May 2, 2011

Results from the Single-Chip Camera Evaluation Conducted by Robert Primes, ASC

One of the highlights of NAB's "content theater" screenings was the Single-Chip Camera Evaluation, the result of an exhaustive camera shootout conducted in February by Robert Primes, ASC and a full crew (totaling what was estimated at over 5,000 man- and woman-hours). After seeing the terrific half-hour presentation at Zacuto's booth, I went back for a second look at the full presentation. While the images -- which should be released online in the future -- are far more important than the charts, here are some key results from the screening, which featured cameras ranging from the cheapest Canon DSLR to cameras costing hundreds of thousands.

Unfortunately the RED EPIC wasn't available at the time the test was conducted, so the RED is represented by a RED ONE MX in these tests. These are just screengrabs via ProVideo Coalition of the test itself; stay tuned for Zacuto's behind-the-scenes videos coming in June. There was also mention of a potential Blu-ray release of the test.


Note these are "line pairs," which is double the traditional "TV lines" measurement -- if you're used to the latter, multiply each number above by two. I also suspect that the 1D Mark IV's focus was soft in this test. And, of course, measured resolution does not take into account aliasing and other artifacts that may make one camera's resolution cleaner than another's.


Note this chart starts at 9, not 0, so the differences between each camera are not nearly as pronounced as it seems visually. As with resolution, not all latitude is created equal -- some cameras retain more highlight detail (film especially), whereas others pack in much more shadow detail (of which the RED is a good example).

Ultimately it's easy to get caught up in a numbers game, or to look at a particular test chart and say, "that's ugly." But the fact is, few of the flaws exposed in these tests are going to outright ruin a film. The only issues that I saw as potential deal-breakers are the infamous DSLR aliasing, and the slow CMOS readouts of a few cameras (which will only come into play if you're shooting an action film or otherwise need a very nimble camera). Every single one of the cameras featured was, to me, perfectly adequate to the task of telling a great visual story.

[via ProVideo Coalition]

Your Comment


Any idea if those DSLR latitude numbers are somehow from the technicolor LOG profile? I know it wasn't out when they did the test, but these guys have enough clout to have been able to test a beta version. I'm just asking because those numbers are a full stop above any other dynamic range test I've seen. I'm assuming the F3 footage is REC709 as well.

That would also mean that if that's still REC709, then the new LOG profile would put the DSLRs on par with RED MX, only bested by the ARRI ALEXA and film (and F3 S-LOG)

All of these cameras are amazing, but I'm really exited about the next generation of Canon SLRs and whatever real video camera they decide to make with their next-gen large chip.

May 3, 2011 at 12:06PM, Edited September 4, 10:54AM

Dustin Sims

It wasn't the Technicolor LOG profile. I agree, the DSLR latitude numbers are at least a stop above what I'd seen in the past...

May 3, 2011 at 2:45PM, Edited September 4, 10:54AM

Ryan Koo

DxO rates the 5D2 and the 7D at 11.9 and 11.7 stops, and that's for stills in RAW mode

no matter if this is Technicolor or any other ultra flat picture style, it's using nearly all the DR that the sensor can provide

May 3, 2011 at 6:45PM, Edited September 4, 10:54AM


@Samuel, I can hardly believe the 7D or 5D are only 0.7 or 0.9 stops better in RAW stills than in video mode.

I mean I always shoot RAW stills because they have a lot more latitude than a normal JPEG can hold.
I don't know, what is the latitude a normal 8-bit JPEG picture can display? Because it seems to me a Canon RAW from the 7D has at least 2 stops more latitude than that.
I think the h.264 cannot under any circumstances be better than a JPEG still, so there must be more difference in latitude between the h.264 and the chip itself!

May 5, 2011 at 5:43PM, Edited September 4, 10:54AM


There's no limit to how many stops an 8-bit JPEG could represent. But JPEGs are meant to be ready for display or print, and monitors and prints have relatively small dynamic ranges, so the camera's image processor will truncate the range in order to get an image that looks properly exposed and suitably contrasty on those mediums.

May 6, 2011 at 7:40PM, Edited September 4, 10:54AM


Saw the presentation at your booth at NAB. Thanks. I was just wondering why the Panasonic 3700 was not in the mix. Any insights regarding the criteria with regard to cameras included in the comparison?

May 13, 2011 at 6:12PM, Edited September 4, 10:54AM

Dennis Boni

I wonder why thet did not include Panasonic GH2. Maybe it does the same as AF100 though.

May 13, 2011 at 9:07PM, Edited September 4, 10:54AM