Should Nikon Z9 shooters avoid ISO 400?

[agrumpyoldsod]

I have posted a new blog entry titled Should Nikon Z9 shooters avoid ISO 400?

To get to the punchline - "Obviously the answer to the question "...." is NO, but if you find yourself needing to shoot at ISO levels between 300-400 slightly better quality results might be achieved if you try moving the ISO setting up to ISO 500 and see if there is a difference - the data indicates there will be." The Z9 uses a dual-gain approach to its signal processing which "dramatically" impacts the Dynamic Range and Signal to Noise levels at the point it "kicks in" (~ between ISO 400 and ISO 500).

Please have a read and provide POV in this thread.

 

[tomfranic]

Thnx 4 the info

 

[Patrick M]

Great article. Though I understood little of it. Doh….

 

[bleirer]

That is interesting. I notice Canon r5 has an almost identical thing, except the bump is at 400 so shooting at 400 is better than 200 or 300 but 100 is still higher. Sony a1 also similar, bump at 500. So I'm wondering if the dual gain is something that is more common than not in modern cameras at least at a certain price point?

 

[arbitrage]

Been shooting at the dual gain ISO value or higher since my D500 days.
All the cameras I've owned since then (D850, A9, A9II, A7RIV, R5, Z9 and A1) have all had a dual gain sensor with pretty much ISO invariance from there on up. Since I'm usually always in need of more shutter speed I rarely drop my ISO low enough to get to the values (Usually ISO 100-200 but depends on the camera) where it makes sense to drop ISO below the dual gain value.

 

[DRwyoming]

Please have a read and provide POV in this thread.

Dual gain architectures do give that jump in dynamic range but it's not something I worry about in actual shooting. It's a discussion closely related to all the angst that Nikon shooters had when the D5 came out with its less than stellar DR performance at low ISO but above its dual gain bump better performance at high ISO settings. The thing is that gain bump occurs in most cameras at a fairly low ISO setting at that also means at a fairly high point on the DR curve.

So yeah, if you shoot the Z9 at ISO 500 you'll get around 9.5 stops of dynamic range (give or take based on the noise floor SNR definition) and if you shoot just below the gain bump at ISO 400 you'll get roughly 8.7 stops of DR (again depending on the noise floor definition). But either is more than sufficient for a high quality print or display image assuming it was well shot to begin with and won't need something like several stops of shadow pull. in my experience high quality prints might have up to five stops of DR depending on the paper and the process and most monitors when viewed in anything but a darkened room might have a stop or two more DR but neither typically exceed 8 stops of DR unless there are ideal viewing conditions. That's basically why we can get away shooting at sky high ISO these days where shooting a camera like a Z9 at ISO 6400 and its associated sub 6 stops of DR is a walk in the park and with good processing the camera can be shot upwards of ISO 10,000 and 5 or fewer stops of DR.

When the D5 came out a lot of folks had trouble with the DR curve that below ISO 1000 or so hovers in the 8 to 9 stops of DR region but in practice the D5 delivers great images when shot at low or high ISO and of course its big advantage happens at high ISO where it outpaces many if not most cameras as the ISO climbs. Sure it would be great to see the low end of its curve up in the 10 to 12 stop region but for well exposed photos in decent light we can't really see the difference in prints or even most monitor use in anything but ideal viewing environments.

To me the gain bump is significant in another way as when you get above it you typically move into the ISO Invariant region of the ISO/DR curve and that has good implications to shooting and processing decisions. IOW, cameras aren't ISO invariant in the region around that bump where a small shift of ISO in post is very different than the same small shift of ISO in-camera but get above that gain bump and most modern sensors (really overall camera architectures as the gain bump is in the camera's signal processing not strictly in the sensor) become ISO invariant. That point is useful to know for folks like astro photographers that might choose to pull exposure up in post or folks who shoot with a bit of negative exposure compensation for insurance against clipped highlights and find themselves adjusting exposure in post.

So yeah, if you're shooting a tricky image where you know you'll need all the dynamic range you can get such as tough lighting with deep shadow detail that will need to be pulled and bright highlights that also need to be protected (I'm thinking midday outdoor wedding shooters struggling with hard shadows and bright highlights and not filling with flash or similar scenarios) then maybe shooting above the bump makes more sense than shooting just below the gain bump but shooting even lower (e.g. base ISO up to ISO 200) in the Z9 would make even more sense if you'll need all the DR you can get.

