Am I the only one with Z9 high ISO issue?

[myalhi]

Hi there,

I didn't notice that many people complaining about a terrible ISO performance of Z9 espacially comparing to my previous D6. I mean even in daylight with a 500 ISO for ex. I get noise clearly visible in the shadow areas. I don't even mention above ISO 2000 in low light. I'm shooting RAW at all times but feeling like I went 10 years back technology in terms of ISO performance. And that is really upsetting me with such a pro body in my hands. Any ideas/solutions highly appreciated.

Please check attached image at ISO 2500 during daylight (just about 1 hour before it gets dark). I wasn't noticing that much noise with my D6 at even 6400 ISO.







--
Mehmet
@meyaajans
@i.c.h.i.g.o_i.c.h.i.e
@f2.8.125.bw

 

[WWeber]

At what level do you compare the nois of D6 and Z9? At 100%? If so, you have to keep in mind that the magification of the Image of the Z9 is ca. 40% larger (linear) than the D6.

For a objective comparison you have to downsample the Z9 image from 45 to 20.5MP.

Under that conditions the noise level of the Z9 is as good as the noise level of my D5.

 

[SCoombs]

At what level do you compare the nois of D6 and Z9? At 100%? If so, you have to keep in mind that the magification of the Image of the Z9 is ca. 40% larger (linear) than the D6.

For a objective comparison you have to downsample the Z9 image from 45 to 20.5MP.

Under that conditions the noise level of the Z9 is as good as the noise level of my D5.

This doesn't make much sense to me. If one puts a Z9 in DX node so that it's using approximately 20MP of the sensor, if should be possible to compare apples to apples with a 20MP camera body. My experience with this (on the Z8) is that the noise seems a lot worse, especially in better lighting conditions where you don't expect to see much noise.

 

[redeeman]

At what level do you compare the nois of D6 and Z9? At 100%? If so, you have to keep in mind that the magification of the Image of the Z9 is ca. 40% larger (linear) than the D6.

For a objective comparison you have to downsample the Z9 image from 45 to 20.5MP.

Under that conditions the noise level of the Z9 is as good as the noise level of my D5.

This doesn't make much sense to me. If one puts a Z9 in DX node so that it's using approximately 20MP of the sensor, if should be possible to compare apples to apples with a 20MP camera body. My experience with this (on the Z8) is that the noise seems a lot worse, especially in better lighting conditions where you don't expect to see much noise.

no, you need to use the full sensor size to compare. So take a full 45mp photo, downscale to resolution of camera you compare to

 

[Mako2011]

At what level do you compare the nois of D6 and Z9? At 100%? If so, you have to keep in mind that the magification of the Image of the Z9 is ca. 40% larger (linear) than the D6.

For a objective comparison you have to downsample the Z9 image from 45 to 20.5MP.

Under that conditions the noise level of the Z9 is as good as the noise level of my D5.

This doesn't make much sense to me. If one puts a Z9 in DX node so that it's using approximately 20MP of the sensor, if should be possible to compare apples to apples with a 20MP camera body.

That's not really the case as you are not comparing equal resolutions. Cropping is not down-sampling

 

[ghostfox_1]

At what level do you compare the nois of D6 and Z9? At 100%? If so, you have to keep in mind that the magification of the Image of the Z9 is ca. 40% larger (linear) than the D6.

For a objective comparison you have to downsample the Z9 image from 45 to 20.5MP.

Under that conditions the noise level of the Z9 is as good as the noise level of my D5.

This doesn't make much sense to me. If one puts a Z9 in DX node so that it's using approximately 20MP of the sensor, if should be possible to compare apples to apples with a 20MP camera body. My experience with this (on the Z8) is that the noise seems a lot worse, especially in better lighting conditions where you don't expect to see much noise.

To elaborate on the other replies, you're still comparing smaller pixels to larger ones. Downsampling effectively bins the pixels making it an actual 1:1 comparison (or closer to, anyway).

If I had a dollar for every time this came up, I'd be rich.

 

[maverick256]

At what level do you compare the nois of D6 and Z9? At 100%? If so, you have to keep in mind that the magification of the Image of the Z9 is ca. 40% larger (linear) than the D6.

