Shooting in Dx mode

[Orson]

I have been shooting more in Dx mode lately on my Z9 and 600 Pf.
Positives for me : smaller file size , better AF.
Negatives : smaller FOV
I heard recently someone telling me that Dynamic range is smaller when shooting in DX mode: can anyone confirm this?
When looking at Photons to Photos website , you can see the difference of 1 stop depending on the ISO value.
What is your preference?

Thanks

 

[DRwyoming]

I heard recently someone telling me that Dynamic range is smaller when shooting in DX mode: can anyone confirm this?
When looking at Photons to Photos website , you can see the difference of 1 stop depending on the ISO value.

Yes, regardless of whether you shoot in DX crop mode or crop to DX dimensions in post you lose a bit over a stop of Dynamic Range compared to shooting full frame. As you posted this is easy to see on the DR curves over at PhotonstoPhotos or other places. Basically it comes not from the cropping itself but the subsequent image resizing to display both the uncropped and cropped image at the same output size. If you look carefully at the PhotonstoPhotos DR chart description you'll see they normalize for output image size for that particular chart (Sensor Read Noise charts are not output size normalized).

There's no free lunch but if you need to crop then it doesn't matter from a DR and image shadow noise standpoint whether that cropping is in-camera or in post.

What is your preference?

If I know I'll crop to at least DX dimensions anyway and I don't have a lot of risk of over-cropping and clipping wingtips or having a fast moving subject go out of the frame I'll often crop in-camera. My Z8 and Z9 subject detection, especially eye detection tends to work better when I shoot in DX crop mode and if I'll crop that far anyway then there's not a lot of reason to take home much larger files.

That said, you retain more cropping flexibility in terms of positioning the subject for composition reasons and you do get a bit more insurance against sudden movement or tracking fast subjects or even allowing for more subjects to enter the frame when you shoot full frame and make cropping decisions in post.

 

[SCoombs]

Yes, regardless of whether you shoot in DX crop mode or crop to DX dimensions in post you lose a bit over a stop of Dynamic Range compared to shooting full frame. As you posted this is easy to see on the DR curves over at PhotonstoPhotos or other places. Basically it comes not from the cropping itself but the subsequent image resizing to display both the uncropped and cropped image at the same output size. If you look carefully at the PhotonstoPhotos DR chart description you'll see they normalize for output image size which for that particular chart (Sensor Read Noise charts are not output size normalized).

There's no free lunch but if you need to crop then it doesn't matter from a DR and image shadow noise standpoint whether that cropping is in-camera or in post.

This doesn't make a whole lot of sense to me.

First, cropping an image is not the same as resizing. In fact, the reality is that uncropped photos are getting resized more often than cropped ones. If I take a 45MP image and display it full screen on an HD monitor it has to be scaled down by something like 95% to fit on the monitor. Meanwhile, if I crop a photo to 2MP or so then it doesn't have to be resized to view.

So I'm not quite sure why one would expect cropping to decrease the dynamic range. It's not as if cropping is ever necessitating the image be upscaled to display, and all images need to be resized downward to display.

Second, I'm not sure why resizing would change the dynamic range anyhow. Dynamic range is a measure of how wide a range of variation in light levels a sensor can record. Once the data is recorded, it's recorded and once the RAW data is interpreted to a visual format to display all those light levels are encoded as values in an range-limited array. If a certain pixel has some data value of, say, 250 and another has the same data value at 192, cropping off other pixels from the edge of the frame doesn't change these values to other things.

Put more practically, take any photo on your hard disk. View it at full resolution. Now crop the photo and view the same part of the photo at the cropped version's full resolution. The two will look identical.

 

[ButlerKid]

From Steve Perry...

The Cropping Epidemic - Backcountry Gallery

I see it all the time – lens jockeys hacking off pixels and tossing them out like they were last week’s kitty litter. It pains me to witness such rampant cropping with so little regard for the consequences – especially since I know how potentially devastating it is to a photo. What really makes...

backcountrygallery.com

 

[DRwyoming]

This doesn't make a whole lot of sense to me.

