D800 vs 5D Mk3 lab test

[DavidTrap]

Found this on Nikon Rumors.

Link (http://nikonrumors.com/2012/04/19/dxomark-verdict-nikon-d800-95-canon-5d-mark-iii-81.aspx/#more-38344)

I'll await more real world pics about high ISO performance, but I will admit I was impressed with H's cat photo at 12800 with his new 5D Mk3.

The really impressive result from my point of view, is the extra 2 stops of dynamic range at 100ISO. Less need for HDR!!!

DT

[StarLane]

Yep, Nikon is way ahead of Canon with their new gen cameras on the stills front, and catching up fast with video...not sure 36mp is really needed though. I'm saving to upgrade from my D700.

[gregbradley]

D800 DXO score of 95 for the sensor is the highest DXO have ever recorded. 5D3 score of 81 versus 5D2 of 79 tells a story of lack of Canon development. 1/3 stop improvement in RAW over 5D2 is not very impressive and still less than score for 1D iii and iv that are several years old. Canon guys are not happy.

D800 matches 5D3 in RAW at high ISO if not slightly better. In JPEG 5D3 goes to 25600 whereas D800 goes to more like 12800. So 5D3 has edge there albeit with some aggressive noise reduction then heavy sharpening afterwards.

I saw some Aussie guy post zoo photos with a D800 and a 500mm lens and ISO up to 4000. They are superb. Images from D800 are consistently sensational (Fuji XPro 1 also, it has no AA filter).

Both are great cameras. D800 is all about that sensational hi res low noise sensor. Sony really has advanced sensor tech with high mp yet great low light ability once thought impossible. Other aspects of the 2 cameras are too similar.

In the end most will stick with whose glass they have the most of unless they really want that 36mp super sensor (a lot do).

The 2 hot cameras at the moment seem to me to be D800/800E and Fuji XPro 1. Sony Nex 7 seems very good also. XPro 1 though is being touted as better than Leica M9. Image samples from XPro 1 are just fabulous.

A D800E and an XPro 1 for lightweight travel camera would be a good combo.

Greg.

[Octane]

The only Canon guys that aren't happy, are the gearheads over on DPR. They are old men with lots of money and no talent. Never mind the 5D Mark III not being good enough for them, their talents wouldn't even outdo a 350D. They are armchair sensor designers. Morons. I read the forum every day just for the arguments. It's a riot.

I am absolutely ecstatic with my camera. Absolutely floored at what it can do. I'm really looking forward to buggering back off to New Zealand in July to really give it a workout.

DxO Mark never was and never will be the final authority on /anything/.

There were two cameras that DxO Mark "tested" (one a Nikon, I can't remember which, and, no idea what the other was) that had the exact same sensors in them. One was given a higher ranking than the other. Of course, it was the Nikon that got the better rating.

They gave the 70-200mm f/2.8L IS II USM a lower ranking than the 70-200mm f/2.8L IS USM. You only have to look at the MTF charts to see that the II blows the original out of the park. Furthermore, they scored a Sony NEX camera higher than a Phase One MFDB. Right.

As the famous LL article once mentioned -- use your eyes, not the numbers. David saw the high ISO performance prevalent in my simple obligatory high ISO snapshot of my cat. Going by the DxO Mark scores, you'd think that the camera wasn't capable of doing anything because it didn't score a 96.

I'm still not sure why the 5D Mark III is being compared to the D800. They are completely different cameras. The /only/ thing that ties them together is that they're in a similar price bracket. Compare a 1D X and a D4, and, that's more than a fair comparison. But, comparing a 36 megapixel monster to a 22 megapixel camera doesn't work. Everyone knows that the higher megapixel camera is going to have a greater dynamic range and awesome noise characteristics. It's just the way sensors work -- just like a full frame camera having more depth-of-field than a cropped sensor.

Paul, can you define "way ahead"?

H

[gregbradley]

Comparing gear is always a bit of sensitive issue as it easily can come across as being critical of someones new toy or raining on their parade.
Thats certainly not the intention here, just a friendly discussion about technical merits of sensors and cameras.