For a objective comparison you have to downsample the Z9 image from 45 to 20.5MP.

Under that conditions the noise level of the Z9 is as good as the noise level of my D5.

This doesn't make much sense to me. If one puts a Z9 in DX node so that it's using approximately 20MP of the sensor, if should be possible to compare apples to apples with a 20MP camera body. My experience with this (on the Z8) is that the noise seems a lot worse, especially in better lighting conditions where you don't expect to see much noise.

D5/6 have 20MP FX sensors, compared to 47MP FX sensors on the Z8/9. Each pixel on the Z8/9 is less than half the size compared to D5/6, and therefore only receive less than half the light. So yeah, if you compare a 20MP DX crop on the Z8/9 to the 20MP FX on D5/6, you would expect the Z8/9 to have one (and a bit) more stop of noise.

That's just a consequence of the physics of light, regardless of sensor technology. I don't think this should be surprise to anyone...

 

[SCoombs]

At what level do you compare the nois of D6 and Z9? At 100%? If so, you have to keep in mind that the magification of the Image of the Z9 is ca. 40% larger (linear) than the D6.

For a objective comparison you have to downsample the Z9 image from 45 to 20.5MP.

Under that conditions the noise level of the Z9 is as good as the noise level of my D5.

This doesn't make much sense to me. If one puts a Z9 in DX node so that it's using approximately 20MP of the sensor, if should be possible to compare apples to apples with a 20MP camera body.

That's not really the case as you are not comparing equal resolutions. Cropping is not down-sampling

I think it's probably quite a pedantic concern to worry about a tiny difference in resolution, but fine, crop to the exact same number of pixels. How is this not an equal comparison?

 

[ghostfox_1]

At what level do you compare the nois of D6 and Z9? At 100%? If so, you have to keep in mind that the magification of the Image of the Z9 is ca. 40% larger (linear) than the D6.

For a objective comparison you have to downsample the Z9 image from 45 to 20.5MP.

Under that conditions the noise level of the Z9 is as good as the noise level of my D5.

This doesn't make much sense to me. If one puts a Z9 in DX node so that it's using approximately 20MP of the sensor, if should be possible to compare apples to apples with a 20MP camera body.

That's not really the case as you are not comparing equal resolutions. Cropping is not down-sampling

I think it's probably quite a pedantic concern to worry about a tiny difference in resolution, but fine, crop to the exact same number of pixels. How is this not an equal comparison?

Because you're comparing a full sensor to less than a full sensors worth of data.

It's like being shocked that crop sensors are noisier than ff, and mft more so.

It's literally the same thing. I don't understand where the confusion is on your part.

 

[SCoombs]

At what level do you compare the nois of D6 and Z9? At 100%? If so, you have to keep in mind that the magification of the Image of the Z9 is ca. 40% larger (linear) than the D6.

For a objective comparison you have to downsample the Z9 image from 45 to 20.5MP.

Under that conditions the noise level of the Z9 is as good as the noise level of my D5.

This doesn't make much sense to me. If one puts a Z9 in DX node so that it's using approximately 20MP of the sensor, if should be possible to compare apples to apples with a 20MP camera body. My experience with this (on the Z8) is that the noise seems a lot worse, especially in better lighting conditions where you don't expect to see much noise.

D5/6 have 20MP FX sensors, compared to 47MP FX sensors on the Z8/9. Each pixel on the Z8/9 is less than half the size compared to D5/6, and therefore only receive less than half the light. So yeah, if you compare a 20MP DX crop on the Z8/9 to the 20MP FX on D5/6, you would expect the Z8/9 to have one (and a bit) more stop of noise.

That's just a consequence of the physics of light, regardless of sensor technology. I don't think this should be surprise to anyone...

If this is the case then I don't understand what the advantage of a larger megapixel sensor camera would be. They are supposed to provide more resolution, more detail, more ability to crop without issues, etc. I have never seen any of this disputed. Yet if the pixels are smaller in this way, meaning you've also got greater distance between the pixels, would you not have essentially *leas* detail in a full frame 45 MP image?

 

[ghostfox_1]

At what level do you compare the nois of D6 and Z9? At 100%? If so, you have to keep in mind that the magification of the Image of the Z9 is ca. 40% larger (linear) than the D6.