Just my 2 cents...

-Dave

 

[Tibor Hrnko]

What Dave says, just above my post is a valid observation for ISO invariance. Plus following are my observations based on Sensor Analysis Primer on "Photons to Photos" site.

• Important thing to remember is that photographic dynamic range is mainly concerning the recovery of shadows. If there are highlights in raw file close enough to bit 14, they will be clipped by increasing ISO.
  • by definition mentioned in its Engineering and Photographic Dynamic Range section PDR charts are using signal to noise ratio (SNR) 20 instead of SNR 1 as reference point - effectively cutting off low 4 bits of 14 bit raw and considering them black/indistinguishable from random pixel noise
  • based on the Photon Transfer Chart for Z9 I would guess that Z9 is not using blacked pixels to remove random noise in NEF files (as it does not converge to 0 but to 1 - compare with D850) and assumes that it is low enough to have no significant impact on the photographic dynamic range for scenes that are bright.
  • Increased analogue gain at ISO 400 assumes that there are no highlights available at that point that will be clipped by amplification, if there are any in the scene, the picture will be worse than the one on base ISO IMO as there will be less noise in shadows but at the expense of clipped highlights.
• I would recommend to use that ISO 400 analogue gain increase if scene is flat enough to fit into 9-10 bits of dynamic range because it will improve shadows retrieval.
• I would recommend to underexpose and use base ISO for scenes where highlights are high and important, because a viewer will notice the burned highlights more than the underexposed shadows. But the picture after postprocess will be more noisy if shadows recovery is attempted in tools like LR or PS.

 

[nautiboy]

A few years ago I ran across an interesting (to me) article on the dual gain on the D850 and ISO invariance. After reading it and reviewing all of the samples, I came away with the following conclusion (which might be the conclusion from the article, I don't remember, I just have these conclusions in my notes file ... ). While most probably know this, just in case - the D850's extra gain starts at ISO400 whereas the Z9 is at ISO500, so adjust the directions accordingly.

1. If that ISO is 64 or as much as one stop over, set the ISO to 64 and make the exposure. If that ISO is 400 or as much as three stops over, set the ISO to 400 and make the exposure. As you crank the ISO incrementally up from 64 or 400, you’ll get less highlight protection, and it won’t gain you anything significant in shadow noise.
2. If that ISO is between 160 and 250, decide if you need the last bit of shadow signal-to-noise ratio (SNR). If you don’t, set the ISO to 160, take the shot, and get some extra highlight protection.
3. If that ISO is 320, decide if you could do with bit less highlight protection. If no, set the ISO to 320. If yes, set it to 400 and enjoy lower super-deep-shadow SNR.
4. If that ISO is between 400 and 2500, and you think you’ve got the highlights handled, set the ISO to 400, and use that exposure. This will produce images that are, by conventional metering standards, more and more “underexposured” as the metered ISO rises.
5. If that ISO is over 3200, set the ISO setting to three stops under the metered ISO. Three stops of extra highlight protection should be enough for almost any scene, and pushing too much in postproduction can cause some color shifts.

For me, preserving highlights has often been high on my list and for at least 20 years I've tended to underexpose most of my shots by at least -0.7 and pushed them in post.

I generally have mostly followed these guidelines with my D850, though often I follow a slightly more simplified version and largely just focus on using ISO64 and ISO400. The only issue I've run into with this strategy is that reviewing the photos on the camera is a little more difficult.

With the Z9, I've tended not to follow these quite as much, mostly due to the EVF. I like the WYSIWYG, but if I am set for underexposing too much, then it gets hard to view in the EVF and I believe may also impact the AF. So I'm back to mostly letting the ISO float, but often with a -0.7 (or more, depending on the scene) exposure compensation. Stressing too much about incremental differences in the ISO300-ISO500 range hasn't seemed worth it to me, at least when dealing with moving/dynamic scenes.

