September 17, 2014, 11:38:35 AM

Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.


Messages - sarangiman

Pages: [1] 2 3 ... 19
1
Reviews / Re: Tony Northrup - D810 vs. 5D Mk3
« on: Today at 05:01:37 AM »
Maybe this video released by Sony will help get my point across:
https://www.youtube.com/watch?v=wy8TAGFC95o

Note how at 1:33 they stress how the AF system doesn't get 'distracted' - because it understands its subject from scene analysis of the imaging sensor. Same thing with the RGB metering sensors in the 1D X, and in Nikon cameras (albeit with much less resolution). And that has been my point all along.

Perhaps on-sensor PDAF and/or DPAF will eventually outperform what we see with DSLRs. I'm all for that. But for now, the best subject tracking is available in cameras that *have* a dedicated color sensor for subject tracking. Be it an image sensor, or a dedicated RGB metering sensor. Not a camera like the 5D Mark III that lacks both of these inputs when it comes to subject tracking.

2
EOS Bodies / Re: How can we improve on 5D3 to 5D4?
« on: Today at 03:07:03 AM »
...
Falk Lumo did some great work that showed AFMA inconsistencies across the AF sensor, on top of a skew that resulted in the left AF problem:



This is taken from his article here: http://www.falklumo.com/lumolabs/articles/D800Focus/SensorArray.html

That looks like lens manufacturing issues to me where the lens isn't properly centered. Look at the recent lens testing on lensrentals to see how warped the light coming through the lens is. It would appear that this warping of light does more than impact IQ, it also impacts AF.

That is unless the above AFMA AF mapping is the same for every lens on that camera ... seems more likely to be a lens manufacturing issue than camera one.

Has anyone done a similar test for a Canon lens/camera combination?

Dilbert - I believe what Falk Lumo was showing there was the AF sensor miscalibration fiasco Nikon underwent with the D800.

I do know Roger's (incredible, laudable, and informative) work - yes, the PDAF sensors are making measurements from non-central light rays, and have to then be corrected for the image-forming light which uses both central & non-central rays. This is lens-specific, and also, at least partly, where AFMA comes in.

My point was: manufacturing tolerances are not so tight that you can assume that the AF sensor surface is perfectly flat, or that it's perfectly aligned compared to the image sensor. Therefore, different focus points may need different AFMA values for optimal focus, and who knows - this might vary from lens to lens as well. We really just don't know exactly what AFMA is doing, or the math involved, although some suggest it's a correction to that correction value that corrects for the PDAF sensor's non-central light-ray measurements. Which itself will be dependent upon the degree of spherical aberration for any given lens.

I wonder if Reikan's developer has collated this sort of data from people using his software... would be incredibly interesting.

3
EOS Bodies / Re: How can we improve on 5D3 to 5D4?
« on: Today at 02:55:37 AM »
So... I'm just wondering how many here realize that the D810, D800, and D750 already have every single one of those features...
Perhaps just the few of us who've actually made the switch and learned what features our new toys, uhm, tools possess.
Frankly, many of those cool features I rarely have use for.  I merely moved to get much improved raw files for landscape work.
I do want to borrow a buddy's 150-600mm Tamron and try the D800's AF tracking for BiF shots.
I don't care if I don't have a high frame rate, just good focus.

Haha, well said.

Initially I'd wanted to switch for the RAW dynamic range, which is great for landscapes, as well as for exposure latitude in post-processing. But then I realized just how powerful all those other features are.

Particularly: subject tracking in 3D focus tracking mode. Which I never used on the 5D3 b/c it was far too unreliable, no matter what use-case I chose in the menu nor how I optimized any one of those use-cases.

But then there's also programmable auto ISO, which is so much better than constantly having to manually change the minimum shutter speed as I switch primes at a wedding (since the 1/focal-length rule is not always applicable).
 
And now b/c of the ease of EC in M mode with Auto ISO - and b/c of the incredible sensor performance - I rarely have to worry about blown highlights even in high ISO situations b/c I just dial in a massive amount of negative EC. In low light scenarios that'd require ISO 800 and above anyway, this usually means I'm not paying any shot noise cost by dialing in negative EC, since all it's doing is lowering the actual ISO amplification used (not changing the focal plane exposure). So, for example, instead of ISO 1600, EC -3 will use ISO 200. Since downstream read noise is low, I pay very little (albeit non-zero) noise cost compared to just using ISO 1600. Meanwhile, I gain 3 stops of highlight detail b/c of the lower ISO. I then selectively raise exposure in post, protecting highlights.

The funny thing is: if Canon were to properly implement EC in M mode with Auto ISO, I'd actually use it in an opposite manner to what I described above. In other words: to brighten the image (via ISO amplification) to near where I want the final image brightness to be, not to apply negative EC - since there *is* a noise cost to brightening in post-processing vs. raising the ISO in-camera. So one could make the argument that if any of these cameras needs EC in M mode with Auto ISO, it's Canon, not Nikon - with the latter you can just select your shutter speed and aperture, then dial in a relatively low ISO and then choose your exposure in post-processing via the exposure slider.