For a objective comparison you have to downsample the Z9 image from 45 to 20.5MP.

Under that conditions the noise level of the Z9 is as good as the noise level of my D5.

This doesn't make much sense to me. If one puts a Z9 in DX node so that it's using approximately 20MP of the sensor, if should be possible to compare apples to apples with a 20MP camera body. My experience with this (on the Z8) is that the noise seems a lot worse, especially in better lighting conditions where you don't expect to see much noise.

D5/6 have 20MP FX sensors, compared to 47MP FX sensors on the Z8/9. Each pixel on the Z8/9 is less than half the size compared to D5/6, and therefore only receive less than half the light. So yeah, if you compare a 20MP DX crop on the Z8/9 to the 20MP FX on D5/6, you would expect the Z8/9 to have one (and a bit) more stop of noise.

That's just a consequence of the physics of light, regardless of sensor technology. I don't think this should be surprise to anyone...

If this is the case then I don't understand what the advantage of a larger megapixel sensor camera would be. They are supposed to provide more resolution, more detail, more ability to crop without issues, etc. I have never seen any of this disputed. Yet if the pixels are smaller in this way, meaning you've also got greater distance between the pixels, would you not have essentially *leas* detail in a full frame 45 MP image?

I think you have a lot of misunderstandings about how all of this works, which is unfortunate given how long you've apparently been shooting. The pixels are smaller, which allows for finer detail. How do you believe "there's more distance" between them? There's literally more pixels on the sensor, all still next to each other.

Maybe you should find some YouTube videos or similar on the subject, though I don't have any recommendations off hand...

 

[SCoombs]

At what level do you compare the nois of D6 and Z9? At 100%? If so, you have to keep in mind that the magification of the Image of the Z9 is ca. 40% larger (linear) than the D6.

For a objective comparison you have to downsample the Z9 image from 45 to 20.5MP.

Under that conditions the noise level of the Z9 is as good as the noise level of my D5.

This doesn't make much sense to me. If one puts a Z9 in DX node so that it's using approximately 20MP of the sensor, if should be possible to compare apples to apples with a 20MP camera body.

That's not really the case as you are not comparing equal resolutions. Cropping is not down-sampling

I think it's probably quite a pedantic concern to worry about a tiny difference in resolution, but fine, crop to the exact same number of pixels. How is this not an equal comparison?

Because you're comparing a full sensor to less than a full sensors worth of data.

It's like being shocked that crop sensors are noisier than ff, and mft more so.

It's literally the same thing. I don't understand where the confusion is on your part.

But you're comparing the SAME number of pixels. Who cares what the overall size of the sensor is? A sensor is not some magical unit that contains some magically comprehensive data set. It is, to speak simply, a wafer with photosites on it. The photosites do the job - not the overall sensor.

Moreover, you mention that crop sensors are noisier than full frame - an uncontroversial statement - but if what you're saying is true I don't see how this could be. You say the smaller pixels of the full frame sensor make for less light and more noise than the larger APS-C pixels, and you say that viewing a full frame photo at its full size will look noisier than viewing the APS-C photo at its full size. So how do we reason that the full frame has less noise?

Let me make this simple: if I take a photo on my Z8 and then take the same photo on my old D500 and then view them both filling all of my 1080P monitor, the D500 image looks less noisy.

 

[ghostfox_1]

At what level do you compare the nois of D6 and Z9? At 100%? If so, you have to keep in mind that the magification of the Image of the Z9 is ca. 40% larger (linear) than the D6.

For a objective comparison you have to downsample the Z9 image from 45 to 20.5MP.

Under that conditions the noise level of the Z9 is as good as the noise level of my D5.

This doesn't make much sense to me. If one puts a Z9 in DX node so that it's using approximately 20MP of the sensor, if should be possible to compare apples to apples with a 20MP camera body.

That's not really the case as you are not comparing equal resolutions. Cropping is not down-sampling

I think it's probably quite a pedantic concern to worry about a tiny difference in resolution, but fine, crop to the exact same number of pixels. How is this not an equal comparison?

Because you're comparing a full sensor to less than a full sensors worth of data.

