Canon EOS 5DS / 5DS R First Impressions Review

That is correct - at higher ISOs the differences become minimal, b/c the camera's analog 'ISO' amplifier amplifies the signal as to get it above the high read noise floor of the Canon camera. The differences against Sony/Nikon/Pentax cameras are most drastic at lower ISOs (100, 200), where minimal signal amplification is applied in-camera.

You can read more about all this in our read noise article.

so on page 7 it should read...

'Compared to its immediate competition, the 5DS R has near identical ISO-invariant as a NikonD810.'
?

If I look at the Raw DR: Exposure Latitude widget, with the 5DS R alongside the D810, in terms of matching noise I find the D810 can be set to have +1 EV extra pull to the 5DS R, but no more?

If the D810 is given +2 EV, or 4x lower exposure, it has far more noise than the 5DS R.

+1 EV pull is 2x less exposure, not 5x. At 5x, or +2.5 EV, the noise doesn't match at all. I'd say all we're talking about is one stop (+1 EV), or less once the 5DS R is down-sized to 36MP?

@left eye: If you look at the SNR curves here, for ~6.6 dB SNR, the D810 only needs a Raw value of 0.0474 while the 5DS needs a Raw value of 0.223. 0.223 / 0.0474 = 4.7. So on the lower end, the D810 can withstand nearly 5x lower exposure for the same SNR.

Our Raw DR scene does not tell the full story, b/c it doesn't have tones ranging from black to white (lowest to highest end of what these sensors can record, that is). If you're looking at some midtone, then you're also conflating shot noise effects.

In other words, for a brighter signal, of course 4x lower exposure on the D810 gives you more noise than 4x higher exposure on a 5DS R --> b/c 4x as much light means much less shot noise, which is a property that scales more with sensor size (& these sensors are the same size).

In other words, brighter tones are shot-noise limited, & their SNR will depend largely on sensor size/efficiency, which is similar btwn these cameras. Different story for dark tones.

In DPP just turn down the sharpness if it bothers you.

Correct. It is a matter of costs. Just make all the cameras with and put the cancelling filter on the ones without. But nothing in the path is the best.

Yep.

actually changning the thickness of the sensor stack by removing the AA filter is a little more complicated than that.

The thickness of the sensor stack. How thick, well, no.....No AA filter is better then canceling. Extra stuff in front of the sensor can never be good.The newest Nikon, D810 is made that way, which is an improvement over using a canceling filter, it was lower cost. Cancelling ok, no filter better.

Cmon, there is purple fringing (LoCA) all over the image, from center to the corners. For instance, supposed black text on white background with 'EF 80mm f/1.8 USM' become purple text on white background. Conclusion is that lens is not up to the task or it's not adequately calibrated / micro adjusted, in other words, correctly focused.

Regarding tests and DPRev policy, some modernization is needed asap. ;) For now, we are lucky to get even test shoots with modern third party lenses.

I am very glad DPR did these test shots with the Otus-lenses. The 85 f/1.8 are good at resolving detail (sharpness), but it is quite clear to me that the 85 f/1.8 falls behind when it comes to contrast and purple fringing. Further, it is also very clear to me, that the Nikon f/1.4 outperforms both the f/1.8, and is on par with the Otus at f/5.6.

I still can't get over DPR's "policy" when it comes to lenses for testing. Why not use Canon 70-200 f/2.8L II IS? It is sharper than the 85 f/1.8, has better contrast and has less fringing. Further - and important - if you go to BHPhoto, selects Canon lenses, and sort by best sellers, the 70-200 comes in at number 4 on the list, and the 85 f/1.8 is number 7. This indicates that people are more likely to use the 70-200 on a Canon FF body, than they are to use the 85 f/1.8. If you do the same with Nikon, the 85 f/1.8 comes in at number 4, and the 85 f/1.4 comes in at number 18. So why exactly do you choose the 85 f/1.4, but not the EF 70-200L?

Myself, and many others, question why DPR uses the way more expensive Nikon 85 f/1.4 on the D810, and compare it to the cheaper Canon 85 f/1.8. If the Nikon f/1.8 is so good, why don't DPR change the "official" test pictures of the D810, so that one can compare directly with lets say the D800 and D800E (which has pictures from the 85 f/1.8)?

It is fair of DPR to publish pictures of the D810 and 85 f/1.8, but most people will only compare with the prictures you get when choosing the D810 in the comparison widget - and that is the 85 f/1.4.

So, that is why I say, that if DPR choose the way more expensive Nikon 85 f/1.4 over the f/1.8, and that gives the D810 a little edge, why don't DPR try to give the Canon 5DSr the same advantage by using a better optic - especially when it is a lens that very many potential buyers are likely to have.