Anyway, I'm rambling now.

What Canon stuff do I miss?
I do miss cross-type points all over the frame, the wireless flash system, and some Canon glass, though. And sometimes I get the feeling that Canon's center AF point focused faster and more confidently in very low light than the D810 - which wouldn't be surprising. Although both systems are rated down to EV -2, the 5D Mark III is 'looking' for much more detail with its horizontal, vertical, and dual-diagonal sensors.

4
EOS Bodies / Re: How can we improve on 5D3 to 5D4?
« on: Today at 01:21:18 AM »
My list...designed to be the all-arounder:

 - Histogram based on RAW *!* (screw JPEG! :P)
 - Higher frame rate (8fps, using CP-ADC for low noise, high speed readout)
 - More dynamic range (high and low ISO...two stops low, as much as possible high)
 - More resolution (~50mp)
 - Layered sensor (drop the bayer! with binning capability, so I could bin 2x2, 3x3, maybe even 4x4 for very high ISO, as I'd rather have the SNR than anything :P...yes, this would mean 150 million photodiodes)
 - iTR metering
 - f/8 AF with center zone support (say 13 center af points usable at f/8)
 - AF-point linked meter
 - DPAF-automated AFMA (and, therefor, DPAF)
 - Dual CF (w/ CFast2 support)

"iTR metering"?  :o

iTR is the AF tracking mode that uses the metering sensor for subject recognition and tracking...

I must say, I'm quite surprised at the number of people in this thread asking for:

  • More DR at base ISO
  • More resolution
  • Spot metering linked to AF point
  • Programmable Auto ISO
  • EC in M mode (and while we're at it: in a less stupid implementation than the 1D X's)
  • Face detection & tracking outside of Live View
  • ... and at least one guy asking for better subject tracking across the frame after initially choosing a subject (well that makes 2 of us now, since this is one of my wishes as well)

So... I'm just wondering how many here realize that the D810, D800, and D750 already have every single one of those features...

Or perhaps that's the point - everyone here wants what Nikon already offers? Either way, it's kind of funny :)

I'm going to add one more thing to the list:

- AFMA for every single AF point, and then, yes, DPAF/CDAF-automated AFMA... b/c AFMA is already incredibly annoying as it is - imagine doing it for 65 points!

I want this b/c I'm finding more and more that you can't trust the factory calibration process to have calibrated every focus point perfectly.

Falk Lumo did some great work that showed AFMA inconsistencies across the AF sensor, on top of a skew that resulted in the left AF problem:



This is taken from his article here: http://www.falklumo.com/lumolabs/articles/D800Focus/SensorArray.html

5
EOS Bodies / Re: How can we improve on 5D3 to 5D4?
« on: Today at 12:53:47 AM »
I would also like to have autofocus track a moving subject in a frame without me having to use the joystick (Nikon can do this as well).

Wouldn't that just be enabling the mode that uses all 61 points?

+ servo, yes.

I think he means 'Intelligent Tracking and Recognition', or iTR, using the RGB metering sensor - which does this far more accurately and reliably, for certain shooting scenarios, than the 5D Mark III ever could. The metering sensor also enables 'face detection AF in viewfinder. Exposure/Ettl based on face alone' -- which PhotoCat wanted. And enables spot-metering linked to the AF point as well.

Canon DSLRs below the 1D X have been behind in all these regards compared to Nikon for many years now.

What'd be great about them putting this stuff into the 5D3 would be the marriage of iTR/'3D focus tracking' (what Nikon calls it) with Canon's stellar dual-cross-type technology, and extensive use of cross-type AF points.

6
Reviews / Re: Tony Northrup - D810 vs. 5D Mk3
« on: Today at 12:40:23 AM »
Maybe a D800 would have nailed every shot...I'm a little skeptical of that, given that I still hear pros complaining about Nikon AF, and have read many reviews comparing the 1D X and D4 that clearly indicated Canon AF tracking was superior (although, possibly not your particular close-subject/thin DOF use case.)

Whoa, let's not conflate the 1D X and the 5D3 in this discussion. iTR on the 1D X makes it significantly better than the 5D3 for the type of focus tracking I'm talking about. Furthermore, yes I doubt sports photography stresses the AF system in the manner I'm talking about (remember what I said about DOF of telephoto lenses with large subject distances, and the relative changes in distances to subjects being smaller than for fast wide-angle primes at close subject distances), so there it's possible that the 1D X's diagonal, wider baseline sensors in the center, as well as more cross-type AF points in general, wins.

Also, there was that whole D800 miscalibration fiasco that sunk the collective opinion of Nikon's AF. And I wonder how much of that persists across other cameras - the factories have to individually calibrate every AF point on the sensor! Who does that better? Who knows? I know there are at least a couple of AF points in my 61 point 5D3 sensor that are out of whack compared to the AF points right around it... indicating a miscalibrated point (and not a general skew). This systems are incredibly complex, so I'm not surprised there are a variety of opinions out there.

