D5 Read Noise

Started Mar 27, 2016 | Discussions
bclaff Veteran Member • Posts: 8,295
D5 Read Noise
12

I have Nikon D5 Read Noise in DNs as well as Input-referred Read Noise to complement my earlier Photographic Dynamic Range (PDR) measurements.

The pattern is somewhat unusual for a Nikon camera, particularly the intermediate ISO behavior.
There is no sharp obvious boundary showing dual conversion gain technology.
It looks like something new that I don't understand yet.

-- hide signature --

Bill ( Your trusted source for independent sensor data at http://www.PhotonsToPhotos.net )

Nikon D5
If you believe there are incorrect tags, please send us this post using our feedback form.
Marianne Oelund Veteran Member • Posts: 7,698
Re: D5 Read Noise
11

bclaff wrote:

The pattern is somewhat unusual for a Nikon camera, particularly the intermediate ISO behavior.
There is no sharp obvious boundary showing dual conversion gain technology.
It looks like something new that I don't understand yet.

I would just describe it as more welcome advancements in optimizing the analog chain.

The read noise plot is a practical and useful guide, as it clearly points out which ISO settings are disadvantageous (those with higher read noise than a higher setting, such as 125, 160, 250, 320, etc.).  I'll have to remember to stay at 3200 or 5000 rather than using 4000.

Although sensor data is interesting, it is only part of the story of this camera.  For users who need to rely on camera JPEG output, the advancements in the D5's NR (which is effective and less damaging to detail) are a huge improvement.  The images coming from my D5 at ISO 3200-25600 are far cleaner and better detailed than I've seen from any of my other dSLRs (that includes D3s and D4) - more so than one would expect from subtle sensor improvements.

Of course, in my business this can be a negative:  Skin imperfections are no longer so well concealed by noise.

-- hide signature --

Source credit: Prov 2:6
- Marianne

Alex Dawson
Alex Dawson Forum Member • Posts: 53
Re: D5 Read Noise

Marianne Oelund wrote:

The read noise plot is a practical and useful guide, as it clearly points out which ISO settings are disadvantageous (those with higher read noise than a higher setting, such as 125, 160, 250, 320, etc.). I'll have to remember to stay at 3200 or 5000 rather than using 4000.

Why not ISO 2500 & ISO 5000 instead of 3200? Is there something I don't understand here?

kind regards // Alex

-- hide signature --

Photographer Alex Dawson http://www.dawson-photo.com

 Alex Dawson's gear list:Alex Dawson's gear list
Nikon D850 Nikon AF-S Nikkor 16-35mm F4G ED VR Nikon AF-S Nikkor 14-24mm f/2.8G ED Nikon AF-S Nikkor 85mm f/1.4G Sigma 35mm F1.4 DG HSM Art +2 more
StanyBuyle
StanyBuyle Senior Member • Posts: 1,293
Re: D5 Read Noise
1

Marianne Oelund wrote:

bclaff wrote:

The pattern is somewhat unusual for a Nikon camera, particularly the intermediate ISO behavior.
There is no sharp obvious boundary showing dual conversion gain technology.
It looks like something new that I don't understand yet.

I would just describe it as more welcome advancements in optimizing the analog chain.

The read noise plot is a practical and useful guide, as it clearly points out which ISO settings are disadvantageous (those with higher read noise than a higher setting, such as 125, 160, 250, 320, etc.). I'll have to remember to stay at 3200 or 5000 rather than using 4000.

Although sensor data is interesting, it is only part of the story of this camera. For users who need to rely on camera JPEG output, the advancements in the D5's NR (which is effective and less damaging to detail) are a huge improvement. The images coming from my D5 at ISO 3200-25600 are far cleaner and better detailed than I've seen from any of my other dSLRs (that includes D3s and D4) - more so than one would expect from subtle sensor improvements.

  • Thanks for your findings. I am convinced about similar improvements with D500 vs D7200.
    http://www.dpreview.com/forums/post/57067552
  • How do you experience the compromise between less DR vs much better high iso results (detail and noise) compared to D4s?
  • Any idea who makes the D5/D500 sensor?

Of course, in my business this can be a negative: Skin imperfections are no longer so well concealed by noise.

-- hide signature --

Source credit: Prov 2:6
- Marianne

-- hide signature --

I like better one decent picture in a day than 10 bad ones in a second.
Registered on Dpreview since 2003.