Thx for sharing these infos, now i need the cash for a 5ds

I have not experimented with sRAW as I can see no point in throwing away information. For reference, sRAW is 12.4 Megapixels and mRAW is 28.5 Megapixels. mRAW doesn't change the max burst frame count and sRAW only increases the burst from 12 to 14 frames. Memory cards are cheap, so I can't see the point.

This doesn't make sense for YOU.

Broody , how is the corner sharpness of wide angle Canon EF lenses via adapter on sony A cameras? From all I have read its pretty poor ...unless your experience is different? Better DR is of course welcome , but i am not so sure if its worth it if it comes at the expense of poor corner sharpness...to use the sony A via adapter with EF lenses is of course very good and tempting..especially MF lenses like the shifts from Canon, when the AF speed doesnt matter...just wondering if there is a IQ loss because of the adapter and thick sensor glass? Generally it would seem most mirrorles systems have poorer corner quality than DSLR's ...but maybe someone with experience can tell us if it is so?

In that case the Canon 6d has a terrible AF system with 1 AF-point in the centre.

Great i would take a D750 anyday over the 6d or 5d mk3 in those conditions.

Many sport photographers shoot jpeg because they need to send the photos fast enough to be published. Sports jpegs looks quite good, and don't need lots of processing. Maybe I'm wrong but if I'm wright, a Sony sensor would have no advantage over Canon in jpeg mode.

@Aramandino @IvanM Honestly a good photographer can do good work with most of what's out there. So in no way am I implying your 5Dxxx are unusable.. I was responding to the comment about the Sony resale price.. which to me is an odd way to pick a camera. I prefer choices Sony is making and so far the posts about the new Pro service are good.. because if I were a Pro.. That is a valid factor and Sony has been very slot to step up on that.. But right now my combination of A77m2 and A6000 and mostly Sony Minolta lenses fits what I want my camera to do control wise use of different angles etc. there is nothing the tempts even though each camera has some strengths are unique to its self. But also as a marketing person.. Canon is clearly dealing with the overall DSLR market size loss with trying to create extra body sales per user.. Vs a single 5dMK4 at this time.

Well the thing is, for many years all we had was a split screen little 'dot' in the centre (MF film days) or a matte screen, just think the 'best and most expensive camera in the world' Leica M, also just has a little centre block to focus with...AF cross points arranged all over the viewfinder is just so cool and works wonders and makes life just so easy...but really its not as if one cannot make do with just one little centre AF spot, now is it? I have used my 6D professionally for the last 18months or so, and sure I really wish the AF points were better, but I have had very very few oof images as a result of it. As BMWZimmer said the centre AF point is very good...and the fastest focus of them all is MF of course...have anyone here ever tried 'zone' focusing?

Yes, the 6d is a real pleasure to use! I now shoot jobs up to 1600iso without even thinking about it...

>>sau sa spunem mai pe romaneste

or to say in Romanian

OK.

Nicolaiecostel is a romanian name ans as I am romanian too I spoke to him in my language

I shot a fashion show the other day on the Canon 5D mk2 at ISO2500 and it turned out perfectly, even for fashion. I was shooting a 70-200 L to be fair and the light was good. At some point I had to shoot ISO4000 to ISO5000, and they were still useable within the purpose, albeit not perfect.

@moga et. al. Yes, I am romanian but this is an english forum. Cheers.

odd. I work with SPP and DP2/1M cameras, and i see no color issues and green corners.

no it isn't deconvolution algorithms allow you to do just that.

@nitroman - deconvution filtering is a pretty cpu expensive operation, if you want to see how your camera looks without a AA filter, run it through DPP"s DLO.

The headphone socket was removed to make room for USB3 - as a still photographer with some interest in tethering, I'm very happy about that :-)

3500 euro vs 4200 euro .. yeah, there's a bit difference

I would assume that at ISO 800 and up it will have better DR than a D5500 based on DXO's results for similar cameras. Low ISO DR is REALLY handy if you need 3+ stop shadow pulls, but few people do, high ISO DR is always useful as that's where the DR finally drops below what you can actually reproduce.

Example - if you want 10 stops of DR you can shoot a D5500 to ISO 1600 and above that you have to combine multiple images shot at different exposures. A 6D can do 10 stops to ISO 4000. A 7DmkII is identical to a D5500 at 1600 (and above).

DR-jon.

If you under expose a shot by 3 stops taken at ISO 200.
You can then overexpose the Raw-file by 3 stops. By then you have the same noise in the shot as the 7d MK2 at ISO 1600

However you have 3-stops more room for your highlights.
So even if you need Higher ISO's a ISO invariant camera can have advantages. Even though extracting it is a bit counter intuitive.

I get that, though it's a pain reviewing the images and you don't have anything to show that much DR on. Plus you can play that game a little bit with the 7DmkII provided you start a bit up the ISO scale (I think it has the same read noise at ISO 1000 as a 5300 at 100). Not that playing ISO invariant games doesn't have a small niche place of course. BTW is there a Nikon sensor that doesn't knock at least a third off its read noise going from 100 to 6400? It's not that invariant (e.g D810 5.5->2.6 so more than twice the DR at 6400 over 100 as the signal's the same).

(Extra bit as I was thinking about this.) A Canon 6D has 2.0 Electrons Read noise at 6400, a D750 5.5 at Base ISO, you sure you'd want to use the Nikon that way at (effective) 6400 vs the Canon? Pretty much all the Canons shot at 6400 will give you more than twice the DR of a Nikon at ISO 100 (Canon Read Noises being in the 2s and Nikons in the 5s). Not that the Nikon base-ISO DR isn't great BTW.

That is becouse the D750 isn't ISO invariant to ISO 100.
Just don't push it so low. ISO 200+ in this case.

That ISO 100 setting is more for having a shitload of light on this camera.

While you can play the same game on a Canon camera. Noise will increase much sooner and harder. The latest Canon sensors have improved in this regard. But they have a large disadvantage left in this aspect

Language, BarnET, please!

Dr_Jon - sure, you don't have anything show that much DR on, but that's what tonemapping is for, and as long as you do it well, keeping some blacks and near-whites in the image, you can make compelling use of that DR, yes?

Yes, although that does start to look less natural after a while so is both an art and a skill - well hopefully anyway ;-)

At low ISO it really does need a suitable subject to use 14 DXO stops, I do 2 stop shadow raises a fair bit but the number of times I'd want 3-4 stops is very limiting. YMMV of course.

The Nikons not being all that invariant is interesting though, as losing more than a stop is a lot when 8ish stops might be all that's available.

Not that the low-ISO Nikon DR isn't a good thing, as I've already said, I'm just not so sure about the ISO invariance stacking up so well unless you might have surprise highlights (if they aren't a surprise you should perhaps be shooting at lower ISO in the first place).

I think I'm going with the r BTW, as the sharpness advantage seems tangible and while I hate (!) moire I've seen it on s shots too. It looks like 2-3 weeks wait though - it seems Canon significantly underestimated demand in the U.K.

"I'm just not so sure about the ISO invariance stacking up so well unless you might have surprise highlights (if they aren't a surprise you should perhaps be shooting at lower ISO in the first place)."

Right, but that's just the thing - if those 'surprise highlights aren't a surprise', then, assuming you wish to keep them, you need to underexpose, be it via actually underexposing your focal plane exposure, or by lowering your ISO.

Either method will run the risk of getting other tones (midtones?) you care about closer to your camera's noise floor, and that noise floor might then show when you try to bring those midtones, that you pushed down to preserve the highlights, back.

And that's simply where a camera with a lower noise floor helps. You run into these issues less frequently. And meanwhile, even 1, 2, or 3 stop pushes are helped, as they still show much less noise than similar pushes from a camera with a higher noise floor (or low base ISO DR).

Sure you know all this though :)

The point I was exploring was that the Nikons do seem to have quite a bit more read noise at lower ISOs, so lowering the ISO and underexposing will add significant noise to the non-highlights.
The D750 at ISO 6400 has half the read noise it has at ISO 200, which isn't quite ISO invariant.

It's quite ISO invariant relative to most cameras... about as ISO invariant as it gets, with the D7200 perhaps being the leader here.

It's all relative, right? As sure the D750 may have half the read noise at ISO 6400 as ISO 100 (since you approach upstream read noise levels at higher ISOs, whereas low ISOs include components from both upstream and downstream read noise), but the 6D has 1/13 the read noise at ISO 6400 compared to 100.

Furthermore, I'm willing to bet the largest source of the extra noise at low ISO for the Sony/Nikon cameras is quantization error from the ADCs being too low bit-depth. A totally different issue compared to the off-chip ADC architectures of Canons.

I'd love to see a 16-bit ADC on the D810 to see if pixel-level DR goes above 13.9, which is suspiciously close to the ADC bit-depth... indicating, to me anyway, that the ADC is limiting (unlike on Canons, where it's the architecture itself that's limiting).

Also, 'half' the read noise sounds bad, but take a look at our ISO-invariance widget to take a look at how much that actually affects tones even after the 5-6 EV push. Since we're only talking about a couple electrons here and there, the biggest difference is down in the deepest shadows where SNR is so low due to shot noise anyway that it's largely irrelevant for most users. 'It' being the noise cost from shooting ISO 200 then boosting 5 EV vs. shooting natively at ISO 6400 in-camera.

In other words, 1 or extra electrons of read noise aren't going to make a dramatic to most of your tones, save for the tones you probably shouldn't be using anyway.

Still, I'd prefer that extra cost to not be there, so I look forward to seeing even more progress in this area, possibly by increasing ADC bit-depth.

Also, it's less about being ISO invariant and more about just having less downstream read noise. In fact, ISO variance could be a good thing if it's not due to downstream read noise...!

i really wasn't making any comparisons to the Canons, just commenting on what I saw just on invariance. My (sole!) point being actually shooting a Nikon at ISO 6400 would seem to be a better idea than underexposing ISO 200 except in unusual circumstances. (Ignoring in-camera review and stuff like that.) Nothing more.

There is of course the point that you bring up that by the time you're looking at shadows (say) 7 stops down you don't have a lot of SNR and so the read noise being greater than the shot noise is only hurting you a little (for a 750 you're talking the read noise being nearly twice the shot noise with ISO 200 and about a bit less than it at 6400 - gotta love a camera where the saturation is about 1k at the ISO you thought of, makes stops so easy to do :-) ).

Also that brings in one of my earlier points about people seeming to think ISO 12800 is so low these days, you still are very limited in many ways going above that as pixels are seeing at most 3-figures of electrons.

Oh and good point about 14-bits, a 750 runs about 5 electrons per ADC step at ISO 200 which is getting tight if you boost a 1200 electron range (light level at ISO 6400) since you can only see 240-ish steps from black to white, halving with each extra stop of boost (so simulating ISO 25600 gives 64 levels). That's assuming it's perfect of course.

Also I make it over 3 electrons of extra noise from 6400 to 100 (2.4 vs. 5.5). The 5.5 being not that far off 2.4 and half an ADC step.

(Edit) Oh and finally this is all just commenting in passing, there are very few changes people are making in cameras these days that would actually help in my picture taking. I'm quite happy at 12 stops DR and ability to lift shadows 2 stops 99.999% of the time. Other factors greatly outweigh any improvements over that, even if they would be occasionally useful. (For example 8 fps burst over 4 fps would get me way more shots than 14 stops DR, even though I only shoot that fast occasionally.)

what high ISO? this things native ISO is capped at 6400!

having a 64GB buffer would solve the problem for some ... 1-2 minutes of continuous shooting, and a memory like that is not that expensive (a 64G SSD drive is about 100$)

Blacks look cleaner on 5DS b/c the default ACR profile crushes blacks on the new 5DS cameras for some reason. We need to inquire with Adobe about this.

no, I mean in the samples gallery, the studio shots don't mean much to me. I want to see real-life performance not 'best scenario' shots which are only really useful for studio shooters.

Because it would look laughable. I have seen ISO400 low light shots and they looked like ISO3200 on a D700.

I'm pretty impressed with this sensor. Noise is surprisingly low. (surprising because the low native max suggested it would be bad, but maybe the choice to have circuitry that could only boost 6 stops of gain is part of the reason it's so clean). It will be very interesting to see how the new 42mp Sony stacks up. It will certainly have higher DR, higher sensitivity and may come close in resolution. But what may be even more interesting is whether Sony will sell that sensor to Pentax for their new full frame. Pixel shift in that, or even with the old 36MP sensor, might blow this away for static subjects.

yeah i'm not sure about the scoring on the 24-70 much like slrgear .. seems like bad copies (which did happen when it came out) caused some funky numbers. it should be closer to the 16-35/4 numbers (IMO).

I don't know about Canon, but I've put the Nikon 24-85, which is okay but not pro level, on both a D600 and a D800. You can tell the difference, even with so-so glass.

Now, will you get even more with great glass? Probably, but there's still an advantage to more megapickles.(sic)

Not every wants a representation of reality. I like the colors that are in Canon jpegs. I tend to increase saturation, sometimes a lot to achieve a more graphic look. If the default is more saturation, that's great for me, especially if I am shooting between 10am and 3pm.

ACR/LR profiles for the 5DS/R are applying a much steeper tone curve than typical. That might be what you're seeing.

Well spotted. The difference is much less pronounced when switched to jpeg, still a little - for want of a more technical term - "punchier" than the others but I like it a lot. Anyone who doesn't like what Adobe does with the raw could remember that it's not actually compulsory to use LR. DPP might even do an OK job, worth a try.

actually with the more saturated look to the 5Ds raws via ACR, one could argue that it is showing more noise against the playing field when compared to other cameras.

shouldn't these cameras be profiled before putting through the "studio scene"?