It's like being shocked that crop sensors are noisier than ff, and mft more so.

It's literally the same thing. I don't understand where the confusion is on your part.

But you're comparing the SAME number of pixels. Who cares what the overall size of the sensor is? A sensor is not some magical unit that contains some magically comprehensive data set. It is, to speak simply, a wafer with photosites on it. The photosites do the job - not the overall sensor.

So do you think mft has the same noise as an ff sensor then? Just curious here.

Moreover, you mention that crop sensors are noisier than full frame - an uncontroversial statement - but if what you're saying is true I don't see how this could be. You say the smaller pixels of the full frame sensor make for less light and more noise than the larger APS-C pixels, and you say that viewing a full frame photo at its full size will look noisier than viewing the APS-C photo at its full size. So how do we reason that the full frame has less noise?

Because when you take it as a whole, the sensor is capturing more light and (on average) the photosites are still larger than crop, allowing for lower noise overall, though we can get into read noise, etc, and there is a lot of info out there on the subject.

Let me make this simple: if I take a photo on my Z8 and then take the same photo on my old D500 and then view them both filling all of my 1080P monitor, the D500 image looks less noisy.

Then you've done something wrong in the process.

 

[SCoombs]

At what level do you compare the nois of D6 and Z9? At 100%? If so, you have to keep in mind that the magification of the Image of the Z9 is ca. 40% larger (linear) than the D6.

For a objective comparison you have to downsample the Z9 image from 45 to 20.5MP.

Under that conditions the noise level of the Z9 is as good as the noise level of my D5.

This doesn't make much sense to me. If one puts a Z9 in DX node so that it's using approximately 20MP of the sensor, if should be possible to compare apples to apples with a 20MP camera body. My experience with this (on the Z8) is that the noise seems a lot worse, especially in better lighting conditions where you don't expect to see much noise.

D5/6 have 20MP FX sensors, compared to 47MP FX sensors on the Z8/9. Each pixel on the Z8/9 is less than half the size compared to D5/6, and therefore only receive less than half the light. So yeah, if you compare a 20MP DX crop on the Z8/9 to the 20MP FX on D5/6, you would expect the Z8/9 to have one (and a bit) more stop of noise.

That's just a consequence of the physics of light, regardless of sensor technology. I don't think this should be surprise to anyone...

If this is the case then I don't understand what the advantage of a larger megapixel sensor camera would be. They are supposed to provide more resolution, more detail, more ability to crop without issues, etc. I have never seen any of this disputed. Yet if the pixels are smaller in this way, meaning you've also got greater distance between the pixels, would you not have essentially *leas* detail in a full frame 45 MP image?

I think you have a lot of misunderstandings about how all of this works, which is unfortunate given how long you've apparently been shooting. The pixels are smaller, which allows for finer detail. How do you believe "there's more distance" between them? There's literally more pixels on the sensor, all still next to each other.

Maybe you should find some YouTube videos or similar on the subject, though I don't have any recommendations off hand...

The full frame sensor is, area wise, much larger than the APS-C sensor. APS-C I'd about 370 square mm. FF is about 864 square mm. With each, that works out to about .05 million pixels per square mm assuming 20MP vs 45MP. But if the 45MP pixels are half the size, then that means there is about twice as much empty space as with the APS-C sensor.

 

[mosswings]

At what level do you compare the nois of D6 and Z9? At 100%? If so, you have to keep in mind that the magification of the Image of the Z9 is ca. 40% larger (linear) than the D6.

For a objective comparison you have to downsample the Z9 image from 45 to 20.5MP.

Under that conditions the noise level of the Z9 is as good as the noise level of my D5.

This doesn't make much sense to me. If one puts a Z9 in DX node so that it's using approximately 20MP of the sensor, if should be possible to compare apples to apples with a 20MP camera body. My experience with this (on the Z8) is that the noise seems a lot worse, especially in better lighting conditions where you don't expect to see much noise.

D5/6 have 20MP FX sensors, compared to 47MP FX sensors on the Z8/9. Each pixel on the Z8/9 is less than half the size compared to D5/6, and therefore only receive less than half the light. So yeah, if you compare a 20MP DX crop on the Z8/9 to the 20MP FX on D5/6, you would expect the Z8/9 to have one (and a bit) more stop of noise.

That's just a consequence of the physics of light, regardless of sensor technology. I don't think this should be surprise to anyone...

If this is the case then I don't understand what the advantage of a larger megapixel sensor camera would be. They are supposed to provide more resolution, more detail, more ability to crop without issues, etc. I have never seen any of this disputed. Yet if the pixels are smaller in this way, meaning you've also got greater distance between the pixels, would you not have essentially *leas* detail in a full frame 45 MP image?

The advantage of a 45MPX FF sensor over a 20MPX FF sensor is resolution. The D5/6 have less resolution than a Z9. They do NOT necessarily have less noise if both are designed to deliver the same frame-level noise levels, which means engineering the sensels to have noise levels that go down as the square root of sensel area. This manufacturers have largely been able to do, with allowances made for features deemed necessary for marketing purposes, such as global shutter. If you look at images with the same magnification and area, the images should appear equally noisy. Remember that typically you are NOT looking at the native resolution of the image file, but that of the viewing medium. Unless you’re pixel peeping, and the assumptions for frame-level performance no longer hold.

 

[ghostfox_1]

At what level do you compare the nois of D6 and Z9? At 100%? If so, you have to keep in mind that the magification of the Image of the Z9 is ca. 40% larger (linear) than the D6.

For a objective comparison you have to downsample the Z9 image from 45 to 20.5MP.

Under that conditions the noise level of the Z9 is as good as the noise level of my D5.

This doesn't make much sense to me. If one puts a Z9 in DX node so that it's using approximately 20MP of the sensor, if should be possible to compare apples to apples with a 20MP camera body. My experience with this (on the Z8) is that the noise seems a lot worse, especially in better lighting conditions where you don't expect to see much noise.

D5/6 have 20MP FX sensors, compared to 47MP FX sensors on the Z8/9. Each pixel on the Z8/9 is less than half the size compared to D5/6, and therefore only receive less than half the light. So yeah, if you compare a 20MP DX crop on the Z8/9 to the 20MP FX on D5/6, you would expect the Z8/9 to have one (and a bit) more stop of noise.

That's just a consequence of the physics of light, regardless of sensor technology. I don't think this should be surprise to anyone...

If this is the case then I don't understand what the advantage of a larger megapixel sensor camera would be. They are supposed to provide more resolution, more detail, more ability to crop without issues, etc. I have never seen any of this disputed. Yet if the pixels are smaller in this way, meaning you've also got greater distance between the pixels, would you not have essentially *leas* detail in a full frame 45 MP image?

I think you have a lot of misunderstandings about how all of this works, which is unfortunate given how long you've apparently been shooting. The pixels are smaller, which allows for finer detail. How do you believe "there's more distance" between them? There's literally more pixels on the sensor, all still next to each other.

Maybe you should find some YouTube videos or similar on the subject, though I don't have any recommendations off hand...

The full frame sensor is, area wise, much larger than the APS-C sensor. APS-C I'd about 370 square mm. FF is about 864 square mm. With each, that works out to about .05 million pixels per square mm assuming 20MP vs 45MP. But if the 45MP pixels are half the size, then that means there is about twice as much empty space as with the APS-C sensor.

That's... Not how that works.

 

[skyrunr]

"With great resolution comes great responsibility." Noise is the main reason why higher end D# cameras have lower resolution than the entry level consumer offerings.
Going from 45mp from 24mp takes some getting used to. You're essentially magnifying everything. You're going to notice more blemishes, noise, and lens imperfections. Your workflow and software will also make a big difference. Lightroom doesn't do a great job with NEF's. For various reasons, Adobe and Nikon don't work together to improve results.

If you want better SOOC results shoot JPG. JPG's from the 45mp Nikon bodies are stunning, and nicer than 24mp NEF's (at least compared to a D750 as I did.) You can create a simple preset to cleanup your images upon import. As an example, LR puts a +40 sharpening on NEF's by default.

When people mention downsizing to compare images, the data is being condensed, and that reduces the visible amount of noise. The same thing happens during print rendering.