Also, someone said earlier how Canon's presence is much more significant at the Olympics. That has changed over the years though...



... and again, I'm not saying one system is better than the other for sports. I am not at all qualified to comment on that, as I haven't tested either system for sports. Actually, my feeling is that you don't strictly need subject recognition and tracking for sports photography, since a lot of the action is happening amongst subjects of similar distances from the camera. So, here, other things might matter more.

I'm only commenting on the type of tracking I've explained previously (and for the sake of my sanity, don't feel like repeating again here!).

7
Reviews / Re: Tony Northrup - D810 vs. 5D Mk3
« on: Today at 12:20:14 AM »
Now, you claim that Canon cameras cannot track in the "Z" axis (which I can only assume means within the depth of the scene, towards or away from the photographer), or track in Z while also tracking in X/Y (vertically and horizontally within the frame).

No no no I don't claim that at all. I would never say the 5D3 isn't good at tracking in the Z-axis; in fact, it's phenomenal at it. You got my statement priorities completely backwards in your statement. I'm talking about the 5D3's ability to automatically move the AF point in the X-Y planes to stick with your initial subject as you recompose or the subject moves quickly & erratically in the X, Y, and Z planes.

I'm just going to nibble your post down to this one part, as I think this point is the clincher. So, first, some clarification is needed. When you do this focus and recompose thing...are you using all of the AF points...or are you in a single-point selection mode? If it is the former...then Canon cameras can do this...maybe not as well, but they can.

If you are doing the latter...using the single center point say, focusing on a subject, then recomposing and the Nikon AF system is then able to maintain the lock using OTHER AF points...then yes. You are correct, Canon cameras DO NOT do that. That is how Sony's new AF system works. I didn't quite get it at first when I first saw some videos explaining what Sony's AF was doing...but once I experimented with the 5D III and couldn't get it to do the same thing, I realized what was "new" about Sony's AF. I think that is a kick-ass feature...and I thought it was only a Sony thing. It would be interesting to me if Nikon has had that for a long time...but I would still like confirmation that that is indeed what your talking about.

jrista: Now please go try it on a Nikon D810, D4s, or D750 or what have you, and you'll be blown away by how 'kick-ass' it is on the Nikon's even without a high-resolution image sensor to do what Sony's doing.

We've essentially reached an impasse where you're not going to understand me any further until you just go and try it.

Now, while you may find Sony's AF 'kick-ass' in this manner (and it really is, especially coming over from Canon), in practice, it's not actually as good as Nikon. The A77 II can't keep up as quickly and accurately as the D810 can. So if you were impressed with the Sony... you're in for a pleasant surprise when you try Nikon's 3D tracking.

If I were to rate this sort of '3D tracking' ability, it'd go something like this:

Nikon D810/D750/D4 > Sony A77 II > Canon 1D X >> Canon 5D Mark III

Since I've tested all of these systems, perhaps you can get some idea of why, in my mind, the 5D Mark III lags so far behind compared to the 3 systems above it in my comparison above that I cannot use any positive descriptor when talking about it.

And - much like you - I want the best of both worlds. So I either want Canon to catch up to Nikon in this regard (b/c it still loses subjects more easily than Nikon and/or Sonly A77 II), or for Nikon to adopt some of Canon's significant benefits: more cross-type AF points, wider baseline sensors, and diagonal sensors, to name a few. So, if the inclusion of the metering sensor* in the 7D II is at all indicative of Canon's intentions to start putting these metering sensors in all their cameras (as Nikon has done for quite some time now), then that's *incredibly* exciting to me. B/c I might actually put up with Canon's sub-par image sensor (in terms of Raw DR) if the 5D Mark III replacement had all cross-type points, those dual cross-type center points, EV -3 focusing, along with a 150,000-pixel RGB metering sensor + iTR. As long as their iTR algorithms caught up to Nikon's 3D focus tracking algorithms, anyway. B/c as much as I love Nikon's huge DR for landscapes, and programmable Auto ISO for more intelligent fast-paced shooting, and EC in M mode with Auto ISO, for that matter, it's focus that remains the largest reason for me having to throw away shots (although, I experienced a huge increase in keepers stepping up from the 5D2 to the 5D3; I just wished I'd gotten either the 1D X or the D800 instead back then... but one was too heavy and the other was plagued by AF sensor miscalibration issues and also didn't really fit in my hand with its poorly designed grip). And therefore I applaud Canon for their focus on focus, if you'll excuse my pun. The 7D Mark II's AF sensor, inclusion of iTR, and dual-pixel AF all indicate Canon cares about AF. And that is, simply put: awesome. Because it's the biggest differentiator between mirrorless ILCs and DSLRs today. 

