I care less about DR because I very rarely do +3 exposure or more. However, when I look at the dpreview studio scene, the high ISO noise at 6400 and 12800 per pixel (not normalized to the same output size) seems at the same level as RP. This is disappointing to me... I'm already not very happy about the high ISO noise of my R5. Sigh.
-
Welcome to the new forums! Please read this first. For known issues we are working to resolve, click here.
R5II High ISO Noise Seems Disappointing to Me
- Thread starter smilerliu
- Start date
J
jjbinx
Guest
Karl_Guttag
Veteran Member
I downloaded the Raw Files for the R5 and R5mk2 at ISO 6,400 and 128,000, and the lighting is quite different between the pictures.
For example, look at the aluminum paint tray in the bottom right corner and the bottles in the middle of the bottom. In the R5 Mk 2, the aluminum paint tray is gray, whereas on the R5 Mk1, it is almost white.
Looking at the gray in the center of the picture the R5mk1 is about 0.5 stops brighter than the R5mk2.
The R5mk2 has much more contrast than the R5 in the overall picture. I reduced the R5mk1 by 0.5 stops to "normalize" it to the R5mk2, and still, the darker gray on the outside is much lighter than the same area in the R5mk2. Also the "gray" in the center of the R5m2 is off the red and blue a bit.
It is hard to compare noise when the pictures are different in lighting, white balance, and contrast.
I'm not complaining about DPReview's method, as I know they are trying to make an objective comparison, but it is tricky to get the exact same shot about 3 years apart. I'm just suggesting that this is not a definitive result.
BTW, on the ISO6,400 shot, I also looked at the Sony a7R picture, and its exposure was about 0.5 stops darker than the R5mk2 or a full stop darker than the R5mk1.
Quarkcharmed
Veteran Member
The DR at base ISO and at high ISO took a hit it seems.
Note that the second gain kicks in at ISO 400 in the R5 and at ISO 800 in the R5II. Also the R5 doesn't apply the infamous noise reduction in raw from ISO 800. So we can compare ISO 100 and ISO 800 between those cameras.
Looking at the deep shadows:
... and comparing the EFCS vs EC in the R5II:
we can tell the R5II has a significantly lower dynamic range at ISO 100 (say around 1 stop worse) and worse high ISO performance.
The ES may be 14 bits but it's still noticeably worse than the EFCS mode.
Yes those aspects look disappointing so far.
I don't think the noise difference is due to the brightness difference. If so, the darker image should look less noise; it will only look noisier after boosting the brightness.
lokatz
Veteran Member
Interesting but is it authoritative? In real world photography who intentionally underexposes by 6 stops at 100 ISO, then boosts by + 6EV in post? I prefer to see correctly exposed images at high ISO to evaluate noise. In that same linked DPR test page, the ISO 6400 image has less noise than ISO 100 +6EV. I have very clean (i.e. acceptable, not completely noise-free) images from my R5 at ISO 12800 and it is rare to need higher. Nor do I often need to “recover” (=boost) shadows by more than about 2 stops. In general, over-boosted shadows are not just noisy but also look unnatural in the context of the overall tonality of the image.
I have no axe to grind over R5ii vs R5 as I won’t be buying one. But I do doubt the concern over high ISO noise in real photography as opposed to tests. I’m happy to wait to see more scientific studies.
lokatz
Veteran Member
You said you find it surprising that Canon would be releasing a long-awaited mark ii body, with a new sensor but the same resolution, that underperforms its predecessor. Well, Petapixel doesn't, and I don't, and the article I referenced explains why.
I also think that the term 'underperform' does not apply here. The R5II outperforms the R5 in most areas, but very few things in this world come without some kind of trade-off.
Last edited:
This thread is supposed to be about high ISO noise and not about DR... I think for amateurs with slow lenses ISO 6400 and 12800 are typical use cases. I'm disappointed that R5II sensor performs worse in such typical use cases.
Of course, opinions vary. No problem there
I only used that term in relation to noise performance. The sensor change is the issue, of course, but this is not such a major change compared to say the Sony A9ii which does indeed take a hit in IQ. And in any case I’m interested in real world photography. The R5ii offers several real improvements in overall utility, just not enough to incentivise me to change on this occasion.
Quarkcharmed
Veteran Member
DR is measured through the noise.
Surely noise and DR are inextricably linked. The OP referenced noise at ISO 100 +6EV. Look at the test page and compare that result with 6400 correctly exposed, which shows far less noise. I too would be disappointed if noise at 12800 was proven to be significantly worse in the mk ii, but I don’t believe that those are “typical use cases” even “for amateurs with slow lenses”. Occasional maybe.
Quarkcharmed
Veteran Member
Pushing the shadows is pretty much normal if you do ETTR. A +6EV push brings deep shadows into midtones which makes them clearly visible; it's normal to push them by 3-4 stops, they still will be visible
What DPR comparison tool shows is the R5II's shadows are noisier than R5's. That means lower dynamic range and less latitude for editing, even though you may not be pushing the shadows by +6 stops.
Time and market will tell if Canon and others have made the good choices. It also points out that maybe we are at the edge of what today's technology can deliver and that now camera makers have to make trade offs.
I was thinking about selling may R7 and R6 II next year to fund the purchase of the R5II. I will use my money for lenses instead and wait for the next camera generation. An R6III with the same read speed and the low noise as the R3 would be nice.
Last edited:
I'm the OP and the thread is about high ISO noise without exposure changes. I think you are referring to a different thread.
The dpreview studio scene (not the DR scene) already shows that R5II high ISO noise is on par with RP at the pixel level and worse than R5.
SigZero
Leading Member
It seems that people think that noise levels on CMOS sensors can go down without any limit. I think that most modern CMOS sensor are at the border. It's very difficult to get into anything better. Especially with stacked designs.
It seems that R5mk2 has dual-gain switch at higher ISO than R5 (seems like something between ISO 400 and 800) - so the case of pushing ISO 100 6 EVs will be worse in that case.
Still don't understand why DPR is not taking into account dual-gain constructions in their shadow-push measurements (at least as an alternative to lower base ISO)
Br, Pawel
It was you who mentioned DR, I was referring to noise as shown in the DPR test images referenced by the OP. I don’t think we are actually in disagreement.
Last edited:
Similar threads
- Question
Keyboard shortcuts
- f
- Forum