Nikon D4 Read Noise Measured and Compared (revised)

[bclaff]

With the help of Alin Popescu who posted More samples Nikon D4 on his blog I have Read Noise values for the D4 measured from NEF files for a limited set of ISO values.

To better share these finding across multiple forums I have posted two articles at my site.
The direct links are:
FX Read Noise Comparison and D4 Read Noise
They are also listed under Top Topics at the top of my site.

I anticipate a great deal of interest in this material.

Please feel free to contact me directly by email if your comment or question is not of general interest to others.


Bill

New data from Mircea Berzegheanu. See his blog

[BernardLanguillier]

Thanks, that sounds impressive as it would indicate that the improvement from D3s to D4 is larger than the one from D3 to D3s.

Does that take into account the increase of resolution, or is it at individual pixel level?

Thank you.

Cheers,
Bernard

[bclaff]

Bernard,

Does that take into account the increase of resolution, or is it at individual pixel level?

The first chart is per photosite, the second is adjusted for photosite size.

Regards,
Bill

[BernardLanguillier]

Bernard,

Does that take into account the increase of resolution, or is it at individual pixel level?

The first chart is per photosite, the second is adjusted for photosite size.

OK, understood, thank you.

Cheers,
Bernard

[bill t.]

It would seem the D4 is more than 1 log2 unit better than the rest.

Is it possible to extrapolate what that might mean on a screen image or print?  Twice the print size for the same visible amount of noise?  Twice the post processing contrast increase for the same apparent visible noise?  I'm just too empirical sometimes.

[BernardLanguillier]

It would seem the D4 is more than 1 log2 unit better than the rest.

Is it possible to extrapolate what that might mean on a screen image or print?  Twice the print size for the same visible amount of noise?  Twice the post processing contrast increase for the same apparent visible noise?  I'm just too empirical sometimes.

I stop better in terms of noise?

Cheers,
Bernard

[bjanes]

With the help of Alin Popescu who posted More samples Nikon D4 on his blog I have Read Noise values for the D4 measured from NEF files for a limited set of ISO values.

To better share these finding across multiple forums I have posted two articles at my site.
The direct links are:
FX Read Noise Comparison and D4 Read Noise

Bill,

Thanks for posting the data. I am confused by the use of the term EV to express the ISO. As generally used EV is a measure of exposure, including both the shutter speed and aperture as discussed on Wikipedia, although exposure compensation can also be expressed in terms of EV. How does one convert your EV value to ISO?

Earlier Nikons have increased read noise at low ISO as indicated by flattening of the DR vs ISO plot at low ISI as explained brilliantly by EJ Martin. The D3x and D7000 have been called ISO-less, since their curve is linear starting at the lowest real ISO (1 stop of DR is lost for each doubling of the ISO). The dotted lines for your D4 curve shows flattening at low ISO, but no data points are shown. Is the D4 ISO-less?

Also, do you have any information on the sensor gain at various ISOs so one could convert the read noise in ADUs to electrons?

Thanks,

Bill

[torger]

It would seem the D4 is more than 1 log2 unit better than the rest.

Is it possible to extrapolate what that might mean on a screen image or print?  Twice the print size for the same visible amount of noise?  Twice the post processing contrast increase for the same apparent visible noise?  I'm just too empirical sometimes.

I'm not sure. These low light kings are usually at their best when working at ridiculously high ISOs, producing (shot) noisy but useful images where lesser cameras cannot produce images at all. Great for photojournalists.

It may not have that much of a practical meaning for a landscape or studio photographer shooting at base ISO. We'll see.

[ErikKaffehr]

Hi,

One EV means twice the ISO. So a one EV improvement may mean that image quality would be the same at 6400 ISO as at 3200 on the other camera.

Read noise affects DR and the darks and not noise in midtones which are usually dominated by photon statistics.

Best regards
Erik

Bill,

Thanks for posting the data. I am confused by the use of the term EV to express the ISO. As generally used EV is a measure of exposure, including both the shutter speed and aperture as discussed on Wikipedia, although exposure compensation can also be expressed in terms of EV. How does one convert your EV value to ISO?

