DPreview First impression review 5D IV

[Loswr]

By the way, I see you've gone silent on your claims that it was us who hadn't 'RTFM' when it came to Auto Area AF with iTR. Since it appears you misunderstood how this mode actually worked, I just wanted to make sure you had a better understanding now, so that you don't continue to erroneously claim that our claims were unfounded and misguided.

(1) It was neuroanatomist that failed to comprehend the manual. See my earlier post.

Well, I had decided to let that pass...but since it seems you'd prefer to double down on your lack of comprehension...

Initial acquisition of a subject and tracking of that subject as it moves are different processes, and Canon's iTR system participates only in the latter. Admittedly, the manual section to which you referred is a bit vague, but one might think you'd have gotten a clue from Canon's choice for the feature's name: iTR – intelligent Tracking and Recognition, not iRT – intelligent Recognition and Tracking. In other words, the tracking part comes first, and the metering data assists with keeping an AF point on the subject, after that subject has been identified.

That's how it's phrased in the 1D X II brochure, for example:
[quote author=Canon USA]
This additional exposure and subject information [from the metering sensor] also helps the EOS-1D X Mark II's iTR (Intelligent Tracking and Recognition) AF system maintain focus on a moving subject with a greater level of consistency.
[/quote]

If a more thorough and verbose explanation would help alleviate your failure to comprehend how iTR works, Canon tech-rep Rudy Winston explains:
[quote author=Canon DLC]
One of the key new innovations is using color information to assist the AF system when it’s in Automatic AF point selection mode. Focus upon a subject with all the AF points active and the AF system initially focuses upon it. But behind the scenes, the RGB metering system registers the color, size and shape of the subject being initially focused upon. Armed with this info, the metering system knows not only where the subject is, but its color characteristics — in essence, what it looks like. The Intelligent Subject Analysis system has taken the first step for the photographer.

And as a subject moves around the viewfinder, the metering system continues to update its position, using color and shape to follow the subject. As it moves, the metering system “tells” the AF system where the subject is. Now, the Automatic AF point selection system can continuously change active AF points to keep the initial subject in focus as it moves around the finder. This is the genesis of Canon’s Intelligent Tracking and Recognition (iTR) feature.
[/quote]

To reiterate...with iTR, Canon uses the data from the metering sensor to maintain AI Servo tracking for a moving subject, but not for initial acquisition of that subject. Rather, initial automatic AF point selection is performed by the AF system, according to the parameters which I previously described, namely 'tends to focus on the closest subject' as stated in the manual. Only then, after that initial focus acquisition, do the data from the RGB metering sensor come into play.

So in fact, it was you who misunderstood the way in which iTR functions. The more detailed explanation provided by Canon should clarify your erroneous, misguided comprehension. Perhaps you'll even consider updating the motocross article to correct the misapprehension that DPR is promulgating by suggesting that the 1D X II is 'confused', rather than performing as designed (as I previously pointed out). But...that may be too much to hope for in this case.

 

[JoeDavid]

Maybe DPReview (Amazon) has a backlog of Sony crap to sell... Canon DSLRs sale themselves.

 

[davidhfe]

So in fact, it was you who misunderstood the way in which iTR functions. The more detailed explanation provided by Canon should clarify your erroneous, misguided comprehension. Perhaps you'll even consider updating the motocross article to correct the misapprehension that DPR is promulgating by suggesting that the 1D X II is 'confused', rather than performing as designed (as I previously pointed out). But...that may be too much to hope for in this case.

Honest question from a 60D owner whose camera does none of this—isn't acquisition supposed to be aided by iSA? In other words, isn't iSA + iTR supposed to handle both acquisition and tracking (a la Nikon's 3D Tracking)?

 

[rishi_sanyal]

By the way, I see you've gone silent on your claims that it was us who hadn't 'RTFM' when it came to Auto Area AF with iTR. Since it appears you misunderstood how this mode actually worked, I just wanted to make sure you had a better understanding now, so that you don't continue to erroneously claim that our claims were unfounded and misguided.