Landscapes on the other hand .... I primarily use either ISO64 or ISO500 and I aim for "expose to the right" exposure settings. And somewhat ironically, I really love the Z9's 20FPS when doing landscapes because I almost always do a 5 shot bracket. I figure bits are free, and it's worth getting just the right exposure levels. With the 20FPS, the shots are so close together, that I can comfortably pick any of the 5 shots to work up in post, or I can pick several or all of them and do an HDR composite, even if the scene isn't entirely static.

 

[Strodav]

FWIW: From DxoMark.com

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

 

[bjanes]

I have posted a new blog entry titled Should Nikon Z9 shooters avoid ISO 400?

To get to the punchline - "Obviously the answer to the question "...." is NO, but if you find yourself needing to shoot at ISO levels between 300-400 slightly better quality results might be achieved if you try moving the ISO setting up to ISO 500 and see if there is a difference - the data indicates there will be." The Z9 uses a dual-gain approach to its signal processing which "dramatically" impacts the Dynamic Range and Signal to Noise levels at the point it "kicks in" (~ between ISO 400 and ISO 500).

Please have a read and provide POV in this thread.

Another opinion by Jim Kasson. DXO print dynamic range is very sensitive to read noise, so using ISO 500 will reduce the read noise and improve engineering dynamic range. However, using ISO 400 will allow more exposure and this will improve the signal to noise ratio (SNR) towards the right of the histogram. Jim's graphic demonstrates this well. ISO 400 will give better SNR until about 6 stops below clipping. For practical photography, the difference is not very significant. If the deep shadows are important as with forensic photography, use ISO 500.

Jim is a very authoritative source and interested Z9 users are advised to check his other posts on the Z9. Simply search for Nikon Z9 in the search bar of the main page.

Cheers,
Bill

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

 

[bjanes]

A few years ago I ran across an interesting (to me) article on the dual gain on the D850 and ISO invariance. After reading it and reviewing all of the samples, I came away with the following conclusion (which might be the conclusion from the article, I don't remember, I just have these conclusions in my notes file ... ). While most probably know this, just in case - the D850's extra gain starts at ISO400 whereas the Z9 is at ISO500, so adjust the directions accordingly.

1. If that ISO is 64 or as much as one stop over, set the ISO to 64 and make the exposure. If that ISO is 400 or as much as three stops over, set the ISO to 400 and make the exposure. As you crank the ISO incrementally up from 64 or 400, you’ll get less highlight protection, and it won’t gain you anything significant in shadow noise.
2. If that ISO is between 160 and 250, decide if you need the last bit of shadow signal-to-noise ratio (SNR). If you don’t, set the ISO to 160, take the shot, and get some extra highlight protection.
3. If that ISO is 320, decide if you could do with bit less highlight protection. If no, set the ISO to 320. If yes, set it to 400 and enjoy lower super-deep-shadow SNR.
4. If that ISO is between 400 and 2500, and you think you’ve got the highlights handled, set the ISO to 400, and use that exposure. This will produce images that are, by conventional metering standards, more and more “underexposured” as the metered ISO rises.
5. If that ISO is over 3200, set the ISO setting to three stops under the metered ISO. Three stops of extra highlight protection should be enough for almost any scene, and pushing too much in postproduction can cause some color shifts.

The source of these recommendations is Jim Kasson. The same strategy can be used with the Z9, noting that the dual gain switch is 500 for the Z9 rather than 400 for the D850. At low ISO the photographic DR of the Z9 is slightly worse than that of the D850, but the difference is not photographically significant.

See here for details.

Bill

 

[ML Wolff]

Is the extra gain for a Z7ii the same as for the Z9? If not, can anyone comment on the 'correct' ISO?

 

[bjanes]

Is the extra gain for a Z7ii the same as for the Z9? If not, can anyone comment on the 'correct' ISO?

With the Z7II, the switch to high gain is at ISO 400 as shown by this Bill Claff chart.

The previous comments cited for the Z9 apply with adjustments for thr Z7II.

Bill

 

[ML Wolff]

thanks

 

[MikePapple]

I believe that the Z9 uses a Sony sensor...

The Sony A7R4 has low noise at ISO 100 and 640, and the A1 at 100 and 500.

I just came back from a rafting trip from Whitehorse, Yukon to Northern BC to Alaskan panhandle (Alsek river).