I'm not saying all this matters for every type of photography. I'm just saying it can open up huge doors for certain types of photography. And can even idiot-proof a DSLR once you consider the impact of proper face detection & tracking with a traditional PDAF AF system (which, again, up until the release of the 7D II, only the 1D X had in the Canon line). Or, at the very least, make it easier for pros to focus on more important things, like lighting, composition, capturing the decisive moment, etc. Of course, I won't use face-detection AF for shallow DOF photography b/c you need to nail the eye -- and I'm not yet convinced that Nikon's 3D tracking and/or Canon's iTR are intelligent enough to focus on the eye of the face after detecting the face using the metering sensor.

*Though, I must say, I can't understand why they'd leave out spot-metering linked to the AF point, as it should be easy now that they've got that high-resolution metering sensor in there...

8
Reviews / Re: Tony Northrup - D810 vs. 5D Mk3
« on: Today at 12:07:40 AM »
Many thanks for taking your time with this, being in the process of choosing a couple of extra bodies at the moment, this cleared a couple of misconceptions i had, so thanks once again.

You're welcome, and thanks for taking the time to say this.

My 6D can do what the sample video show, only not so good when the subject too far from centre. I not sure what wrong with the 5D3.
In the 6D vs D600 video, 6D show it was bad when the subject move from far to near.

I didn't say the 5D3 can't do it, I said it can't do it anywhere near as well as a 1D X with iTR, or any modern Nikon with either a 2,016 or 91k-pixel RGB metering sensor (D7000/7100/5300/5200, D800/810, D4/s, D750, etc.).

And with the 6D's paltry 9 AF points... let's not even go there. The chances of there being an AF point where I actually want the focal point of my subject is vanishingly small.

Are you saying you can slap on a fast prime, focus dead center, and then rotate ("recompose" generally involves much more rotation than translation, obviously) the camera about its own axis so the initial subject as seen by the AF unit goes from one side of the frame to the other (or at least from the borders of the PDAF sensors) and significantly out of the DOF and it will track it?

If so, that's cool and I do not believe my 5D3 could do such.

LOL, thank you for writing this :) And, yes, that's exactly what I'm saying. Again, just watch this video:
https://www.youtube.com/watch?v=C5J7zALyHic

It's literally that good, and even better on the DSLRs with the 91k-pixel RGB sensor.

Importantly, it can still track that well as you're changing distance to the subject, which is exactly where the 5D3 will get confused. That is, the 5D3 will be much more prone to losing the initial subject once you convolute 2D X-Y movement across the frame with distance changes as well. Unless there's only one or two subjects vastly separated from the background moving forward or backward with not too erratic shifts within a small timeframe in the depth/distance axis.

And this should come as no surprise if you follow my thought experiment in my last post in the ** footnote. In fact, once you understand just how the 5D Mark III AF tracking works (by looking at subject distances across all AF points, but never doing any subject/scene analysis to actually understand what your subject is), you'll be able to predict when/where it'll be good, and where a dedicated color/image sensor will be much better.

9
Reviews / Re: Tony Northrup - D810 vs. 5D Mk3
« on: September 16, 2014, 11:51:01 PM »
I think your mesuabating to the point where your blaming gear for your shortcomings. Even if the Nikon is better at tracking, of what good is it if real photographers like jrista or the myriad of others here are already getting the results with the superb 61pt system in the situations you described. That just shows you don't know how to get the frames you want.

No, it means that a system without scene recognition for AF doesn't allow me to get the shots I want for some of my photography. Are you suggesting that 'real photographers like jrista or the myriad of others' represent the entire sample pool of photographers & potential types of photography in the entire world? It really takes some perverse logic to think you understand the needs of every photographer out there so well that you can say 'even if Nikon's AF tracking is better... it doesn't matter.'

Doesn't matter? So the entire focus problem has been completely solved in the industry? Everyone is able to get 100% hit-rate with any prime at f/1.4 under any circumstance?

And - measurebating? Really there's no winning with you. When we're talking about numbers and equations, we're measurebating. When we're talking about real world experience with fast primes and ability of the camera to keep up as a subject moves around erratically, we're still measurebating! I'm saying that a scene recognition system is so good at tracking a subject accurately in 3 dimensions that after you've used it, you just cannot say the 5D III is good at it. Yes it can do it, but it gets confused very easily. And if you just took a minute to think about how the 5DIII is doing it, vs. how the 1D X and 7D II and Nikons are doing it, you wouldn't be at all surprised either. Go back to my thought experiment I posed to jrista in one of my responses, and see if you can understand why an image sensor is much better at tracking than the algorithm the 5D III uses. For your convenience, I've posted it at the end of this post.**

Btw, here's the new Samsung NX1 doing it, with PDAF sensors all over the sensor:
https://www.youtube.com/watch?v=IMYhZ18tKk4

But I suppose that's useless, since the 5DIII is obviously good enough for jrista & a myriad others and therefore every photographer in the entire world.

