Canon 6D Dominates Low Light Focusing Test vs 5Dii, 5Diii & Nikon D600

[MichaelTheMaven]

I did some more tests today with the Canon 6D, and one question I was most interested in was the low light ability of the Canon 6D's central focusing point.

The summary is this:

Between 0 --> -3EV, the Nikon D600 is about 10-15% faster than all of the Canons.
Between 0 --> -7EV, the Canon 6D is almost twice as fast as the 5Dii, ~20% faster than the 5Diii, and the Nikon D600 couldn't achieve any focus at all in extreme low light without it's AF illuminator.

(Note: Here "0" EV was an even exposure at 1/60, 2.8 ISO 400).

For those of you interested in the methodology and actual testing, as well as the specific results, you will want to check out my blog article on the test http://www.michaelthemaven.com/?postID=2577&canon-6d-vs-5diii-vs-5dii-vs-nikon-d600-low-light-focus-speed-comparison.

I also tested the top center, and far right center focus points.

- While the Canon 6D's other focusing points are not nearly as impressive as the other focusing points on the 5Diii and Nikon D600, the 6D's center point is in fact very, very good in low light.

- The Nikon D600 and 5Dii's Central focus points felt much quicker, and snappy. The 5Dii, actually did very, very well in 0-> -3EV range. The 5Diii's & 6D's feel like there is a 2 step process to the focusing.

- The more I play around with the 6D, the more pleasantly surprised I am with it. (When I initially heard the specs I was seriously underwhelmed).

While I certainly wish there were more cross type focus points on the 6D, I think that with the low light focusing and relatively low ISO noise, the Canon 6D is going to be a very solid performer in low light situations.

M

[davidbellissima]

Very interesting test Michael. I own all three versions sof the 5D and I am frustrated by the very slow focus lock time, in very low light, of the 5D3. It can take a second or longer. It does seem like it is a two step process: get 95% there is maybe 30-40% of the time and the image through the viewfinder seems to be in focus, but then there is another 60-70% of the time taken on what must be some sort of focus refinement algorithm. Of course the total focus locking time varies with the lens being used, but it is still a pain. Imagine wedding first dances where the subjects are moving and the camera is taking an age to lock focus.

Anyway, maybe when I get rid on my 5D MK1&2 I will buy a 6D as the low light performance looks very good, if not better than the Mk3 and since I mainly use the centre focus point, that will do for me. That is, as long as the focusing on the 5D is as consitent and repaatible as on the Mk3. The Mk3 focusing accuracy and repeatibility is not the best...

[ScottyP]

Thanks Michael.  Very interesting indeed. A lot of folks wanted to see exactly that.  I wonder if you would be able to confirm/quantify the issue of 5d3 + 600ex speed lite AF assist = even worse lag?  (and comparing 6d on same speedlite AF too)  No one else is going to, and yet a lot of people are very interested in that! 

[zim]

I thought the 6D autofocus in -3EV conditions was a one stop improvement over the 5d Mark III and a two stop improvement over the 5D Mark II? I don't know about the Nikon.
I don't really get testing beyond the rated limit, maybe I missunderstand further explaination of the -7EV thing would be much appreciated. I also seem to recall a video from DigitalRev (in the boxing ring) where the 5D3 excelled against the D600 no where near -3EV?

[ScottyP]

I don't really get testing beyond the rated limit, maybe I missunderstand further explaination of the -7EV thing would be much appreciated. I also seem to recall a video from DigitalRev (in the boxing ring) where the 5D3 excelled against the D600 no where near -3EV?

How can you not understand the interest in testing beyond the rated limit?  The light around you won't stay to the rated limit of your camera, right?   

[neuroanatomist]

(Note: Here "0" EV was an even exposure at 1/60, 2.8 ISO 400).

You're adding a lot of confusion here by arbitrarily defining your own zero.  It's like saying you're going to start counting everything from the number 4, so if there are of 2 of something on a table, you're telling everyone there are 6.  Makes no sense.

0 EV has a real, absolute definition - an exposure of 1 s, f/1.0, ISO 100.

Technically, your "0" EV is actually EV = 5.  That means your -7 EV is really -2 EV (and in fact, the 6D is spec'd to AF at -3 EV, but of course, that may not equate to real world numbers).

It would help avoid confusion if you restate your numbers according to the real definition of EV = 0, otherwise many people will read that you're testing at "EV = -7" and think you're spewing complete bullsh cow manure.

Sorry, but the fact that you don't seem to understand some of the basics, like the fundamental definition of the basis for the specification you are testing, makes me question what you're doing.  I definitely appreciate individual testing (and especially the fact that you did post your definition, so the error could be spotted), but post/blogs like this contribute to the vast and growing pool of bad and/or unreliable information out there. 