On the raft I used A7R4, 24mm GM, F/11, ISO 100 or 640 and varied the shutter speed to expose-to-the-right. Lossless compressed RAW. Lots of glacier pano's. I got splashed on a lot. I took some video in the rapids. On one occasion the wave 'broke' in my face, and on another it 'broke' over my head. This worked well until my lens became 110% saturated with moisture (lots of spit/rain/fog).

I switched to A1 with 100-400 GM. Mostly 100mm, wide open F/5.6, ISO 100 or 500. I could zoom in to 400mm to get glacier close-ups. This is NOT my BIF settings (200-600, manual with auto ISO).

By-the-way, I also used an iPhone 12 Pro Max. Even in airplane mode, it would save GPS coords with each pic. I took pics at each campsite.

Not too many critters on this trip. Grizzly mama and cub, moose and wolf tracks. Some grizzlies at a distance. A few eagles, and a young male moose in the fog. I was expecting mountain goats.

 

[O]

What Dave says, just above my post is a valid observation for ISO invariance. Plus following are my observations based on Sensor Analysis Primer on "Photons to Photos" site.

• Important thing to remember is that photographic dynamic range is mainly concerning the recovery of shadows. If there are highlights in raw file close enough to bit 14, they will be clipped by increasing ISO.
  • by definition mentioned in its Engineering and Photographic Dynamic Range section PDR charts are using signal to noise ratio (SNR) 20 instead of SNR 1 as reference point - effectively cutting off low 4 bits of 14 bit raw and considering them black/indistinguishable from random pixel noise
  • based on the Photon Transfer Chart for Z9 I would guess that Z9 is not using blacked pixels to remove random noise in NEF files (as it does not converge to 0 but to 1 - compare with D850) and assumes that it is low enough to have no significant impact on the photographic dynamic range for scenes that are bright.
  • Increased analogue gain at ISO 400 assumes that there are no highlights available at that point that will be clipped by amplification, if there are any in the scene, the picture will be worse than the one on base ISO IMO as there will be less noise in shadows but at the expense of clipped highlights.
• I would recommend to use that ISO 400 analogue gain increase if scene is flat enough to fit into 9-10 bits of dynamic range because it will improve shadows retrieval.
• I would recommend to underexpose and use base ISO for scenes where highlights are high and important, because a viewer will notice the burned highlights more than the underexposed shadows. But the picture after postprocess will be more noisy if shadows recovery is attempted in tools like LR or PS.

Amazing, may i ask is this the same with the D850.
And i do notice that with my Z9 while excellent i need to watch the highlight a little especially at higher iso.
have i got this wright or am i confused.

I usually use 3200- 4000 ss floated iso to 12800 -07ev F7.1 for fast moving subjects in good light.

 

[O]

Dual gain architectures do give that jump in dynamic range but it's not something I worry about in actual shooting. It's a discussion closely related to all the angst that Nikon shooters had when the D5 came out with its less than stellar DR performance at low ISO but above its dual gain bump better performance at high ISO settings. The thing is that gain bump occurs in most cameras at a fairly low ISO setting at that also means at a fairly high point on the DR curve.

So yeah, if you shoot the Z9 at ISO 500 you'll get around 9.5 stops of dynamic range (give or take based on the noise floor SNR definition) and if you shoot just below the gain bump at ISO 400 you'll get roughly 8.7 stops of DR (again depending on the noise floor definition). But either is more than sufficient for a high quality print or display image assuming it was well shot to begin with and won't need something like several stops of shadow pull. in my experience high quality prints might have up to five stops of DR depending on the paper and the process and most monitors when viewed in anything but a darkened room might have a stop or two more DR but neither typically exceed 8 stops of DR unless there are ideal viewing conditions. That's basically why we can get away shooting at sky high ISO these days where shooting a camera like a Z9 at ISO 6400 and its associated sub 6 stops of DR is a walk in the park and with good processing the camera can be shot upwards of ISO 10,000 and 5 or fewer stops of DR.

