Production Notes on “Sitka” (and a review of the Sony a7R II)

The past year has seen an explosion of blogs and especially vlogs about digital filmmaking, in a pattern of paraphrasing manufacturer specifications and press releases in various safe combinations. I suppose that takes the cynical view, but minding my own business, I always figure that there’s no point in saying what’s already been said. What can always be unique, though, are the circumstances of actual productions, and how these digital gizmos actually do their jobs (or fail). It helps when the writer literally makes films (even though that slows down the blog).

I was just about to create a video guide to the new Sony a7R II, after using it for a month and really putting it to the test on a production. But tonight I realized that the footage I already have, right there in the film embedded above, is the best place to start – especially since the a7R II tries to blend in with other, very different cameras – in effect pitting them against each other.

To set this up, here’s a brief background on the film: It’s called Sitka: A Piano Documentary, named for the spruce wood that replaced a soundboard in a Steinway Concert D grand piano at The Phillips Collection in Washington, D.C. It’s the art museum where Glenn Gould had his U.S. debut, effectively launching his career, and it remains after 75 years one of the premier chamber music concert series in the United States. The piano had to be great. My film shows the process of making the piano sound better, and along the way we learn how pianos work in the first place, because it had to be literally torn apart. I pretty much made the half-hour documentary by myself (taking on all the production roles), but Caroline Mousset at the museum and Keith Kerman at PianoCraft made it all possible, granting rare access to this culture of craftsmanship and musical virtuosity. The film launched yesterday as a non-profit, freely available educational resource at www.sitkadoc.com. (It will survive by word-of-mouth and social media: if you like it, please click here to share it!  UPDATE: It’s now available, still free, at Amazon Video and at Indieflix.)

I’m lately infatuated with miniature stabilizers as a stylistic signature, especially after a dance film residency where I shot most everything on 3-axis brushless motorized gimbal stabilizers like the Filmpower Nebula 4000 Lite and CAME-7800. One year ago, I had taken a first run with the Nebula 4000 Lite on a short film called Highstep, and that combination with a Blackmagic Pocket Cinema Camera (BMPCC) and Olympus 9-18mm zoom lens was a dream come true.

The opening shot of Sitka uses just that. It’s the first example of a guiding theme: you always choose the camera best suited for specific environments (not that you needed me to tell you). Going between the dark inside of a truck, and the bright sunlight under cloud cover outside all in one shot, I needed that high bitrate and 10-bit 4:2:2 colorspace of the BMPCC. My other two cameras on hand, the a7R II and the Panasonic GH4, couldn’t deliver the same dynamic range and headroom for post-production clean-up. (I later keyframed the luminance and color temperature during that transition, for one thing.)

That leaves me with an aesthetic decision: As soon as I’m done with the shot, should I jump to my overall better camera? It helps to take a step back and audit the structure of a film: it’s usually in three acts or so, with distinct changes of scene. This one has mostly just three: moving the piano into the museum under sunlight outside, then inside the museum under (mostly) artificial light, and lastly inside the piano restoration facility under (mostly) fluorescent light. Not just a color temperature issue, it makes sense to restrict the unique aesthetic properties of each camera sensor to each of the three “acts,” whereby audiences forgive those differences because of the structural changes in scene.

Thankfully, despite prior misgivings on the customer support side of so-called “Rubber Monkey,” FilmConvert (promo code FOCUSPULLING for 10% off) has caught up offering profiles for common cameras. And after Panasonic finally came around (embarrassingly) with V-Log after making everyone wait a year, even the GH4 could attach to a specific common film stock from its own profile that could (mostly) match the BMPCC and a7R II using their own profiles. As an aside on this, I’ve noticed much hemming and hawing about the virtues of flat log acquisition for adding dynamic range (all true), while in the practical world, its best feature is getting closer to common ground between mismatched cameras. You just plop FilmConvert onto each camera’s log clips, match up the profiles, choose one brand of film stock emulation between them, and you’re mostly there. (It wasn’t so easy in the days long past of “Cine-Like D” on the GH4.)

If you were mostly here to read about the a7R II, we’ve arrived – but you might not like what you read. Given the film’s structure that I decided upon, the Blackmagic got assigned to the museum stuff with an assist from the GH4 whenever off-gimbal, but for the heart of the film, in the restoration facility called PianoCraft in Gaithersburg, Maryland, I assigned the new a7R II for the job – literally on its first day out of the box (and really into the marketplace altogether). I also paired it with the new CAME-Single 3-axis brushless motorized gimbal stabilizer, as a better match for the a7R II’s heft compared to the lighter BMPCC body and lens on the Nebula. The CAME-Single also has 32-bit firmware with encoders that basically ensure even smoother fuzzy logic for that sense of floating in the air.

