Tony Northrup - D810 vs. 5D Mk3

[dtaylor]

I did see a lot of pictures, taken on the same moment from the same scene, but if we mix them, we can't tell from who the picture was (5d3, D800, D810), unless.... we dive into the pixels, and then I agree, but that's not the way we look at a photo?

Some people can't see the forest for the trees. A few people can't even see the trees because they prefer to look at pixels instead of pictures.

Some people can't see the trees because they were shot at midnight on a moonless night and haven't been pushed 5 stops yet ;D

 

[dtaylor]

The point of having better data to start with is to reduce the amount of effort required to produce better data in the end.

Despite having more pixels, the D810 did not "capture better data" in the IR studio scene. So I'm curious as to how it would save any time, especially when it takes your computer longer to process each step with 36 MP.

I'm not against higher resolution when it yields a benefit. But jumping from 22/24 to 36 MP with today's sensor technology seems to yield no real benefit.

 

[dtaylor]

Your still misunderstanding. You could make the 5D III image as sharp, by sharpening. My point is you wouldn't have to do the extra step of sharpening the 5D III image to achieve the same results.

I would mess with sharpening on ANY photo. The two Nikon guys I know pretty well...both of whom have a ton of Nikon pro equipment...sharpen every one of their photos. The D810 doesn't eliminate that step. What step does it eliminate exactly?

Also, with a more critical eye, I do think that some people could pick out differences between two photos

I would like to see that pair of prints.

There are certain things about D800 milky way photos that I particularly like...a richness of black background sky levels, that are not just pure flat black but still nuanced with detail, as the sky really is, that I don't see in milky way photos taken with a 5D III. There is also a crisper look to stars in milky way photos taken with D800/D810 that just isn't quite there with the 5D III.

Can you point to a pair shot under similar conditions that illustrate this?

 

[scarbo]

Clearly, Sony have the edge when it comes to sensor technology, but that does not mean Canon produce bad sensors. They are just not the current leaders in sensor technology when it comes to image quality. Can they really be blamed for not producing the absolute best sensor, when everyone else comes second to Sony too?

Watching the link below, it seems that Sony have hit upon some innovations which have made them the world's leading camera sensor manufacturer and, by the looks of it, it will be some time before their competitors discover equally efficient innovations. That's just the way the cookie crumbles sometimes. Happenstance would have it that one engineer or a group of engineers who discovered these innovations were employed by Sony and not Canon.

https://www.youtube.com/watch?v=AXp0u1XsrZw

All it takes is one instance of thinking outside the box and securing that new idea against immitation to change the fortunes of a company for many years to come. It seems that Sony have experienced something like this with camera sensors. Of course other manufacturers may discover better or equally good innovations, but that's easier said than done and becomes progressively more difficult as technology matures. By flipping the switch like this, Sony look odds on to remain ahead of the game despite general advances in technology. Their innovations will simply benefit from more general advances. It will take a new, exclusive and more efficient innovation from another manufacturer to displace them.

It's impossible to predict the future, but in the absense of similar yielding innovations from others, a larger part of Sony's business may morph into a sensor manufacturing business for many of the other camera companies, in the way Intel produce processors for computer companies. Sony already make sensors for Nikon, Hasselblad and Apple. Canon will also have the option to become one of their clients, which might be the right move if their sensor technology is not forthcoming.

In any case, I think it's unfair to compare the 5D III with the D810 given the latter was released 2 years later. Yes, the D800 and D800E were strong offerings too (and so was/is the 5D III), but back then both companies were entering the market blind to what the other would be doing (assuming there were no spies in either camp). Now that they know more about their main competitor's direction, I'm sure their decisions will be more informed going forward. We should see what this yields before we predict doom and gloom. It's only been 2 years.

 

[sarangiman]

I find this resistance to improved technology incredibly strange...to the point where I simply don't believe it.

But that's actually fairly normal behaviour, culture changes in generational steps. In many areas of society you literally have to wait for the "old guard" to die off before new ideas can be taken seriously.[/b]

Can I ask...in the context of this discussion, how does going from ~12 to ~14 stops of DR, or going from 22 to 36 MP, represent 'new ideas' requiring the 'old guard' to die off before they're adopted? If you're talking about the switch from film to digital, or from vinyl to CDs, that's fine...but those are paradigm shifts in technology. To suggest that the differences between current Canon and SoNikon sensors are a paradigm shift is ludicrous. Rather, those differences are minor, incremental improvements. Real improvements, yes...but minor.

