Nikon Z6, Z7 and Canon EOS R low-light ISO scores?

Started Dec 4, 2018 | Discussions thread
John Sheehy Forum Pro • Posts: 24,315
Re: Nikon Z6, Z7 and Canon EOS R low-light ISO scores?

Maarten Droogne wrote:

John Sheehy wrote:

Maarten Droogne wrote:

Badwater wrote:

Maarten Droogne wrote:

With the DxO site apparently busy with moving, I was wondering if there is any other site/organisation or whatsoever with objective measurements on those sensors? I wanna compare them to the Sony A7 III and A7R III but most sites dont have any info about this specificaly.

Thanks in advance!

Is there a reason you did not look at the DPR info?

Sorry for the late reply, not always alot time to answer, but very interesting reads, so thanks everyone for that!

I did actually have a look at the DPR scene info, and wow! That is such an incredibly usefull tool. Together with the ISO scores, it does really help to get some insight into it all. Not sure how to notice ' read noise though'... In any case I'm now just waiting for the new sony APS-C camera, and I'm probably buying that one. More than enough reasons to go for that one as long as the ISO score doesnt dissapoint.

What is an "ISO score"?

If you are talking about a metric like DxO "Low light Sports ISO" or whatever they call it, where they cite the highest ISO that meets an IQ standard, those are usually pretty bogus for gauging actual high-ISO performance. They are often meaningless measures of noise in the brighter tones of medium ISOs, and have almost nothing to do with real high-ISO performance, because they report relatively benign photon noise with much more weight than the very ugly read noise that haunts deep shadows and even brighter tones in truly high ISOs.

I have accepted to not take the DxO scores at their words, and I was basing myself on the photon scores. Those might also not be perfect, but I do like numbers. Could you recommend a place to look at numbers that take into account the more various types of serious underlying noise?

You really also need to look at noise, IMO. The numbers are nice, but they don't tell the whole story, and sometimes don't even tell the same story that you think they do. PDR, for example, does not report or intend to report noise vs exposure. PDR is about SNR below the assumed RAW clipping point, and the RAW clipping point for highlights varies quite a bit across cameras in regard to how much headroom there is between middle grey and it. It can be as little as 2.2 stops or so, or be several stops or so. High DR or DPR sometimes only is measured because

What you will see in the DPR studio comparison tool is different than what you will see at DxO, and different than what you see at Photons2Photos. I use all three, aware of the individual weaknesses of all three.

DxO's main weakness in the DR charts is that it ignores the spatial correlation in noise, which is just as important as the amount of noise. It moves the data points horizontally, away from the ISO settings, as needed, to compensate for the variation in headroom, so that the trendlines drawn through them are roughly accurate for representing real measured read noise relative to a standard exposure.

PDR doesn't take sptatal correlation into account, either, nor actual exposure. That's why all the data points are on exact ISO settings, in vertical columns when comparing cameras. The PDR of a camera would not change if you threw a 2-stop ND layer on the top of each photosite; it is only a relative measure across the tonal range at ISO settings, where the number of the setting means nothing exact in relationship to the PDR.

The DPR studio tool attempts to show what the cameras actually produce with the same exposure, but is subject to NR and sharpening difference, especially when comparing in JPEG mode, or across manufacturers. You must interpret what you see there in light of possible NR and sharpening differences (and sometimes small tonal curve differences). Extended highlights are lost in the Studio tool, so you are always looking at input-referred noise, and never DR, per se, with the unknown headroom that isn't shown.

Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow