Richard Butler

Richard Butler

DPReview Administrator
Lives in United Kingdom Seattle, United Kingdom
Joined on Nov 7, 2007

Comments

Total: 5080, showing: 1 – 20
« First‹ Previous12345Next ›Last »
In reply to:

Village9991: what if this kind of test is not correct? Sensors have more information in the highlight are than in the dark one so you can get better (much better) images DARKENING digital images instead of LIGHTENING. so exposing to the right let you obtaing much better images. It would be interesting (for me instead) this kind of test.

It looks a lot to me like your 'overexposed' image has only clipped white regions, so you won't notice if they're unrecoverable. Everything else in the scene is within the camera's dynamic range and is cleaner because you've given it more light.

From what you've presented, that doesn't demonstrate anything about highlight DR. All you appear to have shown is that you get optimal noise performance by exposing to the right.

Link | Posted on Jul 24, 2017 at 17:23 UTC
In reply to:

Cswitch: Maybe 5DIV sales are not as high as expected ...
To cripple the 6D II sensorwise might be a way to "convince" potential buyers to purchase the (outside US too expensive) 5DIV

Given the decision over the chip would have been made long before the 5D IV hit the market, it's fairly safe to say that, whatever Canon's thinking was on this, it has nothing to do with 5D IV sales.

I also think it's supremely unlikely that Canon chose this chip with the aim of *limiting* DR: it's far more likely they decided to iterate on an existing design, concluding that people had seemed happy with it, previously.

Link | Posted on Jul 24, 2017 at 00:38 UTC
In reply to:

Robemo: Can DPR please give the RAW file of the beach scene?

I'm really sorry about that. Looks like I missed the 'http://' off the beginning of the link. It should now be fixed.

Link | Posted on Jul 24, 2017 at 00:33 UTC
In reply to:

Mais78: wow, I sold my 6D to upgrade but the Mk II seems even worse in the noise department. I regret selling.

It's true, you can add more noise reduction and hide the issue.

Bill's data is taken from different images than the ones we present and, so far as I know, are processed using different software. His findings our consistent with what we're seeing.

Link | Posted on Jul 24, 2017 at 00:26 UTC
In reply to:

AwkwardSwine: In Seattle you can see Dunkirk in full 70mm at the Cinerama. In fact they have a full roster of classic 70mm films playing this summer.

https://www.cinerama.com/News/May-2017/Cinerama-to-show-Christopher-Nolan-s-Dunkirk-in-70.aspx

AwkwardSwine - I think you're both describing the same thing. The second image down in the infographic *is* 70mm, which means it's cropped, relative to the 1.43:1 ratio it was shot in.

Link | Posted on Jul 22, 2017 at 01:47 UTC
In reply to:

Robemo: Can DPR please give the RAW file of the beach scene?

I've added a link in the caption.

Link | Posted on Jul 21, 2017 at 22:51 UTC
In reply to:

graybalanced: The article says the decision to shoot with film has "had unexpected benefits" but when I followed the link, the "unexpected benefits" seemed to be the ability to save the sunken film from the sea through an old technique of keeping it wet until developed.

In the link Christopher Nolan said "Try doing that with a digital camera!" But DPReview has published numerous accounts of digital camera cards being retrieved from various bodies of water with the images intact, some submerged for many months, without having to resort to any workarounds whatsoever, so the way the film had to be babied just to get it to survive doesn't seem like a "benefit."

What it sounds like was Nolan was happy about being able to save film with a clever workaround, without recognizing that the problem probably wouldn't have been a problem at all, if the footage had been shot on a digital camera card.

I'm not anti-film...just sayin'

Those SD card stories crossed my mind. Not sure the same is true for SSDs.

Link | Posted on Jul 21, 2017 at 20:59 UTC
In reply to:

keeponkeepingon: So I look a the images at 0 EV 100 ISO and I notice that the 6DII has the sharpest, best looking rendering of the D750/80D/6D both in the center but especially in the corners.

My first thought isn't. Wow the 6DII is rocking the studio sample it's " wow Another useless dpreview apples and oranges comparison."

I'll voice dpreviews/fanboys reply for them: "Say what? It's a CONTROLLED studio test you idiot!!!!"

Here's the thing, in case anyone else has failed to point out the the dpreview studio test has no clothes:

D750 lens: 50mm F1.4 (T stop F1.4)
80D lens: 50mm F1.4 (T Stop F1.6)
6DII lens: 85mm F1.8 (T Stop F2.1)

Gosh even on the same platform you can't use the same lens? And even with the same size sensor you can't use the focal length?

At least for the 6D/6DII the comparison the lens is the same but the lighting looks a bit different and for some reason the 6DII "info" mentions mirror lockup while the "info" on the 6D does not (was mirror lock up used for the 6D?).

We use 85mm lenses on FF and 50mm lenses on ASP-C, wherever possible. The D750 was shot with the 85mm F1.8, not the 50mm (it was a data entry error - I've checked back against the EXIF).

Link | Posted on Jul 21, 2017 at 19:51 UTC
In reply to:

Scorehound_ca: I recall not too long ago the Sony a9 had one of its tests redone because of an error by DPR staff. Again, we have all these comments based on one person's use and not real world shooting in real situations.

I think it would be best to wait until real world reviews come out from photographers using retail shipped cameras.

Well, two independently conducted tests (ours and Bill Claffs, based on different methodologies and processes) and real-world shooting of two production bodies by a series of reviewers.

That, significantly, give results that are wholly consistent with older Canon sensor designs.

Link | Posted on Jul 21, 2017 at 17:47 UTC
In reply to:

PhotoUniverse: WOW. My Nikon D500 is better and it's a APS-C!

I know it's not everything for a camera, but still

I've not seen any explanation of any design or technique for achieving good high ISO performance that impinges on low ISO DR. As such, I'm not sure any such trade-off is being made.

(Actually, that's not quite true: the huge pixels of the a7S and S II mean that it doesn't have sufficient bit-depth to capture all its DR at base ISO, but that's a different issue).

There might be a trade-off between *speed* and low ISO DR. Or, just as likely, it may simply be that Nikon iterated on the design of the D4S chip, rather than risk moving across to the column ADC that it hasn't used in high-speed applications.

Link | Posted on Jul 21, 2017 at 17:32 UTC
In reply to:

RaghavBaijal: ROFL...

https://www.dpreview.com/reviews/image-comparison/fullscreen?attr144_0=canon_eos6dmkii&attr144_1=fujifilm_xt20&attr144_2=canon_eos6dmkii&attr144_3=sony_a6000&attr146_0=100_5&attr146_1=200_5&attr146_2=100_5&attr146_3=100_5&normalization=compare&widget=542&x=0.124786377&y=0.5025702

You're more than welcome. We provide the information and the files precisely because we can't second-guess everybody's needs and use-cases.

If this lets you achieve what you want to do (and, as the gallery shows, the JPEG colour is rather attractive), then that's great.

Link | Posted on Jul 21, 2017 at 03:11 UTC
In reply to:

RaghavBaijal: ROFL...

https://www.dpreview.com/reviews/image-comparison/fullscreen?attr144_0=canon_eos6dmkii&attr144_1=fujifilm_xt20&attr144_2=canon_eos6dmkii&attr144_3=sony_a6000&attr146_0=100_5&attr146_1=200_5&attr146_2=100_5&attr146_3=100_5&normalization=compare&widget=542&x=0.124786377&y=0.5025702

We don't put numbers on DR. Bill's number is something like 9.1, since I think using a >1 SNR threshold makes more sense.

If you're printing to 5 or 6 EV paper then yes, you'll have some flexibility in terms of what you include and what you clip to black but you'll still have several stops less than if you used most other modern cameras.

As I said in the article, you'll have more than the ~8.3EV that typical JPEG would include (though perhaps not much more than a JPEG with ALO turned on before your shadows started to get noisy). If that doesn't matter to you, that's great and it means you can start with Canon's rather pleasant JPEG colour. However, it's significantly behind what contemporary cameras produce and gives you less processing options than an 80D. That's newsworthy.

Link | Posted on Jul 21, 2017 at 00:00 UTC
In reply to:

ThomasH_always: This is getting ridiculous: Every technical solution has levels of compromise, and in this case we are looking at, what: corrective settings of +3EV, +4EV? When in gods name do you need it? Only when you deal with a badly failed photograph, which has a great personal value to you (memories, special situation). Technical perfection irrelevant, content priceless.

This craze resembles indeed the Hi-Fi madness in the 80thies: We red brochures and made decisions based on 0.012% versus 0.013% noise ratio or signal distortion. We do the same with the cameras these days. (Thanks Thom Hogan for reminding us about that in his great article "Stop complaining".)

6D II will be a success, despite these extreme corner cases, in which someone else has a better sensor. I cannot remember having have done a +3EV correction even once, and I just turned in "45 years of taking pictures", since the Nikkormat's ruled the planet.

If your pictures need +3EV, +5EV correction, rethink your technique.

Again, the point isn't about pushing an image by 6EV or by the lesser amount in the sunset image. It's that, as soon as you start pulling shadows up (and there are many instances where you might want to), this camera will be noisier, not only than the competition but also than the EOS 80D, which it should comfortably outperform.

Link | Posted on Jul 20, 2017 at 23:44 UTC
In reply to:

Frank_BR: It looks like DPR is making a mountain out of a molehill. The ultra-professional Nikon D5 which costs $ 6500 has dynamic range more or less equal to the Canon 6D Mark II for ISO below 1600. Check out here:

http://www.photonstophotos.net/Charts/PDR.htm#Canon%20EOS%206D%20Mark%20II,Nikon%20D520Mark%20II,Nikon%20D5

[This is not news to us](https://www.dpreview.com/news/9402203921/nikon-d5-shows-drop-in-dynamic-range).

We think Nikon reverted to one of its older (high speed) designs, which means its results slipped behind the column ADC chips it'd used elsewhere. We called that out, too.