(1) It was neuroanatomist that failed to comprehend the manual. See my earlier post.

Well, I had decided to let that pass...but since it seems you'd prefer to double down on your lack of comprehension...

Initial acquisition of a subject and tracking of that subject as it moves are different processes, and Canon's iTR system participates only in the latter. Admittedly, the manual section to which you referred is a bit vague, but one might think you'd have gotten a clue from Canon's choice for the feature's name: iTR – intelligent Tracking and Recognition, not iRT – intelligent Recognition and Tracking. In other words, the tracking part comes first, and the metering data assists with keeping an AF point on the subject, after that subject has been identified.

So because of the order of the words, tracking must come before recognition? That is some convoluted logic predicated upon a lot of assumptions.

I suppose then that since Nikon's '3D Tracking' doesn't even have the word 'recognition' in it - must mean it doesn't do any recognition at all? And yet ends up being precise enough to target the eye of a face...

How long does the recognition part wait before the camera's 'tracked' enough, since you say it comes after tracking?

To reiterate...with iTR, Canon uses the data from the metering sensor to maintain AI Servo tracking for a moving subject, but not for initial acquisition of that subject. Rather, initial automatic AF point selection is performed by the AF system, according to the parameters which I previously described, namely 'tends to focus on the closest subject' as stated in the manual. Only then, after that initial focus acquisition, do the data from the RGB metering sensor come into play.

So in fact, it was you who misunderstood the way in which iTR functions. The more detailed explanation provided by Canon should clarify your erroneous, misguided comprehension. Perhaps you'll even consider updating the motocross article to correct the misapprehension that DPR is promulgating by suggesting that the 1D X II is 'confused', rather than performing as designed (as I previously pointed out). But...that may be too much to hope for in this case.

So the PDAF sensor, which only knows subject distance, can recognize faces? How does it do that?

Also, how does any of this impact upon my more important point: that in the shootout, one worked for the photographer better than the other. So, again, I ask: what did we say that was actually factually wrong or misleading?

-Rishi

 

[rishi_sanyal]

So in fact, it was you who misunderstood the way in which iTR functions. The more detailed explanation provided by Canon should clarify your erroneous, misguided comprehension. Perhaps you'll even consider updating the motocross article to correct the misapprehension that DPR is promulgating by suggesting that the 1D X II is 'confused', rather than performing as designed (as I previously pointed out). But...that may be too much to hope for in this case.

Honest question from a 60D owner whose camera does none of this—isn't acquisition supposed to be aided by iSA? In other words, isn't iSA + iTR supposed to handle both acquisition and tracking (a la Nikon's 3D Tracking)?

Yes, of course. How else would these newer cameras detect faces in viewfinder shooting? Some sort of sentience built into the PDAF module that can find faces off of just distance information? That'd be some cool magic...

 

[Don Haines]

So in fact, it was you who misunderstood the way in which iTR functions. The more detailed explanation provided by Canon should clarify your erroneous, misguided comprehension. Perhaps you'll even consider updating the motocross article to correct the misapprehension that DPR is promulgating by suggesting that the 1D X II is 'confused', rather than performing as designed (as I previously pointed out). But...that may be too much to hope for in this case.

Honest question from a 60D owner whose camera does none of this—isn't acquisition supposed to be aided by iSA? In other words, isn't iSA + iTR supposed to handle both acquisition and tracking (a la Nikon's 3D Tracking)?

Yes, of course. How else would these newer cameras detect faces in viewfinder shooting? Some sort of sentience built into the PDAF module that can find faces off of just distance information? That'd be some cool magic...

There is some interesting software out there in cameras..... I have a P/S camera with "dog" mode and "cat" mode.....when in dog mode it will not track the cat, but in cat mode, it will. Not only does this mean that you can track an animal, but the software is smart enough to know the difference between a dog and a cat!

As far as faces are concerned, I can associate a particular person with an image and it has the smarts to do facial recognition and tag that person in photos!