You must have heard by now at least some reports about the a7R II overheating and shutting down. Among early adopters, and still to this day, I am in the extreme minority who pulled no punches on the subject. Overheating happened to me a hell of a lot. The high-pitched defense of Sony, by a majority of “bloggers,” tells you a lot about how consumer advocacy is sort of on vacation these days. I always think of manufacturers delivering goods to consumers as a fundamentally adversarial relationship. When I hear about a product defect that several others are having (1%, 5%, 10%, it doesn’t matter) – I tentatively scowl at the multinational corporation first, not the little guy.

Camera gear has sailed high in cost, way past the rate of inflation, at a concerning pace only recently. I remember when the Canon 5D Mark III launched at $3.5k, and the Canon C300 at $20k, and indie filmmakers started realizing that the world might pass them by, after a few years of joyriding. The word “pro” has eternally bandied about, but it started getting petty when “pros” (that’s not a qualitative artistic term) began calling these unrecoverable price tags “affordable.” The self-identifying “pro” breed generally has used, for points of comparison, whatever the accounting departments of champagne production houses have paid for shoulder-cams and P2 cards in big corporate contracts – say, broadcasters.

The Sony a7R II costs $3.2k, but Sony has still made boatloads of money, as it holds the record for the “most pre-ordered camera ever.” It’s a portable mirrorless camera you can fit in a purse, and Sony has arguably spent more time marketing its video features than its ability to take pictures. But if you were going to spend that kind of money to use this as an A-camera, making more than just family movies, fuhghettaboutit.

Not that you can’t do your best, working around the bugs: so, back to Sitka. I’ve been trying to experiment with shallower focus while flying on a gimbal, with mixed results, even though the easiest thing to do is just close up the aperture as much as possible to get deep focus and not have to worry about it. The promising feature of the a7R II, however, was its combination of phase and contrast detection for auto-focusing, which even the subsequent a7S II lacks (because it re-purposed the sensor from the old a7S). So, for all of the footage inside the piano workshop during flight on a gimbal, I was in auto-focus mode (the cardinal sin of “pro” filmmaking). I think it worked great! Without even an attached external monitor, and just peeking from behind the CAME-Single into the rear screen, I avoided the worst situations that I knew from experience would cause confusion, by composing shots around that, but in general I think that it held shallow focus well.

As for lenses, I was able to do most everything with just two primes: the native Zeiss 55mm f/1.8 (usually closed up a click or two from wide open), and the native 28mm f/2.0 for those rare wider shots. That’s whenever I was rolling hand-held, showing how the in-body 5-axis image stabilization really shines, especially when you consider that the lenses themselves lack Sony’s Optical SteadyShot technology (which can otherwise combine with the in-body stabilization for even better performance). There are rare cases when I had to later drop Adobe Warp Stabilizer onto sequences with bad motion bumps, or when some rolling shutter distortion happened.

Whenever flying on the CAME-Single, though, I made use of the workhorse native 10-18mm continuous f/4 zoom lens, from the “old” crop sensor days of E-mount, which actually covers the full-frame sensor at around 14mm, tack-sharp with minimal vignetting. And that raises another big point about shooting on the a7R II. Through a weird quirk of post-hoc marketing rationale, the a7R II settled into being officially known as a Super 35mm camcorder of sorts, even though it’s fundamentally a full-frame camera (of course!). The reason is, after jacking the megapixel count up to the standards of competing still cameras, it becomes necessary to use only the central portion of the sensor where there would need to be less pixel-binning (averaging out) compared to a full sensor read-out that scales the 42 megapixels down to a relatively tiny Ultra-HD frame. 4K isn’t so big anymore when you’re talking photography.

One familiar benefit of less pixel-binning is of course the reduction (but not elimination) in aliasing, which was always the bane of every video filmmaker’s work for about a decade. (Wide shots from Canon 5D Mark II’s look awful now, don’t they?) As for Sitka, I was running-and-gunning solo, in documentary style, where a piano was getting ripped apart (and I didn’t want to miss anything). So even though degrading the sensor size of the a7R II for the sake of better video quality became doctrine, I disobeyed whenever I needed a wider shot from my prime lens, quickly. As such, you’ll see, starting at 00:05:25 for example, aliasing (even accounting for scaling on your playback monitor here).

Another benefit of less pixel-binning is the decreased amount of number crunching necessary to scale on-the-fly, which of course affects picture quality considering the lossy quality of XAVC-S compression in Ultra-HD resolution. And it also affects…overheating!

For once and for all: you can’t judge the extent of this under-reported product defect by just putting the a7R II on a tripod, pointing it at a cactus, and seeing how long it runs before it shuts down from overheating. You couldn’t give the camera an easier gig than that: no motion, thus no 5-axis stabilization, thus no number-crunching of complex and changing arrays of pixels. Ever heard a CPU fan speed up while rendering out footage to H.264? And it doesn’t matter whether it’s a continuous shot, or shooting in takes. Both workflows are totally legitimate forms of so-called “pro” filmmaking. But you can bet that fanboy debates flared anyway, about whether it’s even valid to run a extended shot.