 StanyBuyle's gear list:StanyBuyle's gear list
Nikon D1 Nikon D1X Nikon D1H Nikon D2H Nikon D2X +1 more
Marianne Oelund Veteran Member • Posts: 7,698
Re: D5 Read Noise
5

Alex Dawson wrote:

Why not ISO 2500 & ISO 5000 instead of 3200? Is there something I don't understand here?

ISO 2500 is even better than 3200, but I often do not have the option of going that low.

Overall, settings best avoided are 160, 320, 1000, 2000 and 4000.  Settings best to use would be 100, 200, 400, 640, 1250, 2500, 5000.  Again, this is just for sensor DR considerations; including the camera's JPEG engine characteristics would make things a little more convoluted.

-- hide signature --

Source credit: Prov 2:6
- Marianne

Dibyendu Majumdar Regular Member • Posts: 494
Re: D5 Read Noise

Thank you for your analysis. I was wondering if any comparison is possible with the 1DX MII from the samples posted here (apparently raw samples are available privately):

http://www.canonrumors.com/forum/index.php?topic=29425.0

Regards

JimPearce
JimPearce Veteran Member • Posts: 9,188
superficially Bill...

There looks to be similarities between the D5 and D3s, in terms of both the read noise and PDR curves. Something new from Nikon, or something old?

-- hide signature --

Jim

 JimPearce's gear list:JimPearce's gear list
Nikon D7100 Nikon D500
OP bclaff Veteran Member • Posts: 8,295
Re: D5 Read Noise

Dibyendu Majumdar wrote:

Thank you for your analysis. I was wondering if any comparison is possible with the 1DX MII from the samples posted here (apparently raw samples are available privately):

http://www.canonrumors.com/forum/index.php?topic=29425.0

I need raw files taken in a particular way.
Perhaps someone with a 1DX MkII will be in touch sometime soon.

-- hide signature --

Bill ( Your trusted source for independent sensor data at http://www.PhotonsToPhotos.net )

JimPearce
JimPearce Veteran Member • Posts: 9,188
a question Marianne...

For those of us who shoot RAW, would we likely need to use Nikon software to realize these improvements?

-- hide signature --

Jim

 JimPearce's gear list:JimPearce's gear list
Nikon D7100 Nikon D500
Marianne Oelund Veteran Member • Posts: 7,698
Re: a question Marianne...

JimPearce wrote:

For those of us who shoot RAW, would we likely need to use Nikon software to realize these improvements?

I would say no.  There are so many converters and NR tools available today, that even though the D5's internal processing has improved, it still can't equal the results obtainable in post processing.

Nikon's software might make it easier to simulate what the camera does, but don't expect it to be identical to the D5 processing (Canon does, or did, offer tools that faithfully replicate what the camera does, but Nikon's approach is different).

-- hide signature --

Source credit: Prov 2:6
- Marianne

just Tony
just Tony Senior Member • Posts: 2,519
Looks like Grandpa

JimPearce wrote:

There looks to be similarities between the D5 and D3s, in terms of both the read noise and PDR curves. Something new from Nikon, or something old?

Rather similar features in the curves:

http://www.photonstophotos.net/Charts/PDR.htm#Nikon%20D3,Nikon%20D3s,Nikon%20D5

OP bclaff Veteran Member • Posts: 8,295
Re: Looks like Grandpa

just Tony wrote:

JimPearce wrote:

There looks to be similarities between the D5 and D3s, in terms of both the read noise and PDR curves. Something new from Nikon, or something old?

Rather similar features in the curves:

http://www.photonstophotos.net/Charts/PDR.htm#Nikon%20D3,Nikon%20D3s,Nikon%20D5

However, I've never seen a camera with intermediate ISO behavior that changed across ISO settings.
Notice how the D5 shifts from a pattern that starts with the whole ISO setting to one that is centered on the whole ISO setting at about ISO 640.
Starts with ISO 100, 200, 400; centered on ISO 800, 1600, 3200, 6400
Easiest to see on the Input-referred Read Noise chart:

-- hide signature --

Bill ( Your trusted source for independent sensor data at http://www.PhotonsToPhotos.net )

Marianne Oelund Veteran Member • Posts: 7,698
Re: Looks like Grandpa
2

bclaff wrote:

However, I've never seen a camera with intermediate ISO behavior that changed across ISO settings.
Notice how the D5 shifts from a pattern that starts with the whole ISO setting to one that is centered on the whole ISO setting at about ISO 640.
Starts with ISO 100, 200, 400; centered on ISO 800, 1600, 3200, 6400

There is no engineering reason to stay with full-stop intervals, in fact we may see more shorter intervals in future cameras, as optimization becomes more complex and well-tuned.