When the D5 came out a lot of folks had trouble with the DR curve that below ISO 1000 or so hovers in the 8 to 9 stops of DR region but in practice the D5 delivers great images when shot at low or high ISO and of course its big advantage happens at high ISO where it outpaces many if not most cameras as the ISO climbs. Sure it would be great to see the low end of its curve up in the 10 to 12 stop region but for well exposed photos in decent light we can't really see the difference in prints or even most monitor use in anything but ideal viewing environments.

To me the gain bump is significant in another way as when you get above it you typically move into the ISO Invariant region of the ISO/DR curve and that has good implications to shooting and processing decisions. IOW, cameras aren't ISO invariant in the region around that bump where a small shift of ISO in post is very different than the same small shift of ISO in-camera but get above that gain bump and most modern sensors (really overall camera architectures as the gain bump is in the camera's signal processing not strictly in the sensor) become ISO invariant. That point is useful to know for folks like astro photographers that might choose to pull exposure up in post or folks who shoot with a bit of negative exposure compensation for insurance against clipped highlights and find themselves adjusting exposure in post.

So yeah, if you're shooting a tricky image where you know you'll need all the dynamic range you can get such as tough lighting with deep shadow detail that will need to be pulled and bright highlights that also need to be protected (I'm thinking midday outdoor wedding shooters struggling with hard shadows and bright highlights and not filling with flash or similar scenarios) then maybe shooting above the bump makes more sense than shooting just below the gain bump but shooting even lower (e.g. base ISO up to ISO 200) in the Z9 would make even more sense if you'll need all the DR you can get.

Just my 2 cents...

-Dave

Interesting........ so if i want maximum dynamic range out of my Z9 or D850 i should be at and below 200 iso and above 500 iso or have i got this wrong.

 

[Strodav]

Interesting........ so if i want maximum dynamic range out of my Z9 or D850 i should be at and below 200 iso and above 500 iso or have i got this wrong.

For the D850, the higher the ISO the lower the dynamic range. The curve is monotonically decreasing. I use ISO 64 for portraits (studio lighting) and landscapes.

For the Z9, as shown in the curves above, there is a bit of an anomaly between ISO 200 and ISO 800. From ISO 64 to ISO 200, the higher the ISO the lower the dynamic range. At ISO 200, the DR is 12.62 Ev. At ISO 400, the DR is 11.78 Ev, or about a stop difference. At ISO 800, the DR is 12.62. The curve is monotonically decreasing from there where the Z9 and D850 curves basically lay on top of each other.

The question becomes, will you nottice a difference between 11.78 Ev and 12.62Ev? It is a whole stop difference. In most situations, I wouldn't worry about it, but in an HDR situation, I might skip over ISO 400 and go right to ISO 800 with my Z9.

 

[O]

For the D850, the higher the ISO the lower the dynamic range. The curve is monotonically decreasing. I use ISO 64 for portraits (studio lighting) and landscapes.

For the Z9, as shown in the curves above, there is a bit of an anomaly between ISO 200 and ISO 800. From ISO 64 to ISO 200, the higher the ISO the lower the dynamic range. At ISO 200, the DR is 12.62 Ev. At ISO 400, the DR is 11.78 Ev, or about a stop difference. At ISO 800, the DR is 12.62. The curve is monotonically decreasing from there where the Z9 and D850 curves basically lay on top of each other.

The question becomes, will you nottice a difference between 11.78 Ev and 12.62Ev? It is a whole stop difference. In most situations, I wouldn't worry about it, but in an HDR situation, I might skip over ISO 400 and go right to ISO 800 with my Z9.

excellent, thank you
so setting the iso to a minimum of 800 may be handy with the iso floating.

 

[agrumpyoldsod]

Is the extra gain for a Z7ii the same as for the Z9? If not, can anyone comment on the 'correct' ISO?

No it is as I reported at the top of this thread -- the Z9 was "tweaked" to be better for low light and action. Just follow the link to my blog.

 

[Tibor Hrnko]

Amazing, may i ask is this the same with the D850.
And i do notice that with my Z9 while excellent i need to watch the highlight a little especially at higher iso.
have i got this wright or am i confused.

I usually use 3200- 4000 ss floated iso to 12800 -07ev F7.1 for fast moving subjects in good light.