Why do many want 45mp or 60mp? You can just see the text on the sticker on the bottom of the cane in your example image. This would not be visible from a lower resolution (24mp) camera. Which is why many of us are waiting for the Z6III.

--
SkyRunR
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
TIPS: Be kind, RT#M, use gear not signature, limit/shorten replies with quotes!
'The first casualty, when war comes, is truth' - Hiram Johnson (1866-1945)

 

[Mako2011]

But you're comparing the SAME number of pixels. Who cares what the overall size of the sensor is? A sensor is not some magical unit that contains some magically comprehensive data set. It is, to speak simply, a wafer with photosites on it. The photosites do the job - not the overall sensor.

When comparing apples to apples (regards noise) you really have to downsample one to match other

Let me make this simple: if I take a photo on my Z8 and then take the same photo on my old D500 and then view them both filling all of my 1080P monitor, the D500 image looks less noisy.

Because you are not making an equivalent comparison (subject distances are different if the FoV for the image is the same when you view both on your screen). Or one is being magnified more than the other when you set both to 1080p.

 

[SCoombs]

At what level do you compare the nois of D6 and Z9? At 100%? If so, you have to keep in mind that the magification of the Image of the Z9 is ca. 40% larger (linear) than the D6.

For a objective comparison you have to downsample the Z9 image from 45 to 20.5MP.

Under that conditions the noise level of the Z9 is as good as the noise level of my D5.

This doesn't make much sense to me. If one puts a Z9 in DX node so that it's using approximately 20MP of the sensor, if should be possible to compare apples to apples with a 20MP camera body.

That's not really the case as you are not comparing equal resolutions. Cropping is not down-sampling

I think it's probably quite a pedantic concern to worry about a tiny difference in resolution, but fine, crop to the exact same number of pixels. How is this not an equal comparison?

Because you're comparing a full sensor to less than a full sensors worth of data.

It's like being shocked that crop sensors are noisier than ff, and mft more so.

It's literally the same thing. I don't understand where the confusion is on your part.

But you're comparing the SAME number of pixels. Who cares what the overall size of the sensor is? A sensor is not some magical unit that contains some magically comprehensive data set. It is, to speak simply, a wafer with photosites on it. The photosites do the job - not the overall sensor.

So do you think mft has the same noise as an ff sensor then? Just curious here.

I accept the conventional wisdom is that larger sensors have less noise, which seems contrary to what's been suggested here. If the full frame sensor only has less noise if you downsample it, then having the FF sensor seems like a bit of a "lie" and the conventional wisdom wouldn't seem honest to me. It would be like saying that one 40 hour a week job pays more than another 20 hour a week job - but without telling you that in the 40 hour job half the pay is going to be taken away for taxes while you get the full 20 hour paycheck.

Moreover, you mention that crop sensors are noisier than full frame - an uncontroversial statement - but if what you're saying is true I don't see how this could be. You say the smaller pixels of the full frame sensor make for less light and more noise than the larger APS-C pixels, and you say that viewing a full frame photo at its full size will look noisier than viewing the APS-C photo at its full size. So how do we reason that the full frame has less noise?

Because when you take it as a whole, the sensor is capturing more light and (on average) the photosites are still larger than crop, allowing for lower noise overall, though we can get into read noise, etc, and there is a lot of info out there on the subject.

Let me make this simple: if I take a photo on my Z8 and then take the same photo on my old D500 and then view them both filling all of my 1080P monitor, the D500 image looks less noisy.

Then you've done something wrong in the process.

Straight out of camera. Import to Lightroom, no edits, export.

 

[Horshack]

At what level do you compare the nois of D6 and Z9? At 100%? If so, you have to keep in mind that the magification of the Image of the Z9 is ca. 40% larger (linear) than the D6.

For a objective comparison you have to downsample the Z9 image from 45 to 20.5MP.

Under that conditions the noise level of the Z9 is as good as the noise level of my D5.

Here's a comparison of the D800 vs D3s at High ISO, first at their native sizes (1:1), then with the D800 downsampled to the D3s's resolution. This is from 12 years ago, to give you an idea of how long this has been discussed.


D3s vs D800, native resolution, 100% crop


D3s vs D800, D800 downsampled to 12MP, 100% crop