Earlier Nikons have increased read noise at low ISO as indicated by flattening of the DR vs ISO plot at low ISI as explained brilliantly by EJ Martin. The D3x and D7000 have been called ISO-less, since their curve is linear starting at the lowest real ISO (1 stop of DR is lost for each doubling of the ISO). The dotted lines for your D4 curve shows flattening at low ISO, but no data points are shown. Is the D4 ISO-less?

Also, do you have any information on the sensor gain at various ISOs so one could convert the read noise in ADUs to electrons?

Thanks,

Bill

[bjanes]

One EV means twice the ISO. So a one EV improvement may mean that image quality would be the same at 6400 ISO as at 3200 on the other camera.

That's what I meant when I said that EV can be used to express exposure compensation, but how does one translate an EV of 6.0 as in Bill's graphs to the actual ISO? The values are relative and not absolute. However, if one starts at an ISO of 1.0 as EV 1, then EV 7 would correspond to an ISO of 128.

Read noise affects DR and the darks and not noise in midtones which are usually dominated by photon statistics.

That is true, but how does it relate to my post?

Regards,

Bill

[bclaff]

Bill,

Yes, I use the term EV in two ways.
The first is as you indicate as part of the APES (or APEX) system of exposure.
Lurkers see Additive Photographic Exposure System (APES)

The second is simple as a unit of measure, instead of log2

In this case I really should have been saying EVS
I should start being more careful! 

Regards,

[ErikKaffehr]

Bill,

I was not aware that I were responding to your post. The reason I mentioned it was that there where some questions of what the figures would mean in practice and I wanted to shed some light on the relation between EVs and image quality.

The issue is that I guess that image quality is mostly affected by the number of photons collected and not so much by the noise floor which is related to read noise. I'm not sure that this is evident for all readers.

Best regards
Erik

That is true, but how does it relate to my post?

Regards,

Bill

[bclaff]

Erik,

Not sure to which of the two Bill's you directed your comment. 

I would prefer to use what I call "Photographic Dynamic Range" (PDR) rather than "Engineering Dynamic Range" (DR) to evaluate.
But at this point I only have DR figures to present.
The difference is the PDR only extends into the shadows for far as the final SNR is acceptable as opposed to all the way down to the read noise.
However, even though the low end has photon noise, read noise dominates so the read noise floor remains a good indicator.

Regards,
Bill

[ErikKaffehr]

Hi,

I was addressing 'bjanes"...

I guess that I absolutely agree. What I wanted to indicate is that we have two different noise scenarios, one in mid tones and one in the shadows. Read noise can be reduced and that would make shadows cleaner, but mid tones would not be affected.

There were some questions on the forum what EVs mean in photographical terms. So what I really wanted to say was:

1) A 1EV advantage corresponds to a doubling of ISO

2) An advantage of one EV in read noise does not necessarily lead to a corresponding increase in midtone quality.

Best regards
Erik

Erik,

Not sure to which of the two Bill's you directed your comment. 

I would prefer to use what I call "Photographic Dynamic Range" (PDR) rather than "Engineering Dynamic Range" (DR) to evaluate.
But at this point I only have DR figures to present.
The difference is the PDR only extends into the shadows for far as the final SNR is acceptable as opposed to all the way down to the read noise.
However, even though the low end has photon noise, read noise dominates so the read noise floor remains a good indicator.

Regards,
Bill

[bclaff]

The Nikon D4 Read Noise measurements has fueled some speculation about incredible improvement over the other FX models.
So I'm doing something I usually don't do, I'm adding estimates for the D4 to a couple of my charts.
The estimates are based not only on the Read Noise measurements but also on Full Well Capacity and gain estimates.
I hope this puts the Read Noise data more into context.

The direct links are:
Photographic Dynamic Range Chart
Photographic Dynamic Range Chart ISOs at PDR of 6.5 Chart


Bill

[bclaff]

I have revised my D4 estimates downward but the D4 still looks like a "winner".
For details see the articles and charts at my site.

Bill
P.S. - Alin's blog link no longer works

[bclaff]

If this topic interests you then you may want to revisit the links in the original post as there is more D3S and D4 data that has been gathered.


Bill