Shooting Anchors: The First Feature Film on the 5D Mark III
In mid April, I had the honor of being the first cinematographer to put the Canon 5D Mark III to the test on a feature film, David Wexler’s “Anchors,” which stars Blue Bloods’ Will Estes and The Chernobyl Diaries’ Devin Kelley.
In this post I’ll give you the lowdown on how the eagerly awaited update to the camera that started the DSLR revolution performed under pressure…serious pressure. And since plenty has already been said about the differences between the Mark II and Mark III by experts like Philip Bloom, Vincent LaForet and Shane Hurlbut, I won’t spend time listing them, but will instead focus on how the camera performed in specific situations and address the differences as they come up. Cool? Cool.
As some of you probably know, at the time of the shoot the Mark III had been out for less than a few weeks, and although I hadn’t really had the opportunity to do much more than casual testing around the Endless Picnic office and at home, my extensive on set experience and thorough testing of the MKII (see EP’s tests here) led me to believe that I would be comfortable with the camera on a tight shooting schedule–if it behaved similarly to its predecessor.
Nonetheless, taking a brand new camera on location and asking it to perform 12+ hours a day–without a backup body, which the production’s tight budget precluded–certainly involved some risks. In a sense, I felt like we were doing a beta test for the camera, and that there was a chance that it would fail at the worst possible time! (But more on that later.)
The low-budget nature of the project meant the shooting schedule was going to be tough, too. How tough, you ask? Try nine total shooting days, with only eight at our principal location, the beautiful Gild Hall hotel in lower Manhattan. Which meant covering a rather terrifying 8-14 pages a day.
Luckily I had my wizard-like right hand man on the show–cameraman Andrew Baxter–who was initially brought on as first AC, but wound up doing double duty as gaffer. (His background as a rock climber makes him the best grip and rigger I’ve ever seen, too.) The man can stand pressure like no one I’ve ever met.
Although we had a relatively minimal amount of prep time–both Baxter and I were coming off a hectic month of shooting other projects–we were able to pull together a rig we were confident was going to be relatively versatile, and help us avoid the usual (and now too familiar) drawbacks of working with a “modified still camera” on a movie set: the MKIII body, EP’s customized Red Rock 15mm studio rod system, matte box and handheld system and an Arri FF4 follow focus. Last but not least, we conjured a sweet set of 4 Zeiss CP2 primes (35mm, 50mm, 85mm, 100mm.)
Andrew and I had worked on numerous occasions with both the Canon L series and Zeiss ZE DSLR glass with “add on” gears, but given the incredible time crunch we knew we would face every day, the geared cinema style Zeiss CP2s–with proper focus scales, uniform size/weight and speed–would allow us to work much faster. (For the record, a typical lens change with geared ZEs or L series on this type of rig generally takes between 3-5 minutes due to the adjustments that need to be made to the rod system and follow focus, whereas the cinema style CP2s could usually be swapped in less than 30 seconds!)
For monitoring–always a thorny issue in the world of HDMI–we decided to go from the camera’s mini-HDMI port (gripe: Canon didn’t give us a regular HDMI out on the Mark III) to our trusty upgraded Marshall 7″ V-LCD70XP-HDMI. The signal came out of that unit’s HDMI out via regular HDMI cable to the Zacuto EVF, and finally from that unit’s HDMI passthrough to the workhorse 17″ Panasonic BTLH-1710 in video village.
Now, ideally we would have gone from the EVF to the Marshall 7″ to the Panasonic, but we could not. The BTLH-1710, as good as it is, doesn’t accept an HDMI input (just its sibling, DVI) and the Marshall wouldn’t output anything but digital noise from its HDMI passthrough to an HDMI to DVI cable. (We tried for several hours to correct this problem, but decided categorically the problem lies in the Marshall’s passthrough circuitry.)
As a result, no matter what configuration of the rig and monitoring we wanted to use, in order to get to the Panny in video village we had to incorporate the Zacuto EVF, who’s HDMI passthrough was able to send a useable signal from the HDMI to DVI cable into the 17″. This wasn’t always welcome, especially when the rig needed to be stripped down and squeezed into a corner or lifted high into the air for a bird’s eye shot or slider dolly, a common occurence on the set of Anchors. In those situations one more set of cables and one more step in the signal chain often to be proved a significant burden, and slowed things down.
And since we’d decided to shoot the film 2.35:1–the director and I agreed that it lent itself to bold, aggressive framing and a highly cinematic look–Andrew and I spent a good part of our final prep day getting the 2.35:1 frame lines in place on every monitor throughout the chain.
With the Zacuto EVF, it was a piece of cake–just pop through its menus and turn them on–but while the Marshall initially allowed us to dial in 2.35 markers with its “user” setting, throughout the shoot the guides kept disappearing, and only a full reset of the unit to factory defaults–not the sort of thing you want to be doing on set when time is at an absolute premium–could get them back. (We did try various forms of taping the lines directly onto the 7″ Marshall’s display, but I’m a pretty fussy operator, and, to be blunt, they drove me crazy and Andrew always wound up ripping them off and trying to get the Marshall to behave right before a shot.)
Unfortunately, unlike the EVF and the Marshall, the 17″ Panasonic doesn’t have an HDLSR scaling feature–let’s not forget that the 5D, updated or no, still outputs a 16:9 frame within the 3:2 still frame windowbox, which means the frame size is hardly traditional. As a result, there was some careful lining up of frame lines and taping of the monitor to be sure video village had the same frame that the camera team was seeing. (If you’re following at home, that means 2.35:1 nested within 16:9, letterboxed within 3:2, windowboxed on a 16:9 display. Whew! It wasn’t easy.)