I think D800 and 5D3 are in fact in the same market segment and what most would compare when choosing a high end full frame camera. Both Nikon and Canon would regard them that way, so do the bulk of their customers. Someone thinking about buying a 5D3 would naturally consider a D800. They have similar capabilities and accessories. Similar autofocus, similar metering, similar weather sealing, similar LCD display, controls, 2 memory card slots, same sized sensor, similar price bracket, similar frames per second, similar low light performance. So when you step back from it there isn't a lot of difference between them. A few more fps here, more mp there, great low light performance from both, nice HDR feature on 5D3, nice time lapse feature on D800 etc etc.

The big difference is the sensor hence the importance of the DXo mark.

For weddings I think 5D3 would be my choice from reading the users reports. Smaller file sizes, quiet mode shutter that actually works, great lens choices, softer video but no moire, better low light video performance, better jpeg low light performance.

I agree with you about DPreview sites. They are terrible. But you do pick up some interesting data, especially for me as someone wanting to improve my knowledge of photography.

I am not sure about your comment about more mp means greater dynamic range and lower noise. Its the opposite. Larger pixels perform better than smaller pixels and would have lower noise because they accumulate more signal per pixel for a similar level of read noise, dark noise.

Dynamic range is defined by the read noise and the well depth. Larger pixels almost always have much larger well depth.

So for a sensor maker to increase dynamic range, lower noise, increase megapixels is an impressive feat that few seem to be able do. These are conflicting goals. Look at the specs of the Kodak chips and you'll see their larger chips all have much lower QE and tend to be slightly lower dynamic range.

So hat's off to Sony.

Also when measuring a sensor the fact of identical sensors is only part of the equation. The surrounding electronics can make a large difference.
STL 11 and FLI Proline 11002 use identical chips. But the surrounding electronics makes a huge difference. Look at 2 darks one from each camera. They certainly don't look like the same chip. One is almost noise free, the other very snowy. Same with download speeds, cooling. Same with cameras. So the DXO mark there would be completely valid in my opinion iwth different scores for 2 different cameras with the same sensor. To exaggerate the differences, imagine 2 cameras one with a 12 bit analogue to digital converter and the other a 16 bit.

Sony sensors for a long time seemed small little planetary chips that dominate in their area. Most are quite small. Obviously they are pushing ahead in the CMOS arena bigtime. The quest in the electronic sensor game is to get smaller and smaller circuitry and advance the architecture. You may not know but Nikon also make printing gear these chip manufacturers use to print the circuits on the chips. Not sure if Sony use Nikon stuff. Canon make them too but only for themselves. Perhaps that is limiting cashflow as these machines are worth many many millions and a plant in the billion dollar bracket.

On another note there's an interesting video comparing D800 to Hassalblad HD40. HD40 uses Kodak KAF40000 which is a true sense chip (not Bayer matrix). Its also 16 bit and its interesting to see how subtle skin tones are rendered by Hassy over the D800. It wasn't a professional comparison as it has flaws but you can see where DSLRs will have to go eventually. 16 bit, larger sensors to medium format.

As far as Nex 7 and another Hassy it would depend on the criteria. These Hassys bomb out past ISO400. D800 beat the Latest Hassy HD0 in this regard. I am sure 5D3 would also. Dynamic Range is better in the D800 which is Sony Exmor chip. The Nex 7 is Sony Exmor chip just the "smaller" 24mp one. But in a fashion shoot of course you'd reach for the Hassy. All the other aspects make it rule. But as far as chip performance goes I don't know that Kodak are the king anymore, Sony is. Kodak's a long gone company unfortunately so I don't expect a comeback unless the new owner of the chip company is on the ball. No indications are there to think that that is the case (they are still using the old Kodak website - what's a new website cost at the quality of the old Kodak one - $10,000?

I am expecting a D800E possibly as early as next week. I'd love to go on a photo shoot somewhere and learn from you. You could have a go with it and slum it with a Nikon for a while:rofl:. I've got a few nice lenses.

Enjoy your 5D3, I agree its a great camera and I would have gotten one for sure if the D800E hadn't've come out that was more seductive to me.For me its all about the sensor. I think the 5D3 is probably the more refined all round camera. Its just 36mp man, whoa - can't resist it!!