By the way, did you know that most current Nikon DSLR can focus on faces outside of Live View* (using traditional PDAF and OVF viewing)? Down to the D5200, I believe. And any Sony SLT, of course? Meanwhile, up until the release of the 7D II, only the 1D X could. Because it was the only camera with a separate color sensor for scene analysis. If your face happened to be the closest subject, sure a 5D III would focus on it in 'auto AF area' mode. Have a flower, or any other subject in front of the face, and the 5DIII focuses on that instead. Switch to a 1D X with iTR, and it'll focus on the closest (or biggest, I'm not sure) face, and track it as well. Not a big deal for me, but great when I hand off the camera to a family member to take a photo.

So now with the introduction of the 7D Mark II, do you think the inclusion of iTR with the RGB metering sensor is just a bunch of marketing hype? A 'me too' feature? Or do you think perhaps Canon is including it now b/c its actually of some utility?

Because, at the crux of it, you're essentially arguing that iTR is completely useless. And I completely, radically disagree.

Now if you want to complain about something legitimate, then point out the lack of AF point metering because that really does suck at times but saying the 5D3s AF is bad at tracking means that you didn't RTFM.

I did point that out. Almost every Nikon camera, down to the D5200, has spot-metering linked to the AF point. Because that's yet another thing the RGB metering sensor enables.

Just b/c you don't think the metering sensor 'seeing the scene' and providing face-detection & subject tracking isn't useful, doesn't mean it's not, or that I didn't RTFM.

What's your point of repeating the phrase 'RTFM' other than to incite me? Do you think that phrase is conducive to intelligent discussion? 

and yes, the nikkor G primes still AF like a slug and that alone shifts AF speed to canons for weddings. (And I've shot a lot of them only with primes.)

Demonstrably false, and yet another blanket, unsubstantiatad statement. The Nikkor 24/1.4 keeps up no better than the 24/1.4 on a 5DIII in terms of speed of Z-axis tracking in my limited testing of them side by side. Perhaps a very controlled scientific study might demonstrate otherwise, but like a slug? I don't think so.

The Sigma 35mm f/1.4 is even better (and meanwhile retains full 3D focus tracking ability, since it reports distance information).

And why don't we talk about 85mm primes now, a workhorse for weddings? The Canon 85/1.2 is so slow to focus that I've missed many shots, finally opting to go with a 85/1.8, only to find its focus precision was quite poor. The Nikkor 85mm primes, OTOH, are great in terms of focus speed. Why would you leave that out of your 'Canon primes are faster for weddings' blanket statement?

Not to mention the 85mm f/1.2's enormous CA on the sides such that it doesn't sharpen up until somewhere between f/2.8 and f/4:

Here's the Canon 85/1.2 vs Nikkor primes wide open:


Here's the 85/1.2 at f/2.8, where it's still not as sharp/uniform as the Nikon 85 primes at f/2 and f/1.8, respectively.


If you want to talk about advantages of the 5D Mark III focus system, talk about its cross-type points all over the frame, and its wider baseline diagonal points in the center of the frame.

*But to be fair, it's doubtful the D5200-D7100, with their 2,016-pixel RGB sensors, do it anywhere near as well the more pro-level bodies with their 91,000-pixel sensors. The higher the resolution of the metering sensor, the better. Then again, I was surprised even the 2,016-pixel metering sensor in the D7000 could aid subject tracking well, but it does a reasonably good job. Certainly much better than my 5D Mark III.

**Here's an example of how the 5D Mark III focus system works. Let's say the center point of the AF system detects a subject 10ft away, then you recompose, then the camera notices a subject 10ft away is now over the left-most AF point, and meanwhile there's now nothing at 10ft away at the center point. Therefore, the camera decides your subject has moved (or you've recomposed such that the subject is now at) the left-most AF point. But what if your subject moved to 9ft away during this time as well? Well, with some clever algorithms you could analyze all the focus points and see if there was some progression of a subject like this (I've assigned letters to specific focus points for ease of discussion):

  • Subject in center point (C) 10ft away
  • Center point C no longer detects anything at 10ft, but the point just to its right (D) has a subject at 9.8ft
  • Point D no longer detects a subject at 9.8ft, but the point to its right (E) detects a subject at 9.5ft.
  • Point E no longer detects a subject at 9.5ft, but a point 6 points to the left (F) now detects a subject at 9.2ft.

... and so on and so forth.

Are you starting to see how incredibly complex this can get, and how prone to failure this might be if the subject is moving like this in 3-axes and/or the movement is convoluted with you recomposing? Or another subject entering the frame at a similar depth?

Are you starting to see how using an image sensor (Sony SLT, or all mirrorless ILCs really), or a color sensor with some finite resolution to recognize color patterns (enough to detect a face, which we know RGB sensors can do given their face-detection ability) that communicates with the PDAF sensor might have the potential to perform significantly better?