How can you not understand the interest in testing beyond the rated limit?  The light around you won't stay to the rated limit of your camera, right?   

See...this is my point.  He's not testing beyond the rated limit of the 6D, he hasn't even tested it as low as the rated limit.

[prjkt]

as a side note, in terms of AF assist with a 600EX, there was no noticeable difference in time between focus acquisition with the flash on (AF assist on) or switched off while shooting social shots in a nightclub for me

[RS2021]

Ugh... Hope tests like this don't sway people. I wouldn't even know where to start with the holes and misinformation here. 

[wellfedCanuck]

Ugh... Hope tests like this don't sway people. I wouldn't even know where to start with the holes and misinformation here.

I don't have the technical expertise to argue one way or the other- but if you're going to  make statements like that here then you owe the forum something to substantiate your assertion.

[DavidB]

Thanks for the test results Michael.  I'm curious though if you did any tests comparing the expansion points on the 5d3 vs the 6d?  In my experience having 4 pt expansion greatly increases the speed of AF in low light on the 5d3. 

[skitron]

Ugh... Hope tests like this don't sway people. I wouldn't even know where to start with the holes and misinformation here.

Given the point was to do a relative comparison and not an absolute measurement, I don't see much of a problem. Especially when he cited his definition of 'EV'. As neuro pointed out, it might add to the confusion a bit but it doesn't really invalidate anything.

That said, I'd be interested to delve in a bit deeper and test it's low light performance with respect to the ability to track and lock a moving subject relative the others. For me this will be the determining factor whether I send mine back for something else. Mine is showing up tomorrow, so I'll test and post...and I promise not to couch the results in terms of EV whether used correctly or otherwise... 

[neuroanatomist]

Given the point was to do a relative comparison and not an absolute measurement, I don't see much of a problem. Especially when he cited his definition of 'EV'. As neuro pointed out, it might add to the confusion a bit but it doesn't really invalidate anything.

No, but now that I've looked at the methodology, there are other holes to be poked in the 'analysis'.  He's testing targets illuminated to different light levels, yes, but they seem to be different targets shot from different angles, therefore they may not be comparable for the same camera (and shooting an orthogonally-shaped target from an angle isn't the best way to test AF performance). 

The better way to have done this test would be to have pointed the cameras straight at the same target, and vary the light levels only.  I'd have set the illumination at 2 EV, then stepped down the illumination to -4 EV in 1-stop increments using ND filters on the lens.

Having said that, a generic conclusion that the 6D center point AF performs very well in low light is quite reasonable.

[3kramd5]

0 EV has a real, absolute definition - an exposure of 1 s, f/1.0, ISO 100.

Then EV is amusingly useless since there are varying definitions of ISO 100.

I always thought exposure value was independent of sensitivity (i.e. merely a function of f number and exposure time). f/1 for 1 second happens to be EV 0, but EV 0 isn't defined as such.

Perhaps I'm mistaken.

[neuroanatomist]

Then EV is amusingly useless since there are varying definitions of ISO 100.

LOL...now that's amusing.  "ISO 100" is defined unequivocally by ISO 12232:2006 as published by the International Standards Organization (which is why we call it ISO). 

But you're point is valid.  In actuality, when set to ISO 100 the cameras in question are really at a sensitivity of ISO 73 - 80 based on the actual standard.  That's one of the things that DxOMark measures.

[3kramd5]

Then EV is amusingly useless since there are varying definitions of ISO 100.

LOL...now that's amusing.  "ISO 100" is defined unequivocally by ISO 12232:2006 as published by the International Standards Organization (which is why we call it ISO). 

But you're point is valid.  In actuality, when set to ISO 100 the cameras in question are really at a sensitivity of ISO 73 - 80 based on the actual standard.  That's one of the things that DxOMark measures.

I don't have access to the standard, but per wiki:

"The ISO standard ISO 12232:2006[55] gives digital still camera manufacturers a choice of five different techniques for determining the exposure index rating at each sensitivity setting provided by a particular camera model. Three of the techniques in ISO 12232:2006 are carried over from the 1998 version of the standard, while two new techniques allowing for measurement of JPEG output files are introduced from CIPA DC-004.[56] Depending on the technique selected, the exposure index rating can depend on the sensor sensitivity, the sensor noise, and the appearance of the resulting image. The standard specifies the measurement of light sensitivity of the entire digital camera system and not of individual components such as digital sensors, although Kodak has reported[57] using a variation to characterize the sensitivity of two of their sensors in 2001."

So while maybe sensitivity has a specific definition, the affect of sensitivity on the result has myriad possible results, hence different cameras recording different luminance levels with equivalent exposure settings.

Also from wiki, EV=log2(Fnumber/ExposureTime).

In any case, it has no bearing on this discussion.