Perhaps this will clear up some of the confusion on why cropping (and subsequent output size normalization) leads to more image noise

In terms of why that impacts Dynamic Range, it comes down to how DR is defined and measured. Some think DR is measured between brightest whites with detail and darkest blacks that still hold detail. But that's not really how it's done. DR is measured between fully saturated whites (255, 255, 255 in 8 bit RGB representation) and a defined level of acceptable shadow noise (e.g. 13dB SNR, 10dB SNR, etc.). That noise level is somewhat arbitrary and folks could define the DR floor as a lower level of acceptable noise or a higher level but still DR is measured as the range between fully saturated whites and a level of acceptable image noise in the darkest parts of the image (shadow areas).

If you simply crop an image and just treat it as a smaller version of the image as in printing it as cutout of the original image or sharing it on the web as a smaller image in pixel dimensions then there's no impact on noise as it's just a smaller portion of the original image. But the more typical use of cropping is to still want the same size output image for web or print use and resize the cropped image to match the uncropped image size. That could be from downsizing the full frame image, upsizing the cropped image or both to achieve the same output image size. In that step you either reduce the visible shadow noise in the full frame image by downsizing (noise averaging) or increase the visible shadow noise in the cropped and interpolated image or both. @Steve's video above describes that quite well.

The concept of output image size normalization for comparison between different media sizes goes back a hundred years or more and is baked into things like how you compare DoF between say an image captured with 8"x10" sheet film and 35mm film. The only way to make an apples to apples comparison that includes things like viewing distance and definition of Circle of Confusion size is to normalize the output images to the same size. That still holds today when comparing images from say a M4/5 sensor to an APS-C crop sensor to a Full Frame 36mmx24mm sensor to medium format and beyond. That assumption of output size normalization is part of every DoF calculator and it applies when making a direct comparison of visible image noise or dynamic range as well.

Bottom line, since DR is a measurement of range between brightest whites and a defined level of acceptable shadow noise and though brights in the final image don't get any brighter but visible shadow noise increases as you crop and then output size normalize, the DR drops as you crop and resize to normalize the output image sizes. If you read the notes on the PhotonstoPhotos DR charts or read through their methodology you'll see that output size normalization is part of how those charts are created.

 

[nmerc_photos]

I very rarely, if ever use DX mode. I don't have anything against it - and when I shot canon I used crop mode a lot of the time.

But as I'm getting better at approaching my subjects, getting them comfortable with me, etc. I find I can stay further back and fill more of the frame - so there's less need for a crop.

DX/Cropping in post achieve the exact same thing - so there's no reason to hesitate using it.

 

[SCoombs]

If you simply crop an image and just treat it as a smaller version of the image as in printing it as cutout of the original image or sharing it on the web as a smaller image in pixel dimensions then there's no impact on noise as it's just a smaller portion of the original image. But the more typical use of cropping is to still want the same size output image for web or print use and resize the cropped image to match the uncropped image size. That could be from downsizing the full frame image, upsizing the cropped image or both to achieve the same output image size. In that step you either reduce the visible shadow noise in the full frame image by downsizing (noise averaging) or increase the visible shadow noise in the cropped and interpolated image or both. Steve's video above describes that quite well.

My point is that unless you're using a 20 year old camera the uncropped image size is already so large that for most print or web use you're not resizing the cropped image to match the uncropped size.

If I have a 45 mp image and view it on an HD monitor, I'm downsizing it by 43 mp. If I crop the image first by half, I still have an image with 20mp more than can be displayed and I downsize by 20mp when viewing it.

If I make an 8x10 print of that image at 300 ppi I again have to downsize by 38mp. If I crop by half and make the same print, I also downsize by 13mp. Either way, I'm taking my image and reducing it, not trying to resize the smaller one to match the size of the larger.

In print, you CAN get to the point where you're doing that if you print very large sizes at high ppi, but otherwise you're not trying to resize a cropped image to equal uncropped. For web, I can't imagine you're almost ever doing so.

This comes up on occasion when discussion about cropping comes up but I just can't imagine a situation where a person is cropping an image and then trying to match the size of the original.

 

[DRwyoming]

My point is that unless you're using a 20 year old camera the uncropped image size is already so large that for most print or web use you're not resizing the cropped image to match the uncropped size.

You should really watch Steve's video.

The point is comparison between the full frame image and the cropped version. To view those at the same output size, whether for print or screen use you'll either downsize the full frame image (reducing it's visible noise through noise averaging), upsize the cropped image (increasing it's visible noise) or some of both. No matter how you slice it, once you normalize the output image size for the same camera settings on the same sensor the cropped image will show higher noise and from that lower dynamic range.