Generally speaking yes, all digital cameras have decreased dynamic range with higher ISO.
One can visualize a digital camera pixel as a small bucket that is filled with tiny droplets of light. If the ISO is increased those drops are bigger and the bucket is filled quicker. Advantage of the setting is that camera is more sensitive. Disadvantage is that it cannot record the signal with the same precision.

For details one has to look at the graphs similar to those mentioned by poster and compare the differences between the implementation of specific camera models.

 

[Griffym]

Surely if ISO400 is appropriate for the creative vision you have in mind, you should use it.
I have never worried about ISO, I take much more notice of aperture and shutter speed. Should I start looking at my collection and seeing how many images I have taken at ISO 400?
If you put up 2 images one taken at ISO 400 and the same image taken at ISO 320, what percentage of viewers could tell which one was ISO 400?
I am very curious about the possible areas of photography where ISO would become a significant issue in trying to reach the desired output. Fascinating!
Is Dave right?

 

[Strodav]

Surely if ISO400 is appropriate for the creative vision you have in mind, you should use it.
I have never worried about ISO, I take much more notice of aperture and shutter speed. Should I start looking at my collection and seeing how many images I have taken at ISO 400?
If you put up 2 images one taken at ISO 400 and the same image taken at ISO 320, what percentage of viewers could tell which one was ISO 400?
I am very curious about the possible areas of photography where ISO would become a significant issue in trying to reach the desired output. Fascinating!
Is Dave right?

For birding and other dynamic situations, I usually shoot raw, manual with auto ISO, so the camera selects the ISO within the range I set. I won't worry if it selects ISO 400 on my Z9. In a static situation, like birds in water, where I have time and am selecting shutter speed, f stop and ISO, I will avoid ISO 400 on the Z9, choosing instead ISO 800. A stop (about 1 Ev) is effectively twice the amount of light.

 

[bjanes]

Dual gain architectures do give that jump in dynamic range but it's not something I worry about in actual shooting. It's a discussion closely related to all the angst that Nikon shooters had when the D5 came out with its less than stellar DR performance at low ISO but above its dual gain bump better performance at high ISO settings. The thing is that gain bump occurs in most cameras at a fairly low ISO setting at that also means at a fairly high point on the DR curve.

Just my 2 cents...

-Dave

Dave's comments are considerably worth more than two cents, but there is an error: the D5 does not have dual gain as shown by the DR plot on Bill Claff's web site. With dual gain there is a discontinuity in the plot at the dual gain switch. With the D5, the plot becomes linear at ISO 2546 and this is the point at which the camera becomes ISO invariant.

The astrophysicist Roger Clark has a good write up on using the Canon 5D MII for astrophotography where it is important to maximize sensor performance. The plot becomes linear at ISO 1600, which is where the sensor becomes ISO invariant. In this type of work, there is no need to increase the ISO beyond 1600. As Roger says, ISO 1600 gets it all.

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

Here are photographic DR plots from Bill Claff's site for some representative cameras with my annotations added. The important points on the plot are when the sensor becomes ISO invariant as shown in the linear part of the graph, when analog amplification ceases (shown by open circles), and where analog amplification is maximal.

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

I hope this helps. For more explanation of these charts, please refer to Bill's site. The plot for the D5 (not shown) is very similar to that of the D6.

Bill

 

[bjanes]

This Claff Photographic Dynamic range Improvement chart shows more clearly the dual gain switch points for the D850 and Z9. The D5 is not dual gain but becomes more or less ISO invariant at ISO 2546. The Z9 dual gain sensor is nearly perfectly ISO invariant from ISO 63 to 400, and again from 400 to max. The D850 is pretty much ISO invariant from ISO 400 and beyond but shows a staggard pattern from base to ISO 400.

If you are using manual exposure with the Z9 you can meter for an ISO 400 ETTR exposure, use this f/stop and shutter speed, and then set the camera ISO to 64 and brighten in post. In both cases the esposure in lux/seconds is the same. This gives 3 extra stops of highlight headroom. The same strategy applies for ISOs above 400, but it is not advisable to brighten more than 3 stops in post. See Jim Kasson for details on how to do this with the D850. For the Z9, the strategy is simpler since the Z9 is ISO invariant between ISO 64 and 400 and Jim's intermediate steps for ISOs between 64 and 400 are not necessary.

Bill