I have my eye on a Fji Xpro 1 (AA less 16mp). The images from those are stunning and its so compact. Lousy autofocus though.


Greg.

[issdaol]

I have always been a Canon owner but the 800E has me considering a change as I have not upgrade for some time. The new Sony sensors are very nice although it is surprising that Sony don't use the 36MP version in their SLR (maybe that is to come? or licensed only to Nikon?).

The Kodak Truesense Chips are still referred to by Kodak as a "Bayer" Filter Chip. They have basically taken the standard Bayer Filter Matrix and added a "Clear" pixel that accepts all wavelengths. Luminance which in the past was information on the Green pixel in now only captured on the new "Clear" pixel. Chrominance data is captured across Red, Blue and Green pixels. There is a trade off that Chrominance now represents 50% of the data and luminance is 50% as well. In the standard Bayer 100% the data is Chrominance.

[DavidTrap]

Agreed!

I also echo Greg's comments about differing results from sensors in different cameras - it's the whole package that makes the image.

H - your cat picture did prompt me to take a similar photo with my D800. I was pretty impressed by my results, but chose not to post on your thread to avoid inappropriate comparisons (comparison shots are really only valid if taken in exactly the same manner). As for all the interest in high ISO performance, I personally think 99% of it is a technical exercise. I can count on one hand with no fingers the number of times I have wanted to take a natural light portrait of a coal miner down the mine at midnight. I used my D800 last night under dark skies 150km west of Brisbane. The sky glow recorded at ISO6400 (30sec widefield images) will probably make them unusable, but I haven't had the chance to process the images yet. ISO3200 is probably going to be enough for me, but if I want to take a pic to record something at a higher ISO, I'll happily turn up the ISO and accept that it's unlikely to be a smooth sharp image.

DT :thumbsup:

[Peter Ward]

The problem with that analysis is whether or not the noise is correlated.

SBIG have sublime temperature regulation (1/100th of a degree)....better than most....hence you can ramp up the gain and get a tad more signal and, yes, you get snow, *but* when reduced the data has bugger-all correlated noise.

I have stacked thousands of frames with SBIG gear and have yet to see a correlated noise pattern build.

[gregbradley]

That is true. Its not really a problem as you point out as it dark subtracts out well but as far as measuring sensor performance goes I was using it as an example with these DSLR sensor comparison performances. Its not all about the chip but also the quality of the surrounding electronics.

Greg.

[CometGuy]

What puzzles me with the DXO lab tests for cameras is there is already an established way of reporting sensor performance, i.e:

Quantum Efficiency Curve
Read Noise
Dark Current @ T
Full Well Capacity
etc.

Why not test and report on these. Sure measuring the Quantum efficiency curve is a bit tricky, but you'd think it would be well within there resources.

Terry

[gregbradley]

Terry,

I just googled geometric quantum efficiency which is as I recall what Sony use.

Here it is:

http://books.google.com.au/books?id=s6rE3Z6PK44C&pg=PA321&lpg=PA321&dq=geometric+quantum+efficiency&source=bl&ots=-FOmCDmYms&sig=ssxko11hjQEw5EfF5GvRDUAYdD8&hl=en&sa=X&ei=9paTT5-xCcmYiAfqjZ2_Dw&ved=0CEUQ6AEwAg#v=onepage&q=geometric%20quantum%20efficiency&f=false

Translation:

You multiply QE by the sensor ability to cover the surface of the chip (fill factor - how much of a gap between pixels?).

I assume fill factor is a number greater than 1 hence the high QE numbers Sony presents. Although fill factor you would think would be a number less than 1 (1 being the whole area??). Otherwise the QE would be much greater than 77% and that would put it up with an ST10. Unlikely or the Sony chipped cameras would be leaving the Kodak ones for dead in astrophotography and they aren't.

Anyway I suspect its a marketing ploy using definitions that present your product in a better than normal way.

Greg.

[CometGuy]

Yeh thats interesting Greg. Quantum efficiency should include Fill factor for sure.

Terry

[Peter Ward]

My understanding is fill factors are always less than 1. Fill factor is simply a percentage of a pixel than can "see" light compared to the bits of the pixel (gates etc.) than can't.