If you're displaying say a web sized image for both, the Full Frame image will have more downsizing and thus more noise reduction than the crop image and again it will show lower noise for the same ISO setting.

Yes, there are plenty of spare pixels to allow cropping but once the two images (FF vs crop) are compared at the same output size the FF image will show less visible noise and this shows up in the dynamic range charts as a difference for the same sensor shot at the same ISO with different levels of crop. It's why for every camera that supports a crop mode you'll see a bit over one stop lower dynamic range when the camera is shot in an APS-C crop vs Full Frame.

 

[RikWriter]

I don't have a Nikon, but I use the crop mode on my Sony A1 and A7RIV all the time for the same reason as you, that I know I'm going to crop in post anyway and it gives me more room both in the memory card and on the buffer. All the noise and DR difference stuff is small enough that I'm confident I can take care of it in post.

 

[DRwyoming]

If I have a 45 mp image and view it on an HD monitor, I'm downsizing it by 43 mp. If I crop the image first by half, I still have an image with 20mp more than can be displayed and I downsize by 20mp when viewing it.

Here's an example of posting to this website with a 2000 x 1332 pixel size, one shot full frame and one as a DX crop of the same image.

This was shot pre-dawn at very high ISO and it's the same photo just one was cropped to DX pixel dimensions before downsizing to the web post size and the other was left at Z8 Full Frame dimensions prior to the downsize to the same web posting dimensions. Posting a fixed size to a website like this is one example of where two images would be normalized to the same output size. Output size normalization does not mean that the smaller image has to be increased though when printing large prints that can happen, it just means the final output size is the same in pixel dimensions regardless of whether one was downsized, both were downsized by differing amounts, one or both were upsized, etc.

Anyway, here's the two images with no other processing beyond the RAW import defaults. IOW, no noise reduction, no exposure, contrast, added sharpening, color or other edits just a DX crop extracted from the full frame image and then both downsized to the same web post dimensions:

You can only see EXIF info for this image if you are logged in.

You can only see EXIF info for this image if you are logged in.

They're both pretty noisy at this ISO but at least on my monitor the cropped image is noticeably noisier.

This is the same effect that's measured on a lab bench to produce the dynamic range charts and regardless of what size is chosen for the normalized comparison size the cropped image will always show more noise and less dynamic range due to the size normalization when the rest of the camera settings like ISO, shutter speed and light remain the same.

 

[Orson]

Here's an example of posting to this website with a 2000 x 1332 pixel size, one shot full frame and one as a DX crop of the same image.

This was shot pre-dawn at very high ISO and it's the same photo just one was cropped to DX pixel dimensions before downsizing to the web post size and the other was left at Z8 Full Frame dimensions prior to the downsize to the same web posting dimensions. Posting a fixed size to a website like this is one example of where two images would be normalized to the same output size. Output size normalization does not mean that the smaller image has to be increased though when printing large prints that can happen, it just means the final output size is the same in pixel dimensions regardless of whether one was downsized, both were downsized by differing amounts, one or both were upsized, etc.

Anyway, here's the two images with no other processing beyond the RAW import defaults. IOW, no noise reduction, no exposure, contrast, added sharpening, color or other edits just a DX crop extracted from the full frame image and then both downsized to the same web post dimensions:

View attachment 86373

View attachment 86374

They're both pretty noisy at this ISO but at least on my monitor the cropped image is noticeably noisier.

This is the same effect that's measured on a lab bench to produce the dynamic range charts and regardless of what size is chosen for the normalized comparison size the cropped image will always show more noise and less dynamic range due to the size normalization when the rest of the camera settings like ISO, shutter speed and light remain the same.

This is a relevant difference! it might be a good idea not to use Dx in camera when Iso is too high.At normal Iso , the difference is less pronounced

 

[DRwyoming]

This is a relevant difference! it might be a good idea not to use Dx in camera when Iso is too high.At normal Iso , the difference is less pronounced

That's a point Steve makes in the video linked above, if you do shoot or crop to DX mode or beyond then keep an eye on ISO and perhaps tradeoff shutter speed or aperture to keep ISO lower when heavy cropping is anticipated.