Link | Posted on Jul 20, 2017 at 23:40 UTC
In reply to:

RaghavBaijal: ROFL...

https://www.dpreview.com/reviews/image-comparison/fullscreen?attr144_0=canon_eos6dmkii&attr144_1=fujifilm_xt20&attr144_2=canon_eos6dmkii&attr144_3=sony_a6000&attr146_0=100_5&attr146_1=200_5&attr146_2=100_5&attr146_3=100_5&normalization=compare&widget=542&x=0.124786377&y=0.5025702

Since DR is the difference between the brightest captured tone and the point where noise becomes *unacceptable*, you can see that it's difficult/impossible to put a number on (which is why I didn't, in the article).

DxO uses an SNR cut-off of 1 (the point at which signal and noise are equal). That's a sensible definition from an engineering perspective but would be much noisier than most people would put up with. Bill Claff uses a different threshold, which assumes you have a lower tolerance for noise.

So, what Bill considers to be a 9.3EV range for the original 6D, DxO interprets as 11.4EV. It's up to you to decide which of these more closely matches your own tolerance for noise.

If you're used to reading DR numbers based on SNR of 1 then yes, 9EV would seem shockingly low. However, both our tests and Bills lead you to the conclusion that there's significantly less processing flexibility in these files than you'd expect and that you'll see this after fairly minor edits.

Link | Posted on Jul 20, 2017 at 23:30 UTC
In reply to:

melgross: This really is odd. I’m wondering if it’s not some defect. Why would Canon do this on purpose? It makes no sense.

However, as I say, it's reasonable to assume that all brands try to map the saturation point in their Raw files pretty closely to the clipping point in their JPEGs (to avoid repeated underexposure). As such, highlight recovery doesn't vary significantly, brand to brand, once you exclude the perceptual error of pulling near-white highlights down to a brightness where you can better distinguish between them.

Link | Posted on Jul 20, 2017 at 23:08 UTC
In reply to:

melgross: This really is odd. I’m wondering if it’s not some defect. Why would Canon do this on purpose? It makes no sense.

The difference is that there's a hard cut-off in the highlights: once the signal exceeds the full well capacity of the pixel (or, at least, whatever proportion of the full well capacity is mapped to the maximum raw number), you get no additional information.

At the shadow end, the end is fuzzy. As signal drops, the signal to noise ratio drops, partly because shot noise becomes proportionally more significant but also because there's a slight background 'hum' from the electronics. Our tests here show that the background electronic noise of this camera makes the shadows unusable earlier than on its rivals.

We've not seen radical differences between brands in terms of how far above (JPEG) middle grey clipping occurs, other than modes such as Fujifilm's DR 200 and 400, and Canon's HTP, which push middle grey a stop further down the Raw file for more highlight capture, at the cost of 1EV shot noise.

Our next generation test will include a Saturation element, just to re-confirm this.

Link | Posted on Jul 20, 2017 at 22:58 UTC
In reply to:

TrojMacReady: Does the quoted 8.3EV for jpegs include the maximum use of ALO (in other words, is that the input DR or output DR)? Because that basically lifts shadows too to compress a larger DR from the scene into a smaller final DR as used in jpegs and other viewable formats.

TrojMacReady - 8.3EV is a typical number for a default tone curve, without any dynamic range compression. Turn on ALO and HTP and you'd add a fair bit to that. At which point you may well see the additional noise of this sensor.

Link | Posted on Jul 20, 2017 at 22:45 UTC
In reply to:

ThomasH_always: This is getting ridiculous: Every technical solution has levels of compromise, and in this case we are looking at, what: corrective settings of +3EV, +4EV? When in gods name do you need it? Only when you deal with a badly failed photograph, which has a great personal value to you (memories, special situation). Technical perfection irrelevant, content priceless.

This craze resembles indeed the Hi-Fi madness in the 80thies: We red brochures and made decisions based on 0.012% versus 0.013% noise ratio or signal distortion. We do the same with the cameras these days. (Thanks Thom Hogan for reminding us about that in his great article "Stop complaining".)

6D II will be a success, despite these extreme corner cases, in which someone else has a better sensor. I cannot remember having have done a +3EV correction even once, and I just turned in "45 years of taking pictures", since the Nikkormat's ruled the planet.

If your pictures need +3EV, +5EV correction, rethink your technique.

@ThomasH_always - it looks like I mis-processed the D750 files when I tried to fix and EXIF glitch this morning. Reload and you should see the correct rendering.

The exposures (which are what define the noise level) are unchanged, it's just that I'd processed them slightly too dark.

Link | Posted on Jul 20, 2017 at 22:38 UTC
In reply to:

melgross: This really is odd. I’m wondering if it’s not some defect. Why would Canon do this on purpose? It makes no sense.

@PredatorsPrey - To be clear, I'm not suggesting trying to recover highlights from the JPEG, I'm talking about recovering Raw highlights over and above what was included in the JPEG (since this is a reasonable definition of whether a highlight is included or needs to be recovered).

If there are multiple EVs of fully usable information above what's already included in the image, then you've underexposed by multiple EVs and your image has more shot noise than necessary.

I don't understand your second point, I'm afraid.

Link | Posted on Jul 20, 2017 at 22:11 UTC
Total: 5080, showing: 1 – 20
« First‹ Previous12345Next ›Last »