You and I both know the difference is actually 3 stops: 11EV vs 14EV. Add FPN to the Canon and the practical DR is even less.

Why are you down-playing the difference? It actually approaches an order of magnitude, if not more.

 

[sarangiman]

I too use Lee's GNDs. I have a bunch of them, in both soft and hard grad. I really love the Lee filter system (although it failed me recently...i had my 2-stop proglass ND in...and when I was photographing rivers it just slipped out and shattered on the rocks...I am not sure why it did, but it was like a $160 filter. ). The thing that set me off not long ago was a bunch of scenes where the skies ended up totally blown out when I exposed to preserve some detail in the shadows...and the skies were patchy...not along a nice horizon where a GND filter could actually be used to fix the problem.

I've bracketed crazy-wide like that as well, but in my experience, at least when you have bright highlights (like the highlights in water, or bright skies backing a dark foreground, or the sun in the frame), you end up with posterization or haloing if your exposures differ by a stop or more. Getting the exposure differential down to 2/3rds of a stop seems to smooth out the highlight transitions, so you don't end up with posteriation or funky CA or color issues or things like that after merging to HDR.

Yup, I know all too well what you're talking about. It's often very, very difficult to merge vastly different exposures - there are problems at the boundaries of very bright to very dark regions.

I also pretty much refuse to use most HDR merging software, and do most of it by hand using luminosity masks and an Intuos and other tricks.

Actually a much nicer way to do this is to expose for the highlights, let the shadows get buried, then just shoot a bunch of images and average them (using something like PhotoAcute). Then process a single file. This way your shadows clean up from the averaging, and you don't have the artifacts HDR software introduces.

And of course it goes without saying that this technique is much easier with a D800/810 than a Canon DSLR.

 

[Loswr]

I find this resistance to improved technology incredibly strange...to the point where I simply don't believe it.

But that's actually fairly normal behaviour, culture changes in generational steps. In many areas of society you literally have to wait for the "old guard" to die off before new ideas can be taken seriously.[/b]

Can I ask...in the context of this discussion, how does going from ~12 to ~14 stops of DR, or going from 22 to 36 MP, represent 'new ideas' requiring the 'old guard' to die off before they're adopted? If you're talking about the switch from film to digital, or from vinyl to CDs, that's fine...but those are paradigm shifts in technology. To suggest that the differences between current Canon and SoNikon sensors are a paradigm shift is ludicrous. Rather, those differences are minor, incremental improvements. Real improvements, yes...but minor.

You and I both know the difference is actually 3 stops: 11EV vs 14EV. Add FPN to the Canon and the practical DR is even less.

Why are you down-playing the difference? It actually approaches an order of magnitude, if not more.

Call it 4 stops if you like. It doesn't change the main point that we're discussing incremental improvements, not a paradigm shift. Of course, we should also stipulate that were talking about DR at base ISO.

 

[jrista]

I find this resistance to improved technology incredibly strange...to the point where I simply don't believe it.

But that's actually fairly normal behaviour, culture changes in generational steps. In many areas of society you literally have to wait for the "old guard" to die off before new ideas can be taken seriously.[/b]

Can I ask...in the context of this discussion, how does going from ~12 to ~14 stops of DR, or going from 22 to 36 MP, represent 'new ideas' requiring the 'old guard' to die off before they're adopted? If you're talking about the switch from film to digital, or from vinyl to CDs, that's fine...but those are paradigm shifts in technology. To suggest that the differences between current Canon and SoNikon sensors are a paradigm shift is ludicrous. Rather, those differences are minor, incremental improvements. Real improvements, yes...but minor.

You and I both know the difference is actually 3 stops: 11EV vs 14EV. Add FPN to the Canon and the practical DR is even less.

Why are you down-playing the difference? It actually approaches an order of magnitude, if not more.

Technically speaking, since stops are base two, it's really more than one order of magnitude. But, I get what your saying.

Just going by the difference in DR by decibels, the difference is 16.8dB (assuming the maximum DR is 13.8...as I don't really like DXO's PrintDR. If we go by Print DR, the difference is more like 19.2dB). It's 82.8dB vs. 66dB. At 16.8dB, the difference is more than an order of magnitude, since every 10dB change IS an order of magnitude, and closer to two orders of magnitude. At 19.2dB, the difference is two orders of magnitude. THAT is very truly significant.