The difference in stops is the same, but a one stop effective difference from say ISO 400 to ISO 800 may not even be noticeable but the effective change from say ISO 12,800 to ISO 25,600 can be dramatic.

 

[SCoombs]

Here's an example of posting to this website with a 2000 x 1332 pixel size, one shot full frame and one as a DX crop of the same image.

This was shot pre-dawn at very high ISO and it's the same photo just one was cropped to DX pixel dimensions before downsizing to the web post size and the other was left at Z8 Full Frame dimensions prior to the downsize to the same web posting dimensions. Posting a fixed size to a website like this is one example of where two images would be normalized to the same output size. Output size normalization does not mean that the smaller image has to be increased though when printing large prints that can happen, it just means the final output size is the same in pixel dimensions regardless of whether one was downsized, both were downsized by differing amounts, one or both were upsized, etc.

Anyway, here's the two images with no other processing beyond the RAW import defaults. IOW, no noise reduction, no exposure, contrast, added sharpening, color or other edits just a DX crop extracted from the full frame image and then both downsized to the same web post dimensions:

View attachment 86373

View attachment 86374

They're both pretty noisy at this ISO but at least on my monitor the cropped image is noticeably noisier.

This is the same effect that's measured on a lab bench to produce the dynamic range charts and regardless of what size is chosen for the normalized comparison size the cropped image will always show more noise and less dynamic range due to the size normalization when the rest of the camera settings like ISO, shutter speed and light remain the same.

I've seen Steve's video in the past. For whatever it's worth, on my monitor both photos look pretty similar and most of what I see is color noise which should be relatively easy to remove, but I realize that's beside the point. I agree that generally speaking you'll see less noise in non-cropped photos.

I think the real "disagreement" here is that when someone says "dynamic range" I don't think of noise. I realize that they are related and that it's common for people to talk about the two in the same breath, but when I think of them I think of things that are related but distinct.

Here is an HDR composite image I made a while back along with a few of the originals. I don't remember at this point which originals I actually used so these may not be the ones I actually used.

You can only see EXIF info for this image if you are logged in.

You can only see EXIF info for this image if you are logged in.

You can only see EXIF info for this image if you are logged in.

The reason I did a composite in this case wasn't really because I was thinking about noise. It was because I knew that if I exposed sufficiently that you could actually see the kids, the sky would be totally blown out. I realize that there's a relationship here with noise because if I were to increase the exposure on the very dark shot you can actually brighten the kids up enough to see, but they are going to be noisy.

Still, the way I think about this has to do with whether I can see things that are dark - literally whether there is any luminosity to the darkest areas ,not whether the dark things are noisy or not - and whether there is detail in the bright areas. As such, I can imagine a photo which is very noisy but which I'd say has better dynamic range than another one that is not very noisy because in the first I can see details in the bright areas and in the dark areas there are is luminosity rather than just black. If you'd originally phrased it purely in terms of noise, I'd have agreed without a second thought that the cropped image will have more noticeable noise.

 

[DRwyoming]

I think the real "disagreement" here is that when someone says "dynamic range" I don't think of noise. I realize that they are related and that it's common for people to talk about the two in the same breath, but when I think of them I think of things that are related but distinct.

The term may be used differently by different folks but in signal processing and engineering it's well defined and in the case of dynamic range charts like those produced over at PhotonstoPhotos the term is very definitely related to the visible noise floor in an image.

But we're getting way off topic of the OPs question on cropping and DX mode. Yes you can extend the dynamic range of an image through HDR image composites but that's another conversation relative to cropping, image size normalization and what that means in terms of noise and dynamic range of a single image capture.

 

[EricBowles]

Take a look at this link comparing a Z9 to a Z( with a DX crop. You can choose any other full frame camera if you prefer.

Photographic Dynamic Range versus ISO Setting

You can see that the relative noise level of a DX crop means roughly 1.1 stops more noise - or lower ISO. What I mean by that is you need to use 1.1 stops lower ISO for a DX crop to have the same effective noise as a full frame image. The uncropped image has the benefit of downsizing which effectively reduces noise. This assumes a common output size.

That does not mean you should not crop - just that there is an impact. If you can use a longer focal length and fill the frame, that's preferred to cropping from a noise perspective. If you need to crop anyway - cropping in the camera or in post won't matter.