Of course, this is a mirrorless camera and is reading the image sensor, far more info than a metering sensor..... but this is the type of thing I expect to see when Canon makes the leap to a "real" mirrorless camera..... Imagine a mirrorless 7D3 (same ergonomics) that can track the bird as it flies through the branches of a tree!.......

 

[Don Haines]

Are there Zebras on the 5D4?

 

[ritholtz]

Hi Rishi,
You skipped these forums when D5 review published. You are very critical of 7D2 DR and explained thoroughly importance of DR and ISO invariance. You declared 7D2 DR situation is dire.
"Getting an optimal ETTR exposure is difficult and usually only best done via extensive bracketing. Given the difficulty of absolutely nailing an optimal exposure, the flexibility offered by a camera with greater dynamic range cannot be understated for situations such as these: they prove more tolerant of any 'mis-exposure' which, in fact, may not be a 'mis-exposure' at all when you're using the in-camera exposure indicators to judge your exposure with higher dynamic range scenes such as this one."

"Note that the dynamic range advantages of cameras with high base ISO dynamic range can extend to higher ISOs as well, where the 'ISO-invariance' of cameras like the D7000 allows you to purposefully underexpose the image by lowering the levels of ISO amplification. This gives you extra highlight headroom in accordance with the amount of reduction in ISO amplification. The 7D Mark II is not amenable to this type of shooting."

Then down played similar DR situation for D5:

"Either way, in our opinion, we'd try not to over-stress the importance of the fact that the D5 has poorer base ISO dynamic range than its current peers (after all, you can buy multiple D810s for the same price, if low ISO DR is important to you). For its intended audience, the D5's high ISO imaging capabilities, advanced autofocus and durability are likely to be much more important."

"In real world shooting, the D5's dynamic range hasn't been a huge problem. Sure, you'll need to watch your exposure in high contrast scenes more than you would on, say, a D750 or D810 - pushed low ISO Raw files out of the D5 exhibit a lot more shadow noise than those other cameras. But that's taken care of by either sacrificing some highlight detail during your exposure, or with a little bit of luminance noise reduction in post. And if you shoot JPEG and nail your exposure, or routinely in low light, there's even more to like."

Don't you think, you are kinda unintentionally playing to strengths of Nikon in your review by downplaying DR.
Barney justified this by saying, 7d2 is a general purpose camera and D5 is a specialized expensive tool which is fine for target users. When pointed out about how 7D2 is compared with 1dx in your review and your recommendation for Canon users to take a look at 7D2 instead of 1DX, he didn't respond.

 

[rishi_sanyal]

Are there Zebras on the 5D4?

No, unfortunately. Video support tools are limited. Actually, I use zebras for stills as well with mirrorless cameras, which then reminds me of how unfortunate it is that no one but Phase One links exposure warnings to Raw and instead bases them off of contrasty JPEG conversions. But that's another matter entirely.

A couple of people were asking about ISO and DR performance. Rather than commenting here, I'll just say that we should have some content for you on Monday.

And to your point - indeed, mirrorless cameras have an advantage with respect to subject tracking. But I feel some of them (like Sony) may have access to too much information, and therefore trip up/slow down. They should be able to get around this by pixel binning to lower resolutions, but I'm not convinced they all do so cleverly enough. Sony's 'Lock-on AF' for example is less reliable than Nikon's 3D tracking even with a 91,000-pixel RGB metering sensor, so a lot of it has to do with the efficiency of your algorithms. Subject tracking with Canon's DPAF, while a bit slow, is far more reliable - I'd pick it any day over Sony's object tracking.

But object tracking will continue to improve as processing power increases.

Cheers,
Rishi

 

[3kramd5]

The 5d mk2 I had at a constant 2 1/3 stop over, my 1dx is at +4/8 offset 0 ev.

The most interesting part of his thread is the use of 4/8. Outside music, I've never seen anyone say 4/8. Most of us say "half."

 

[StudentOfLight]

Are there Zebras on the 5D4?

No, unfortunately. Video support tools are limited. Actually, I use zebras for stills as well with mirrorless cameras, which then reminds me of how unfortunate it is that no one but Phase One links exposure warnings to Raw and instead bases them off of contrasty JPEG conversions. But that's another matter entirely.

