reduce the image?

[nfoto]

The same crop from D3200 (internal Baader U), UV-Nikkor 105mm f/4.5 at f/8, 1.6 sec ISO 100. Photo Ninja. D3200 (internal Baader U) and D600 (full-spectrum modified, using external Baader U) are essentially identical in the final outcome. There are no artefacts of any kind.

 

[nfoto]

Downscaling by a suitable factor will of course "smear" detail and by definition also reduce banding, but the issue will still persist. In the situations when files need to be enlarged the banding makes for an even worse problem.