Incidentally, many types of bird photography are unlikely to stress this type of system much, since you typically have one subject at a very distinctly different depth from everything else (the background or sky). And when the bird moves, its typically going to move along the depth-axis with measurable acceleration or deceleration - which allows the predictive AF to work quite well (and I've already said Canon does this very well). Also, keep in mind the DOF for extreme telephoto lenses at large subject distances. For example, 300mm at f/4 for a subject 30m away has a DOF of 2.3m, giving the AF system more room for error compared to a subject 1m away shot with a 35/1.4 where the DOF is 6.4cm (and where it's very, very easy for the subject to fall out of that DOF or for the photographer to move more than 6cm).

10
Quote from: DP Review
The D750 inherits a version of the 51-point Multi-CAM 3500FX AF system that we've seen used in the D800/E and D810 but its 'II' version is actually more sensitive rated down to -3EV. We haven't been able to really get a feel for how much difference this makes in normal use but in theory, the D750 should offer superior AF reliability in poor light compared to the D810 and D4S, which is quite something (and which might prompt more than a few D800 owners to 'upgrade' to the D750).

As I've stated before, people make too much of this spec.  Consider an example of the difference between -2 EV and -3 EV light levels:  f/2.8, 1/15 s, ISO 51200 vs ISO 102400.  Neither of those is likely to result in a very good image in most situations. 

Whatever (likely hardware?) change that enables focus to work down to -3 EV might, just might, make the system also work better and more reliably at, say 1 or 2 EV. That would be advantageous.

I'm not saying that's the case, I'm just saying it's likely and that if it *is* true then that would be pretty worthwhile.

What'd be nice would be if someone tested this thoroughly. But before that, I think it's premature for anyone to make the comment you did - completely ignoring the potential benefits at slightly higher, yet still low, light levels. We all know very well that precision of AF points drops with lower light levels and lower contrast; anything that increases the SNR of what the AF sensor 'sees' could help performance in situations other than just -3EV.

Let us know if you think that's completely false.

All that said, I do wish Nikon put some more revolutionary changes into the AF system. Canon's been, admirably, iterating quite a bit on their AF sensors, if I understand correctly.

Interesting that this went unanswered. Neuro: if you don't have any solid evidence that the -3 EV rating *doesn't* help in other, less drastically low-light settings, then please don't go around spreading misinformation that its of no utility to shooting scenarios other than those extreme ISOs (51k and above) you mentioned.

I'm saying I don't know, as I haven't had a chance to test it yet, but common sense dictates that if the AF sensor can focus in lower light, then it's probably going to better in other, more reasonable, low light scenarios as well (like EV 2 or 3 or what have you).

This was certainly the case for the A7S, for example. Just b/c it was rated down to EV -4 didn't mean that's the only place it was useful. It'd focus far more quickly and reliably than the A7R in less low-light scenarios as well. And this isn't surprising - if you're not pixel-binning, then each pixel used for focus has significantly lower SNR. Thus, AF is more likely prone to failing in low light, especially with low contrast subjects. And if you're only pixel binning at the software level, you've got all the extra read noise of the extra A7R pixels (which, importantly, don't individually have lower read noise than the A7S' pixels).

Of course, the fact that the A7S' EV -4 rated AF is still only CDAF usually meant that a 5D Mark III, for example, would still outperform it when shooting actual moving subjects in low light (b/c PDAF only needs to make a few measurements to nail focus, whereas CDAF needs to continuously hunt - during which time your subject may move significantly, throwing CDAF completely off.). That said, Sony was correct in saying the A7S would focus in lower light levels than any competitor DSLR - but I found that was only true for static subjects. The sampling intervals get longer in lower light (so AF slows down) and if your subject is moving - good luck to any CDAF system. The subject usually ends up moving before CDAF is able to make enough measurements to complete focus.

11
Quote from: DP Review
The D750 inherits a version of the 51-point Multi-CAM 3500FX AF system that we've seen used in the D800/E and D810 but its 'II' version is actually more sensitive rated down to -3EV. We haven't been able to really get a feel for how much difference this makes in normal use but in theory, the D750 should offer superior AF reliability in poor light compared to the D810 and D4S, which is quite something (and which might prompt more than a few D800 owners to 'upgrade' to the D750).

As I've stated before, people make too much of this spec.  Consider an example of the difference between -2 EV and -3 EV light levels:  f/2.8, 1/15 s, ISO 51200 vs ISO 102400.  Neither of those is likely to result in a very good image in most situations. 

Whatever (likely hardware?) change that enables focus to work down to -3 EV might, just might, make the system also work better and more reliably at, say 1 or 2 EV. That would be advantageous.

I'm not saying that's the case, I'm just saying it's likely and that if it *is* true then that would be pretty worthwhile.

What'd be nice would be if someone tested this thoroughly. But before that, I think it's premature for anyone to make the comment you did - completely ignoring the potential benefits at slightly higher, yet still low, light levels. We all know very well that precision of AF points drops with lower light levels and lower contrast; anything that increases the SNR of what the AF sensor 'sees' could help performance in situations other than just -3EV.

Let us know if you think that's completely false.