A couple of people were asking about ISO and DR performance. Rather than commenting here, I'll just say that we should have some content for you on Monday.

And to your point - indeed, mirrorless cameras have an advantage with respect to subject tracking. But I feel some of them (like Sony) may have access to too much information, and therefore trip up/slow down. They should be able to get around this by pixel binning to lower resolutions, but I'm not convinced they all do so cleverly enough. Sony's 'Lock-on AF' for example is less reliable than Nikon's 3D tracking even with a 91,000-pixel RGB metering sensor, so a lot of it has to do with the efficiency of your algorithms. Subject tracking with Canon's DPAF, while a bit slow, is far more reliable - I'd pick it any day over Sony's object tracking.

But object tracking will continue to improve as processing power increases.

Cheers,
Rishi

Do you guys have the 16-35mm L III yet? I'm keen to see some rooftops

 

[Loswr]

The 5d mk2 I had at a constant 2 1/3 stop over, my 1dx is at +4/8 offset 0 ev.

The most interesting part of his thread is the use of 4/8. Outside music, I've never seen anyone say 4/8. Most of us say "half."

When Viggo referred to the 5DII, he was talking about leaving EC constantly set to +2.33 stops. His reference to the 1D X is not talking about EC, but rather autoexposure microadjustment (AEMA), which is essentially an exposure version of AFMA. The AEMA can be set up to ±1 stop in 1/8-stop increments, which is why Viggo phrased it as a +4/8 offset.

 

[Loswr]

So because of the order of the words, tracking must come before recognition? That is some convoluted logic predicated upon a lot of assumptions.

No, I stated that the order of the words was a clue. Do you understand what that is? No, I suppose not, since you apparently don't have one.

I then went on to quote from Rudy Winston's Canon DLC article describing in specific detail how the iTR system works, which states definitively that the AF system itself first selects the focus point, and then after that the metering data are used.

Would you care to address the statement by a Canon technical mouthpiece, which explicitly contradicts your viewpoint...or would you rather just continue to ignore the facts and change the subject?

So the PDAF sensor, which only knows subject distance, can recognize faces? How does it do that?

Honest question from a 60D owner whose camera does none of this—isn't acquisition supposed to be aided by iSA? In other words, isn't iSA + iTR supposed to handle both acquisition and tracking (a la Nikon's 3D Tracking)?

Yes, of course. How else would these newer cameras detect faces in viewfinder shooting? Some sort of sentience built into the PDAF module that can find faces off of just distance information? That'd be some cool magic...

It's not magic, and of course the PDAF sensor doesn't detect faces. But the metering sensor doesn't simply evaluate the entire region of the field which it covers, and identify faces out of the blue (or red, or green). First, an AF point is selected, either manually or automatically depending on settings (and for automatic selection, there may be more than one point selected). The automatic selection of the focus point occurs exactly as I've already made clear – it 'tends to focus on the nearest subject'. After the focus point is selected, the AF system tells the metering sensor where to look for faces. Here's a quote from a different Rudy Winston article describing that process, which you'll probably ignore like you ignored the first one...

Also, how does any of this impact upon my more important point: that in the shootout, one worked for the photographer better than the other. So, again, I ask: what did we say that was actually factually wrong or misleading?

Downplaying your factual errors just like you downplayed the importance of DR for the Nikon D5, 'eh? Unsurprising.

What is factually wrong is your repeated insistence that the data from the metering sensor are used for the initial automatic selection of an AF point, in spite of statements by Canon which explicitly show that you are incorrect.

What was misleading was the statement in the DPR motocross comparison that the 1D X II was 'easily confused' when in fact, it was the person holding the camera who was confused...about how the manual states the camera should behave and about how to properly configure the camera for iTR tracking in AI Servo (i.e., the steps listed above).

So, can we expect an Editor's Note to be added to that motocross comparison article, indicating that the camera was performing as designed, but the reviewer failed to understand that fact? Somehow, I doubt it.