It's also a concept no one seems to understand. And I'm tired of trying to explain it...

 

[dtaylor]

I could provide evidence. In the past, I have provided evidence.

You provided a long list of links to HDR photographs that you claimed were not HDR photographs. And a sunflower shot with no Exmor version for comparison. The Coke box wasn't yours and while it showed a difference...which everyone acknowledged before the Coke box...it didn't show nearly the difference you claim.

So, until such time as I have a chance to do that in an effective manner, I will be ignoring all requests from you.

Will you be providing evidence for both the HDR without bracketing claim and the new Milky Way claim?

It was very satisfying to watch you back peddle and re-frame after seeing the IR samples btw

 

[Loswr]

Just going by the difference in DR by decibels, the difference is 16.8dB (assuming the maximum DR is 13.8...as I don't really like DXO's PrintDR. If we go by Print DR, the difference is more like 19.2dB). It's 82.8dB vs. 66dB. At 16.8dB, the difference is more than an order of magnitude, since every 10dB change IS an order of magnitude, and closer to two orders of magnitude. At 19.2dB, the difference is two orders of magnitude. THAT is very truly significant.

So...are you saying that the difference is a paradigm shift?

 

[dtaylor]

You and I both know the difference is actually 3 stops: 11EV vs 14EV. Add FPN to the Canon and the practical DR is even less.

Do you know what a 1 stop difference looks like? Because Fred Miranda's test didn't show even 1 stop difference. Tonal range captured was the same, but with more noise on A vs. B.

Even DxO doesn't claim <11 stops for the 5D mark III

 

[raptor3x]

I find this resistance to improved technology incredibly strange...to the point where I simply don't believe it.

But that's actually fairly normal behaviour, culture changes in generational steps. In many areas of society you literally have to wait for the "old guard" to die off before new ideas can be taken seriously.[/b]

Can I ask...in the context of this discussion, how does going from ~12 to ~14 stops of DR, or going from 22 to 36 MP, represent 'new ideas' requiring the 'old guard' to die off before they're adopted? If you're talking about the switch from film to digital, or from vinyl to CDs, that's fine...but those are paradigm shifts in technology. To suggest that the differences between current Canon and SoNikon sensors are a paradigm shift is ludicrous. Rather, those differences are minor, incremental improvements. Real improvements, yes...but minor.

You and I both know the difference is actually 3 stops: 11EV vs 14EV. Add FPN to the Canon and the practical DR is even less.

Why are you down-playing the difference? It actually approaches an order of magnitude, if not more.

Who knew that 12.1 actually rounds down to 11? We learn something every day.

 

[sarangiman]

Quit comparing apples and oranges.

According to DxO, the 5D3 is 10.97EV at the pixel-level. 13.67EV for the D810 at the pixel-level. So 2.7 EV at the pixel level; except that's not a fair comparison with 64% as many pixels on the D810 that can help DR when they're averaged during downsampling.

Normalized difference is 11.7 vs 14.8 for the D810 - that's 3 stops, and possibly more depending on how you account for FPN.

Also, you can maintain near base ISO levels of DR even at higher ISOs with Nikon/Sony cameras if you know how to use your camera right. So as far as I'm concerned, it's far more DR even at higher ISOs.

 

[raptor3x]

If you're going to complain about apples to oranges, at least compare Canon's best against Nikon's best. At pixel level, the D810 has a 1.9EV advantage over the 6D and a 2.2 stop advantage when both are downsampled to 8MP. The 6D has a 1 stop advantage at ISOs over 800 at a pixel level and 2/3 stop advantage when downsampled to 8MP.

 

[Loswr]

If you're going to complain about apples to oranges, at least compare Canon's best against Nikon's best. At pixel level, the D810 has a 1.9EV advantage over the 6D and a 2.2 stop advantage when both are downsampled to 8MP. The 6D has a 1 stop advantage at ISOs over 800 at a pixel level and 2/3 stop advantage when downsampled to 8MP.

Are you suggesting that the 6D performs better than the 5DIII in those metrics? Because that would imply that Canon has improved the performance of their sensors...and we all know that's not true.

 

[Loswr]

I'm saying it's near the beginning of one.