You can improve fill factor by having bigger pixels, smaller gates and micro-lenses.

Sony's QE's I don't believe are anything like 70%. QE is a strict ratio of actual flux to detected flux. Sony sensors bleed both noise and charge so the signal might look clean, but is also reduced in the noise reduction process.

I understand their real QE to be about 30%. Sony only quote "relative" QE specs... a complete waste of time unless all you are concerned about is colour balance.

Don't believe me? Try this easy test: Shoot though a 7nm SII filter with a DSLR. I suspect the you'll find the results will be very sad...;)

[Shiraz]

I understand their real QE to be about 30%.

Not sure about that Peter. Absolute QEs are quoted for a few Sony sensors on the Point Grey site - understand they do their own measurements. http://www.ptgrey.com/support/kb/index.asp?a=4&q=361&ST=quantum+efficiency
Current mono ones (and even the 285, which has been around a while) have peak QEs somewhere between 60 and 80% and >50% through the Bayer filters for the colour ones. Most interesting is the high QE of the new back illuminated CMOS devices - they have no gates to get in the way. Interesting times on the way?
regards ray

[gregbradley]

Thanks for that. I got the wrong term. Yes relative QE is what they quote. So that means QE relative to different colours?

A bit like Kodak quoting QE for RGB in one shot colour sensors?

I found a definition - not totally clear but here it is:

http://www.acronymfinder.com/Relative-Quantum-Efficiency-(RQE).html

So it would mean here some portion of the light radiation hitting the sensor. Presumably the red separately from the green separately from the red. but not the IR or UV.

Greg.

[Peter Ward]

I often ask myself of anything I read on the net "Does that seem reasonable", with a touch of Cal Sagan's "extraordinary claims require extraordinary evidence"

Point Grey's "measured" QE for the Sony IX674 is 70% at 385nm....amazing !!! since that beats the very best (and very expensive...they fly similar sensors on Hubble) E2V back illuminated CCD's currently available. I'd suggest Point grey got absolute vs relative QE's confused....

One wonders if they used a rain gauge :)

[Peter Ward]

Sort of. A relative QE curve usually has the same shape at the absolute curve, except the manufacturer drops the Y-axis peak QE down to 1.0

The problem being they don't state what the peak value is.

I found this useful snippet on the JAI website.... it describes the QE vs noise parameters better than most:

"Charge conversion (also referred to as pixel sensitivity) is a measure of how much voltage is
created from each electron in the pixel well .

If we combine the charge conversion number with the QE number (and hold the fill factor constant), we should be able to get a good baseline for our sensitivity comparison. Unfortunately, while Kodak states what the charge conversion factor is for their imagers, Sony does not. The newer Kodak imagers have a stated charge conversion rate of approximately 31 μV per electron.

Sony doesn’t list this information, but calculations indicate that for the newer Sony imagers, 8-10 μV per electron is typical.

Ultimately, sensitivity is a measure of what minimum amount of light is required for a camera/
sensor to produce an output level where the image (signal) is readily distinguishable from the
noise of the sensor and camera. Sensors with a good combination of fill factor, QE, and charge
convergence can reach this output level in relatively low light. Sensors and cameras that are also
very “quiet” (high signal-to-noise ratio), can produce a meaningful output level with even less
light because more gain can be applied to the signal without producing an unacceptably noisy
image."

Hope that helps :thumbsup:

[bratislav]

That is not a valid test and you know it. DSLRs are nearly blind at 672nm and that is nothing to do with QE and everything to do with built in shaping filter. A valid test would be using Sony equipped CCD and then use SII filter. May I suggest a FLI Microline MLx285 for example.
The results will not be sad at all, I'm willing to bet.

[luigi]

The reports that show the D800 ahead are objective.
The reports that like the 5DIII are subjective.

Nikon made the logical succesor to the 5DII, while Canon made the logical succesor to the D700, oh the irony...

[Peter Ward]

Ah, correct, but the point I was trying to make was the "relative" QE curve might still be great. The Absolute result (albeit due an I/R cut filter) will be tragic.