 

[3kramd5]

The 5d mk2 I had at a constant 2 1/3 stop over, my 1dx is at +4/8 offset 0 ev.

The most interesting part of his thread is the use of 4/8. Outside music, I've never seen anyone say 4/8. Most of us say "half."

When Viggo referred to the 5DII, he was talking about leaving EC constantly set to +2.33 stops. His reference to the 1D X is not talking about EC, but rather autoexposure microadjustment (AEMA), which is essentially an exposure version of AFMA. The AEMA can be set up to ±1 stop in 1/8-stop increments, which is why Viggo phrased it as a +4/8 offset.

a'ha. Fair enough, I conflated the two statements.

Good to know, since I ordered a 1dx this morning and probably won't RTFM

 

[Loswr]

Good to know, since I ordered a 1dx this morning and probably won't RTFM

Touché.

Enjoy the 1D X - it's a great camera!

 

[Speedster]

Are there Zebras on the 5D4?

A couple of people were asking about ISO and DR performance. Rather than commenting here, I'll just say that we should have some content for you on Monday.

Thanks Rishi. I'll look forward to it.

 

[K]

DPR's conclusions were preordained. The 5DIV is improved, and it's a very good camera. Well, good for a Canon...but the D810 is better, and oh boy, just wait for the D820.

It's probably in their unofficial reviewers guide:

• For any given feature, Canon's performance is [insert choice from thesaurus search for 'good'], and Nikon's performance is [insert choice from thesaurus search for 'superlative'].
• In a Canon review, mention Nikon (or sometimes Sony) at every opportunity, in a Nikon review, mostly ignore Canon. Thus, Nikon features which Canon lacks should be discussed in reviews of both brands, whereas Canon features that Nikon lacks should be omitted from Nikon reviews.
• In the unlikely case that you find a Nikon feature not worthy of high praise, downplay the importance of that feature for the target market, in the typical case of presenting Canon deficits, stress that those features are critical for all users.
• In every Canon review, find one feature (preferably a minor one) to highlight, so Rishi can claim to be unbiased (used to be called the 'look, I have a ______ friend' approach).

A+

 

[K]

The point Neuroanatomist has made (the truth of which I can't currently judge), is that you're not "merely pointing out shortcomings," as if you had done so in a purely objective way, or as if it could be done in a purely objective way. All reviews and tests have unintentional bias, we get that. As I read the criticism, they are accusing you of two major errors: (1) failing to learn the Canon product well enough to use it as described in the manual; i.e. trying to apply Nikon-centric principles to the Canon product, instead of learning it for what it is; (2) choosing which features and qualities should be emphasized based on which brand wins that particular battle. You need to engage with these complaints professionally, not emotionally. When you lash out emotionally at critique, whether that critique is valid or not, it makes me wonder if your other work is also emotionally tainted or biased.

Exactly. Consider DPR's title choices...pre/reviews of the 1D X II: "Canon Catching Up?" and "Rock Solid"; DPR's review of the D5: "Setting New Standards". Canon is consistenly damned with faint praise.

Dynamic range and exposure latitude (the ability to underexpose by 5 stops then push in post) were critically important features for all photographers, and ISO invariance was a huge advantage...and in those areas, Canon fell far behind the competition. Then the D5 came out, and suddenly DR and extreme shadow lifting weren't really all that important for the target users. I mean...those things were important for the target users of the 7DII, which was not very good for DR and shadow pushing (well, it was good for a Canon camera, but the 'advantages of the aging Nikon D7000 were quite clear'). So according to DPR, it makes sense to criticize the 7DII for poor DR and exposure latitude while excusing the same things in the D5. Presumably because a fast-action high fps camera like the 7DII has a far different target user base than a fast-action high fps camera like the D5.

Also of note was that in the 7DII review of DR and exposure latitude, they added the Canon 5DIII "as a comparison against full-frame, which should have an advantage," (but of course the aging D7000 beat the 5DIII, too). However, in the Nikon D5 review for some reason they don't mention that for low ISO DR the APS-C sensor in the Canon 80D outperforms the FF sensor in the D5.