As far as the general camera settings were concerned, Baxter and I had decided during prep to be as disciplined as possible regarding our choices. For various reasons–the look, primarily, but also our tiny lighting package and small G&E team–we decided to shoot the entire film wide open at f2.1, and, wherever possible, shoot at ISO 320 for night work, and 100 or 160 for day work. (The sequence that ends the film, while entirely set during the day, is almost entirely lit, with only minimal daylight peeking through the lower Manhattan hotel’s windows–those NYC “canyons” don’t get much light!)
I should also note–and this is a trade secret here, so listen up!–I shoot virtually everything I do on HDSLRs at 24p at a shutter of 1/30 a second. Why? Well, despite the fact that traditional film camera math dictates that the normal shutter speed for 24fps (given a standard 180 degree shutter) works out to 1/48 of a second, 1/30 just looks more organic and natural. And, given the fact that I’m a cinematographer, I trust my eyes.
So now that I’ve filled you in on nearly every little detail of prep and the gear and approach we’d settled on, you’re by now no doubt wondering: “How the hell did this new camera do?” The short answer is…OK.
The biggest problem was that Canon–in my not so humble opinion–decided to change the camera’s general layout to closely resemble the 7D (which split the camera into indistinct but separate video/stills modes.) This meant that I had to rethink virtually everything I was doing in regard to menus, starting and stopping the camera and adjusting settings. Not exactly a perfect scenario for either me or my AC working under pressure. One example? The handy 10x focus magnifier has been moved to an extremely awkward spot on the other side of the camera. It’s nearly impossible to hit it while you’re holding the camera.
The second biggest issue was certainly the menus. After the shoot wrapped, Andrew commented that [they are] “far more elaborate [than the MKII] and for no good reason that I can think of.”
And I agree. While the 5Ds menus were hardly intuitive–not that anyone needs reminding, but we’re dealing with a stills camera here!–they were relatively stripped down and with a bit of practice, could be navigated fairly quickly. I cannot say the same for the MKIII. While some may argue it’s something you’ll get used to over time, and that might be true to an extent–there’s far more clutter and redundancy in the new layout–sometimes to a bewildering and infuriating degree.
Another problem area Baxter cited as major “but that should be avoidable” was switching from stills to video mode. “Occasionally when we turned the camera off to switch a battery or lens, we would have to re-patch the HDMI mini in order to get picture again, which cost us a fair amount of time. Hopefully that’s something that Canon will work on in firmware updates.”
And, as I intimated earlier, there was one major crisis with the camera. About 7 days in, the camera started behaving erratically-the LCD screen flickered, and the camera’s still light meter settings wouldn’t leave the screen when in video mode. Finally, the camera stopped recording. Once it started again, it refused to output a 720p signal (only SD) while in record mode, which meant all of monitors would be screwed up.
We tried various fixes–changing the camera battery, card, cables, turning it on and turning it off, updating the firmware, etc. all to no avail. The camera did eventually start recording again, but the 720 out during recording wouldn’t return. (This caused all sorts of confusion with our monitoring setup.) We then contacted a local rental house and had another body on standby, and called Canon to see what the deal was. That’s when the real fun started.
The short version of this story (and, perhaps, the one that is most kind to Canon) is that no one we spoke to at the company was any help whatsoever. After an hour on the phone with several support staffers, we gave up in frustration. They all insisted that the 5DMKIII does not output anything from the HDMI mini port but 640 x 480!
About an hour later the backup body arrived, and (of course!) from that point on both cameras worked flawlessly through the end of the shoot. (It was very helpful to have the second body on hand to use as a viewfinder for setting up shots and quickly checking exposure on its surprisingly accurate LCD screen, something that was difficult to do quickly on the big rig, with all its HDMI cables precariously arranged and ordered.)
And there were some things that were truly great about the camera. There is clearly a significant improvement in the camera’s sensor. Rolling shutter artificats and aliasing are noticeably reduced as compared to the 5DMKII, and on the whole the camera’s image feels more solid and steady. And while much has been made of how much more sensitive the camera is in the shadows (some literature and reviews have been throwing around “2 stops”) I can say with a fair amount of certainty that while the camera does compare somewhat favorably with the MKII as far as overall latitude goes, I saw nothing close to two more stops of sensitivity in the shadows. (A more definitive test of noise back to back with a MKII would need to be done to say that for sure.) Generally speaking, I felt that the camera held up very much the same way the MKII does in average shooting conditions.
Another improvement that I noticed even during prep was that the color rendition in the 5D does seem to be improved. Greens, in particular–always a tough color in digital video, I feel–seemed far more film like and pleasing.
And having the 720p out during recording–something that some folks at Canon dispute even exists!–when it worked, was a boon. Focus pulling was a bit easier, and the HDMI handshaking delay that plagues the MKII’s HDMI signal as it shifts from 1080p to 640×480 when it records was minimized.
After all was said and done, I’d give the MKIII a B. It boasts several significant improvements over the MKII, and on the whole is moving closer to a proper video camera–although Canon’s insistence on changing the layout and menus was a significant drawback. Video camera manufacturers like Sony and Panasonic know better than that–did the EX3 radically change its menus from its EX1 predecessor? Did the HVX200 mutate into something radically different when it became the HPX170? Is an HDX900 all that different than a HDX3700? The answer is no, because these companies understand that professional camera people value tools that are both familiar and become better with each iteration.
Does the MKIII pass this test? The answer is maybe, although I’m certain that over time I’ll come to find my comfort zone with the camera, just as I did with the MKII. For now, though, for all of its foibles, the MKIII provides a noticeably improved image, and, as my cinematography professor used to say, “What’s up on the screen is the only thing that matters when the lights go down.”
Amen to that!Tweet