All that said, I do wish Nikon put some more revolutionary changes into the AF system. Canon's been, admirably, iterating quite a bit on their AF sensors, if I understand correctly.

12
As a wedding photographer, I think this is a great backup camera, but we shoot with f/1.4 a lot, so the 15 cross type points and the 1/4000th shutter will take its toll.  The -3ev is great though, if only the 15 cross type weren't a limitation.  The mk iii and the 1dx are still the way to go and that's what our studio uses.  The D810 and the D4s and D700 are still Nikon's wedding workhorses. 

I hear you on the somewhat measly 15 cross-type points (by today's standards anyway, after the 5D Mark III and the 1D X), clustered near the center of the frame.

That said, I still prefer Nikon's AF system for my f/1.4 prime wedding photography simply b/c the 3D AF tracking system is so reliable that I can trust it to automatically move the AF point to stay on my subject as I recompose (or the subject moves) after I initially acquire focus using the center point. This is much faster than manually moving the AF point (which of course I could never do fast enough when the subject is moving).

Thankfully, the D750 has the same 91,000-pixel RGB metering sensor in the D810, so 3D AF tracking will likely be just as good (if not identical).

I hope Nikon is working on 'catching up' to Canon's more extensive cross-type offering.

I've also wondered for some time if the high-precision, wider-baseline center points on the 5D Mark III and 1D X outperform Nikon's center AF points. I haven't seen any rigorous tests done here (other than Roger Cicala's - but I believe his tests were done under good lighting with high contrast targets).

Personally, I think the D750 will be a workhorse wedding tool for the Nikon side. It actually offers a potential advantage over the D810 given that it focuses (theoretically) in lower light. And its only disadvantages compared to the D810 are lower resolution and higher base ISO (or slightly lower dynamic range) - which are probably of very little import to many wedding photographers. Meanwhile, the slightly increased FPS is advantageous to wedding photography.

They've definitely packed a lot into this camera, and that's admirable. I just hope we see some more 'revolutionary' improvements in technology from Nikon soon. I suppose that could be said about everyone, though!

13
Reviews / Re: Tony Northrup - D810 vs. 5D Mk3
« on: September 10, 2014, 05:38:01 AM »
OK. It's totally reasonable and fair to doubt someone claiming something you haven't seen, or that you feel is contrary to your experience and/or some sort of general consensus. That's fine, and I should expect it. Though, remember, I didn't say that Canon was poor at this, I said that Canons without a dedicated RGB sensor (everything but the 1D X) fare worse in this regard. The 1D X is far more capable at this sort of tracking than the 5D3 - in my hands/my tests - and I attribute this to iTR. Though not quite up to speed with the D4s/D810.

All I'd ask of some of the guys here is that you try & refrain from being absolutely certain that you're right, and that I'm totally wrong - until you've verified this yourself by shooting the cameras - in their proper modes properly calibrated - side-by-side to see what exactly these cameras are capable of when it comes to subject tracking in 3 dimensions. And yeah, I know, I can ask all I want... doesn't mean I'll get it. Certainly not here, from past experience.

In the meantime, I thought I'd search for videos demonstrating Nikon's 3D AF tracking vs. Canon's AI servo with automatic point selection, or iTR.

While you can find many examples of Nikon's 3D AF tracking on YouTube, I couldn't find anything similar for Canon, save for marketing videos (which only have simulated demonstrations of how auto AF point selection to track a subject should work).

I suppose, in a sense, this is somewhat indicative of how many people actually use automatic AF point selection for subject tracking on Canons. Like I said, I'll just to take my own videos/pics of comparisons to demonstrate the point, short of quoting Ken Rockwell (who actually happens to be right here re: 3D AF tracking's superiority to the analogues in Canon's 7D and 5D3).

For now, in case this adds anything to the conversation, I'll post a couple links demonstrating Nikon's 3D AF tracking. These tests below don't quite stress the AF system as much as some of the things I've been trying, since these examples don't have the subject changing distance very much. But, this is all I could really find in my brief search. Since you all have used the similar mode on Canons, perhaps from memory you can get a rough idea of how this stacks up against what you've seen on your 5D3, 7D, or what have you.

Here's the Nikon D5200's 3D AF tracking... you know, the Rebel competitor. It uses a 2,016-pixel RGB sensor:
https://www.youtube.com/watch?v=C5J7zALyHic

And here's the D4s' 3D tracking, which uses its 91,000-pixel RGB sensor: Not too much lateral (X-Y plane) movement here, but still gives you an idea of how well the AF system sticks to the initial subject. The 1D X can approach this in my experience, but my 5D3 can't... it often gets confused and the selected AF point(s) hop all over the place. And, yes, I've tried increasing the 'Tracking sensitivity', as well as the 'AF pt auto switching', settings.
https://www.youtube.com/watch?v=daL-4kTaLuU

Is that the sort of AF tracking and AF point moving to stay on the subject you typically experience with your 5D3 and 7D?