There is also the impact of final output. If you are downsizing an image for posting on social media, it effectively reduces apparent noise. So you can have a very noisy image, and by downsizing to 1000 pixels on the long edge almost all the noise goes away. If you cropped to 1000 pixels rather than downsizing, it might be unusable in comparison. Likewise if you are upsizing, noise become a greater factor and you likely need noise reduction of other efforts to manage or reduce noise.

 

[Chris K]

“Don’t crop, because it reduces image quality” implies that you have some other means to get the framing you want.

First, frame the image you want, by any means necessary. The image quality you end up with is based on how much lens you’re willing to carry.

For wildlife, you can say ”the 800PF is less noisy than the 500PF”, because the 500PF requires cropping or a TC to get the same framing. The TC increases ISO due to effective aperture, and cropping increases scale of the file. Both reduce image quality.

 

[Kuau]

I have always wondered if DX mode really does help AFC mode on the the Z9
I can only conclude in theory if you use all area AF mode then the camera could take advantage of a smaller area to deal with yet I have not seen any actually evidence to support this.

 

[Chris K]

I imagine it helps area AF, cause it’s not looking at as much clutter. I haven’t noticed better eye tracking or anything like that, however.

One thing is for sure: it sure helps me!

 

[bleirer]

I have always wondered if DX mode really does help AFC mode on the the Z9
I can only conclude in theory if you use all area AF mode then the camera could take advantage of a smaller area to deal with yet I have not seen any actually evidence to support this.

It's easy to prove it to yourself. I was skeptical with my R5 so I tried an experiment. I used a human subject because they would stand still for me. I kept walking backward (with eye detection set to people), until I got far enough away to start losing the eye. Then I switched to crop mode and continued to walk away. I was now able to detect the eye again. I forget the exact result, but if I recall it added about 1/3 more distance to what could be tracked.

 

[apakett]

My point is that unless you're using a 20 year old camera the uncropped image size is already so large that for most print or web use you're not resizing the cropped image to match the uncropped size.

If I have a 45 mp image and view it on an HD monitor, I'm downsizing it by 43 mp. If I crop the image first by half, I still have an image with 20mp more than can be displayed and I downsize by 20mp when viewing it.

If I make an 8x10 print of that image at 300 ppi I again have to downsize by 38mp. If I crop by half and make the same print, I also downsize by 13mp. Either way, I'm taking my image and reducing it, not trying to resize the smaller one to match the size of the larger.

In print, you CAN get to the point where you're doing that if you print very large sizes at high ppi, but otherwise you're not trying to resize a cropped image to equal uncropped. For web, I can't imagine you're almost ever doing so.

This comes up on occasion when discussion about cropping comes up but I just can't imagine a situation where a person is cropping an image and then trying to match the size of the original.

When when downsizing an image you are not throwing away pixels as you do when cropping. Those extra pixels are combined to lower noise and increase dynamic range.

 

[Kuau]

When when downsizing an image you are not throwing away pixels as you do when cropping. Those extra pixels are combined to lower noise and increase dynamic range.

I don’t think it has anything to do with noise or dynamic range

 

[ButlerKid]

I have always wondered if DX mode really does help AFC mode on the the Z9
I can only conclude in theory if you use all area AF mode then the camera could take advantage of a smaller area to deal with yet I have not seen any actually evidence to support this.

It does. Even Steve Perry has commented on this many times. Focus acquisition is helped because the subject is much larger in the frame.

 

[Kuau]

It does. Even Steve Perry has commented on this many times. Focus acquisition is helped because the subject is much larger in the frame.

Understood yet which focus modes benefit from DX mode? It would make no sense if the answer was all of them

 

[ButlerKid]

Understood yet which focus modes benefit from DX mode? It would make no sense if the answer was all of them

Why not? Even when using single point which doesn't have subject detection or tracking, putting the point exactly where you want it would be easier on a larger target.

 

[Kuau]

Why not? Even when using single point which doesn't have subject detection or tracking, putting the point exactly where you want it would be easier on a larger target.

I was just referring to any of the subject detection modes. For the user experience yes smaller focus area means easier to navigate yet was curious with subject detection
Of one used small area large area custom area all area one would think that there should be no difference except having a smaller focus area to move the box around