[gregbradley]

[QUOTE=Peter Ward;844601]Sort of. A relative QE curve usually has the same shape at the absolute curve, except the manufacturer drops the Y-axis peak QE down to 1.0

The problem being they don't state what the peak value is.


Yes that is good. Thanks for posting that.

Greg.

[Shiraz]

Peter, its not just Point Grey who have measured Sony CCD performance.
Did a quick search and found that Apogee, Carl Zeiss, FLI and Photometrics have also published absolute QE curves for the ICX285 that are similar to the Point Grey data (not identical though, so not copies). The (previous generation) ICX285 has an absolute QE of at least 60% over much of the visible band according to all sources - it should not be surprising that the newer Sony chips are even better - they have been in this game for a fair while.
http://www.ccd.com/ascent_disc_qecurves.html (http://www.ccd.com/ascent_disc_qecurves.html)
http://www.zeiss.de/C1256D18002CC306/0/4E9EF8C9122BBA9CC1257149005146D1/$file/48-0065_e.pdf (http://www.zeiss.de/C1256D18002CC306/0/4E9EF8C9122BBA9CC1257149005146D1/$file/48-0065_e.pdf)
http://www.flicamera.com/spec_sheets/MLx285.pdf
http://www.photometrics.com/products/datasheets/HQ2.pdf


However, this discussion on Sony chips really misses the point of David's thread - the new cameras that are appearing are exceptional. Back illuminated 20mp+ chips promise higher QE than affordable existing chips and they will have enough small pixels that they can be adapted to any chosen scope and seeing conditions by way of binning. Provided the manufacturers make them in mono form, there could be some very interesting astro cameras available in the near future. Regards ray

[gregbradley]

It seems like Sony will take the mantle over from Kodak.

57% QE in a CMOS 36mm 4.88micron pixel DSLR full frame chip is a sensational achievement.

I agree, I think once they come out with some larger chips they will be unstoppable.

Greg.

[2stroke]

Nikon just need to fix up there entry level gear now, they need something to go against the 1100D and ill be happy. Lol not all of us have the coin :(

[DavidTrap]

D3200 just released.

DT

[gregbradley]

Yes Nikon D3200 24.2 mp and HD video with wifi adapter available for about $50, plus larger LCD screen. Sample images are amazing.

Canon 1100D 12.2mp, smaller screen. Not sure about video.

Greg.

[2stroke]

The d3200 is a $900 body lol would put that at the 60D not entry level. The d3000 is the same as the 1100d price wise but its to old and the 1100d runs rings around it. Next is the D3100 but priced at 600D level and is near equal to the 1100D, nikon need to get something to compete in this segment of market.

[RickS]

Here's an article on DxO Labs with a little info on how they do their testing. It's a bit light-weight, but I thought it was interesting...

http://news.cnet.com/8301-11386_3-57415777-76/how-dxo-labs-tests-hot-cameras-like-canons-latest-slr/

Cheers,
Rick.

[gb_astro]

Real world comparison:

http://www.fredmiranda.com/5DIII-D800/

Page 2 shows shadow detail comparison at 100 ISO.
The Nikon seems miles ahead in this area.

gb.

[DavidTrap]

I've read that review GB, and I'm not really sure what to take away from it. He certainly demonstrated the dynamic range of the D800, but expressed concern about focusing through LV which is his preferred technique. He's using some fairly exotic glass (tilt-shift lenses) and it sounds like he's been around long enough to have used a 5:4 view camera where such a technique is the norm. Call me biased, but I just got the impression that he didn't really give the Nikon a go.

Ultimately, there are few people who will have the chance to really compare these cameras side-by-side. I've got far too much invested in Nikon lenses to change camps to Canon, and the closest I'll get to a Canon 5D MkIII will be using a friend's camera.

DT

[gb_astro]

Yes David, I thought he could have been a bit more excited about the D800 especially for a landscape photographer where dynamic range and sharpness at low ISOs is critical.

gb.

[DavidTrap]

Had the chance to try out my D800 today in difficult lighting conditions - poorly lit concert hall.

I was shooting at ISO4000, taking photos of my kids singing and dancing. Very impressed with the quality and low noise. Regardless of how it compares to the 5DMkIII, I'm stoked with the new generation of sensors!

DT