What this has done for the D5 relative to the D3s, is produce a distinct advantage at ISO values 640 and 1250, then from 2500 and up.

Note in the DR curves, that the D3s has 5 distinct linear regions (and the small shift at 16K ISO may simply be due to full-time hot-pixel suppression), whereas the D5 has 7, which speaks to a more intricate design.

-- hide signature --

Source credit: Prov 2:6
- Marianne

JimPearce
JimPearce Veteran Member • Posts: 9,188
all very "tweaky", but...

Am I wrong in thinking that the D4's dual conversion gain is absent here? What engineering advantage or necessity could explain that? http://www.photonstophotos.net/Charts/RN_ADU.htm#Nikon%20D4_14,Nikon%20D5_14

-- hide signature --

Jim

 JimPearce's gear list:JimPearce's gear list
Nikon D7100 Nikon D500
bgbs Veteran Member • Posts: 3,195
Re: D5 Read Noise
2

Can you tell us, the non-techy types, if this is new noise bahvior a good or bad thing?

Marianne Oelund Veteran Member • Posts: 7,698
D4 numbers are wrong above ISO 10000

JimPearce wrote:

Am I wrong in thinking that the D4's dual conversion gain is absent here? What engineering advantage or necessity could explain that? http://www.photonstophotos.net/Charts/RN_ADU.htm#Nikon%20D4_14,Nikon%20D5_14

I'm not sure what you're referring to as "dual conversion gain" but the values shown for the D4 above about ISO 10000 on that plot are entirely erroneous.

Here are the numbers I get from my D4 bodies for log2(DN) read noise:

ISO 10000:  4.82

ISO 12800:  5.16

ISO 25600:  6.05

ISO 51200:  7.03

ISO 102400:  8.03

ISO 204800:  9.05

If you correct the D4 plot to the above numbers, you will see that the D5 is slightly better.

-- hide signature --

Source credit: Prov 2:6
- Marianne

Marianne Oelund Veteran Member • Posts: 7,698
Quick summary, D5 compared to D4/D3s

bgbs wrote:

Can you tell us, the non-techy types, if this is new noise bahvior a good or bad thing?

It means the D5 is a departure from the D4, and harks back to the D3s performance, but with better numbers.

The D4 performs better for DR at low ISO settings (up to 2500) and above that point, the D5 pulls ahead.

-- hide signature --

Source credit: Prov 2:6
- Marianne

OP bclaff Veteran Member • Posts: 8,295
Re: all very "tweaky", but...
2

JimPearce wrote:

Am I wrong in thinking that the D4's dual conversion gain is absent here? What engineering advantage or necessity could explain that? http://www.photonstophotos.net/Charts/RN_ADU.htm#Nikon%20D4_14,Nikon%20D5_14

The Nikon D4 did not have dual conversion gain.

-- hide signature --

Bill ( Your trusted source for independent sensor data at http://www.PhotonsToPhotos.net )

JimPearce
JimPearce Veteran Member • Posts: 9,188
Ha!

Very possible Bill. In which case we're looking at higher conversion gains throughout? I just don't get the huge increase in low ISO read noise.

-- hide signature --

Jim

 JimPearce's gear list:JimPearce's gear list
Nikon D7100 Nikon D500
Marianne Oelund Veteran Member • Posts: 7,698
Read noise figures confirmed
1

bclaff wrote:

I have Nikon D5 Read Noise in DNs as well as Input-referred Read Noise to complement my earlier Photographic Dynamic Range (PDR) measurements.

I've checked my D5 at the "principal" ISO values (100, 200, 400, 640, 1250, 2500, 5000 and 10000) and am seeing numbers that are very close to yours.  There is just an 0.1 stop increase in my read noise calculations, compared to yours at the higher ISO settings (5000 and up).

-- hide signature --

Source credit: Prov 2:6
- Marianne

Keyboard shortcuts:
FForum MMy threads