My real point is I think Canon is missing the paradigm shift that is currently happening, and will continue to happen over the next few years.

Time will tell. But consider some of the sensor technology that Canon has patented, in terms of huge sensors (largest CMOS sensor ever), high pixel density sensors (120 MP APS-H), low light imaging (0.01 lux fireflies), etc. If you're right and there is a paradigm shift underway, it's very likely that Canon is ready, and will jump when it makes financial sense. Given the documented recent sales data and the predictions of the near-term market share from Thom Hogan, it looks like it doesn't make fiscal sense yet.

 

[sarangiman]

If you're going to complain about apples to oranges, at least compare Canon's best against Nikon's best. At pixel level, the D810 has a 1.9EV advantage over the 6D and a 2.2 stop advantage when both are downsampled to 8MP. The 6D has a 1 stop advantage at ISOs over 800 at a pixel level and 2/3 stop advantage when downsampled to 8MP.

OK, that's fair.

The D810 still has higher DR at higher ISOs if you use your camera in a smarter way though, so those DxO higher ISO DR values are meaningless to me.

 

[arcanej]

What is the time between design to final product for a CMOS sensor? Or, to ask it a different way, how long after AMD's purchase of Radeon did the first fully AMD designed Radeon graphics card reach the market?

Canon has had patents for designs that could compete with the Sony-Exmor and Samsung-BSI sensors. A product incorporating these patents hasn't released it yet. If the D800 was a shot across the bow for Canon, should we begin to see that next year? That would be three years after Canon saw what Exmor could do. Is that approximately the lead time from the start of a chip design to a product reaching the market?

Taking a step back, what Canon had been doing seemed like a great business plan. It was - and still is - the dominant camera manufacturer. It's nearest peer was trailing in sales and didn't have the capital to sink into a revolutionary, new chip, sink a few hundred million into a new fab and come out with a revolutionary new design.

I'm sure Canon did the math: what would be the incremental revenue that could be gained from a much higher DR chip design and what were the costs associated with bringing that chip to market. If the up front costs were higher than the discounted future cash flows, the best business decision would be to hold off.

 

[raptor3x]

If you're going to complain about apples to oranges, at least compare Canon's best against Nikon's best. At pixel level, the D810 has a 1.9EV advantage over the 6D and a 2.2 stop advantage when both are downsampled to 8MP. The 6D has a 1 stop advantage at ISOs over 800 at a pixel level and 2/3 stop advantage when downsampled to 8MP.

OK, that's fair.

The D810 still has higher DR at higher ISOs if you use your camera in a smarter way though, so those DxO higher ISO DR values are meaningless to me.

Can you explain this further? I shoot an A7 as well as Canon and in my experience the Canon trounces the A7 at high ISO.

 

[raptor3x]

I also wouldn't say Canon "trounces" anything at high ISO. High ISO is primarily limited by physics. Canon has a marginal benefit at high ISO vs. cameras that are now becoming "last" generation, like the D800. The Nikon D810 closes the gap with the 5D III a bit more, and starts to encroach on 1D X territory at really high ISO (I actually think the D810 offers more DR at ultra high ISO than the 5D III...on a normalized basis, the D810 gets 7.7 stops of DR at ISO 51200 to the 5D III's 5.7 stops).

The A7 has quite a nasty quality to the noise at high ISO. I've had both an A7 and A7R as well as the 5D3 and while from a SNR ratio aspect I'd suspect there's not a huge difference, the A7 has a very poor quality to the noise that doesn't clean up with NR very well at all. In practice I'd put the 5D3 ~1 2/3 stops ahead of the A7 and the A7R ~1- 1 1/3 stops ahead.

The only thing out there right now that is really "trouncing" any other camera at high ISO is the A7s. It actually enjoys a two-stop advantage over the 1D X at ISO 51200, bringing nearly 9 stops (8.8 to be exact) of DR at that level. At lower ISO's it actually normalizes out a bit with the 1D X...the A7s' true advantage is at the ultra high ISO settings, and it does a remarkably good job.

The A7S is certainly a very intersting camera. I was really hoping that the D750 was going to be a D810 with the A7S sensor inside. If they had done that I probably would have picked one up along with a 70-200VRII. As for the D810 at 51200, I'm not sure which numbers you're looking at but I see it about equal to the 5D3 and ~2/3rd stop behind the 6D in dynamic range for the normalized print size on DXO.