For the DR/shadow pushing section, DPR's D5 review states, "This content was originally published before we had access to Canon's EOS-1D X Mark II. We are currently working on an updated comparison." They've had access to a 1D X II for quite some time now – the full review of it was published over 2 months ago (and certainly they had it for some time before that), and their full review of the 1D X II states that it outperforms the D5 in those areas. But they haven't updated the D5 review to compare it to the 1D X II. I wonder why?

Well done.

Yet another world-class post that completely exposes the severe anti-Canon bias. This should be a prerequisite reading for anyone across the entire web that wants to engage in a Canon vs. Nikon debate when reading and shopping around for a camera.


It is sad that the level of bias and techniques of bias in photography reviews has reached a level comparable to political journalism.

The folks at DPReview could be making a whole lot more money working for main stream media....

 

[privatebydesign]

But you have added to the captions, why did you feel the need to do that if you were comfortable with the way you presented your piece previously.

No, I only added 'Canon 5D Mark III' to one rollover, an oversight due to the fact that that rollover was copied from another article on read noise where we purposefully left brands/cameras out of the discussion. After someone pointed it out here (in a very rude manner, no less), I agreed it best to clarify right in the caption (since I guess some weren't reading the text around it that clearly indicated the 5D IV's improvements addressed such issues).

No other captions were changed. I am very comfortable with the way the piece was originally presented, which is pretty much the same way it's presented now.

I complained about one caption.

....... All the images are from a 5D MkIII and the example of lack of DR, ............ isn't even clearly labeled, so it deliberately gives the false impression it is from the new camera.

You changed it.

You then said "No", but then, to paraphrase, "I only changed that one you complained about and said wasn't clearly labeled. But only because we made a mistake and it wasn't clearly labeled".

That is why people have such contempt and distrust for politicians and the media. Even when the answer is 'Yes I did change that one you said wasn't labeled' you still have too say 'No'.

Grow a pair, you made a mistake own it, you fuel the fire of distrust and lack of respect by not just saying, "Yes, we missed that one, sorry". The coincidental fact that it is the most misleading and provocative image for you to "unintentionally copy and paste" can only add to the quite lamentable feint praise with which you are so adept at.

 

[privatebydesign]

As for your issues with metering; composition and exposure are the two key aspects of any image. Stand two photographers next to each other and their images look different because of the composition and exposure. Anybody using any camera with selective exposure control should understand how and what a meter is doing and how it will represent a scene in front of them. And lets be honest, all digital cameras will meter practically the same in any situation.

Well, except for ones that won't spot meter off the chosen AF point.

Also, this was kind of my point: that you have to work around the fact that camera meters aren't perfect always for your creative intent, so technologies like EC and latitude both help you, with the latter particularly helping you when you don't have the time to chimp.

Metering off the chosen AF point does what exactly? It still meters for 12.7% grey, if your subject is a 1/2 stop below mid tone and you want them to look 1/2 stop below mid tone then you are good with no EC, still a mental juggling act. But how often is that? Besides, are you talking 1% spot where you can be quite sure you have the area covered by your AF point, or the more common 2.5% which will still, more often than not, contain some of that "unworkable and impossible to control" background?

The point is camera meters are "perfect" in that they will give you consistent information. As photographers we have to decide the relevance of that information and how we can apply it to get our anticipated result.

Exposure latitude is the lazy way. It is the 'safety net', the 'it isn't my fault', 'it should be able to do it for me' mentality that kills all kinds of craftsmanship. I am not being elitist, I am not saying I won't take more DR when I get it, I am not saying it can't save your butt on occasions, I am saying when technology dumbs craftsmanship down to such a level that core aspects of understanding of that craft can be completely ignored from an artistic and creative endeavor then that is not actually helping that artistic and creative endeavor or those doing it.

Few testers can adequately test 200mph super cars as they are meant to be driven, few testers, it seems, can adequately test modern upper range camera systems as they are meant to be used.