Honest question - I'm curious.

P.S. The following video is probably irrelevant b/c no one claims the 6D AF system to very good, but I thought I'd post it anyway. This guy in his 6D vs D600 comparison alludes to the 6D not being able to keep up and track as well across the frame, though things are of course made worse by the severe lack of focus points. And it looks like not too much of a stress test b/c of the DOF. But the D600 clearly outperforms the 6D which, of course, isn't surprising given its 2,016-pixel RGB sensor:
 
http://youtu.be/Dg_6jSaXGgY?t=4m20s

He says: "[The D600] certain seems to track better... the little dot was traveling with her and staying really close on her... I found it was choosing the skin tone."

14
Reviews / Re: Tony Northrup - D810 vs. 5D Mk3
« on: September 10, 2014, 01:47:06 AM »
The ONLY way you can compare these cameras is on price. That's the biggest lowest common denominator between them. Different tools for different work.  Action and Low Light: 5D MKIII. Detail: Nikon. But it's not always so clear. Which would you use to get lots of detail of a bird in flight? At that point, it becomes a preference...

I think Tony Northrup and his wife Chelsea are great. He is one of the few highly-marketed photographers I pay any attention to anymore. Several reasons:

  • Great Gear Reviews. He doesn't pull punches when reviewing gear. But I also think he's fair.
  • Myth Buster. See his video on FF vs APS-C and what it means for focal length and aperture? I've never seen anyone do a video that shows you what the differences is between the two sensor sizes. This is just one example. Why has it been so difficult for anyone else to show the same thing?

I have great respect for Tony and Chelsea. I don't think they are full-time photographers. I believe he has something to do with computer certification programs. He mentioned something briefly about that. I know Chelsea gives private music lessons.  And they still make the time to get enough content in for an hour long, quality youtube video every week. Try doing that yourself and see how much time that takes....

And that's another thing about them. They don't oversell themselves or make themselves look like a bigger deal than what they are. They aren't above doing portraits of employees at their local auto dealership. They tell you they don't make that much money on some of their products. They don't follow the common motto: fake it until you make it. They just work hard. Like what these http://ricardogomezphotography.com/successful-photographers-on-making-it-part-ii/ fine photographers have said it takes...  Sorry. It's a link to my blog. But it's just easier to find the video because I posted it prominently. Twice  ;D  This video has gotten me through some difficult times. The interviewees are simply honest. Love it.

And no. I don't know Tony or Chelsea or have ever corresponded with them. Just good peeps working hard to realize their dreams. With integrity. And Chelsea just cracks me up   ;D

BTW, Tony and Chelsea aren't switching to Nikon after all. Tony released a video...  https://www.youtube.com/watch?v=Jva08HY6uLE Thanks for the video Tony.

He has some good points. One thing that's misleading is his claim that you can compensate for smaller sensors by using brighter lenses. That's only true up to a certain point, b/c pixels have finite well-capacities. So while this will work in low-light (or light-limited) situations, it won't work for, say, landscapes or scenes with appreciable dynamic range where increasing the exposure indiscriminately will simply blow your pixels.

Here, larger sensors will still fare much better (all else held equal).

These are complicated topics so I'm not blaming him. But it's important to understand some of the intricacies/details.

Btw, here's an in-depth look at equivalence as it relates to focal length, aperture, DOF, and noise/ISO:
http://www.dpreview.com/articles/2666934640/what-is-equivalence-and-why-should-i-care

The take-home is that 'equivalent aperture' is a nice way to get an idea of total light gathering ability of the lens, resulting DOF, and diffraction.

15
Reviews / Re: Tony Northrup - D810 vs. 5D Mk3
« on: September 10, 2014, 01:37:23 AM »
+1 He really should RTFM because I tried a d800 for a weekend and it's AF sucked like a vacuum cleaner compared to the 5D3. I don't know about the 810 but it still uses that multicam backbone that was in the d700, which I also used before. Also I remember the G 1.4 primes being a slug compared to my 24L II 50L and 135L.

He probably didn't RTFM and I haven't seen any photos from him either.

Thank you for that very mature and descriptive dialogue; it really added to the conversation. And by that I mean it continues to spread the very misinformation that misleads people into buying into, and then themselves propagating, erroneous myths.

What about the AF? I'm talking about subject tracking in 3D AF vs. Canon's '61-point auto point selection' in AI servo mode.

  • What AF mode were you in?
  • Did the D800 you tested have the completely miscalibrated AF sensor (with the left AF problem?)
  • Were your lenses microadjusted? You don't shoot primes at f/1.4 without 1st checking that they're calibrated to your body properly.

And the D700? The improvements made to that module going to the D800 helped low-light focus significantly, and the dedicated RGB sensor increased 91-fold in resolution going from D700's 1005-pixel sensor to 91,000 pixels.

You think maybe that 100x increase in resolution might, just might, have helped the D800 at subject recognition and tracking?

#facepalm

Pages: [1] 2 3 ... 19