During this production, some things (and even informal interviews) ground to a halt because of overheating. Annoying to everyone! I never took long shots, and the environment was room temperature, but it didn’t matter: the accumulation of many shots (from seconds to minutes) would kill all work after about 30, sometimes 20, one time 15 minutes. I tried all the gimmicks, even re-balancing the gimbal so that I could pull the pivoting screen out from the burning chassis. I tried running external power, replacing with “colder” batteries, and moving to a slightly colder room between takes: no appreciable difference. As experience later bore out using my new Sony a7S II (coming in a later review here), even the external Atomos Shogun recorder would have only minimized the problem, but not made it go away. And the whole point of the a7R II marketing is that it finally records Ultra-HD internally.

This is the spot where most of the Sony evangelists would arrive at their talking point, “so if it didn’t work for you, just choose a camera that does.” And yeah, I did commit to that idea at the top here. But I also wanted phase plus contrast detection auto-focus, flying on a lightweight gimbal to get into tight spots, with internal Ultra-HD recording. I needed the a7R II to deliver, and it sort of didn’t.

The a7S II definitely performs better in this area, and that’s what I’m using now (and would have preferred for this shoot). The holy grail I’m waiting for is something of a mash-up between the new crop-sensor Sony PXW-FS5 (for its small size and clever ergonomics), the Blackmagic URSA Mini (for its better colorspace and indie spirit with a price in low orbit), and the full-frame sensor of these delicate Sony a7x purse-cams that don’t even have balanced audio inputs. It’s just a question of when! Meanwhile, I’m strategically only investing in full-frame glass to get ready. And the only thing holding Sony back from making a proper full-frame camcorder is their fear that existing owners of super expensive cinema lenses that max out at Super 35mm will feel abandoned in the “pro” product niche. Naturally, the only solution to overheating is a proper camcorder body, but here we are.

Sony colorspace still has inexplicable flaws at the margins, manifesting the most in skin tones, and all my a7R II piano workshop footage was shot in S-Log2 Cine Gamma using automatic white balance. That latter part might have been a bad idea, but if I were locked down to one of the fluorescent presets, I would have gotten color shift anyway as I moved in continuous takes between different types of artificial light in the workshop. Ultimately, I had to keyframe color shifts in trouble spots, and it’s mostly alright now. I can live with the skin tones, even after pushing them in post, but barely – and I don’t think FilmConvert is causing the problem. Even the BMPCC is much more satisfying after a simple FilmConvert grade.

Regarding sound, I was more than content to use the internal microphones for all the ambient noises in the workshop, from piano notes to drills and jigsaws. It’s a documentary! Whenever I see dedicated boom operators on shoots like this, I laugh. Much is forgiven when you have extremely good audio on one foreground layer – in this film, it was either an interview voice or a good piano recording – even when poorer ambient noises lurk underneath or even dominate for a few seconds.

As a hold-over from my old (very dusty) NEX-6, I did hitch the odd Sony ECM-XYST1M stereo microphone onto the a7R II for a couple of sequences, sometimes when I was rushed for good sound. By no means a high-quality microphone, it does include some suspension to avoid shock from hand-holding the body, and it gets power from the Multi Interface Shoe bus, leaving at least the fact that it’s better than the internal microphones. I’m still really bothered by the fact that Sony keeps ignoring the totally unanimous, long-running customer complaint: that this product kicks any camera into Automatic Gain Control (AGC), which chases audio levels and introduces pumping. Mandatory! Enabling manual levels in the firmware would take about one minute, but inexplicably, the answer to that request is always: “Thanks for your feedback!” Sony might be worried about angry parents calling them, complaining that baby’s first words across the room were recorded too quiet because they themselves chose low manual audio levels, instead of being forced to use AGC. (Let us never forget that Sony balances its books mostly from mass-market consumers who have very short attention spans – so you proceed into cross-over product lines like this at your own peril.)

You’ll hear how that worked out (or didn’t) under the worst of circumstances at the two-shot interview twenty minutes in, where I didn’t have time to put lavs on the subjects, needing later to send heavy lifting over to Izotope RX 5 Audio Restoration Software in post. Pretty magic stuff (way better than Adobe Audition) in terms of cleaning up sounds, and making others disappear.

When Sitka heads back to the museum from the workshop, the a7R II is finished, and I’m running-and-gunning with the BMPCC on the Nebula. Instead of un-mounting and re-mounting that camera for handheld shots, I compromise by using the Panasonic GH4 and Leica 42.5mm f/1.2 lens. It’s not a great match to the BMPCC, and to show how rolling shutter is just shades of awful on any camera, you’ll see wacky wobbles at 00:22:49 and 00:23:46. Adobe Warp’s advanced analysis mode and rolling shutter correction barely helped.

After dealing with the a7R II constantly overheating, I was glad to exploit the GH4 to the extent I could, though. Panasonic’s budget shooter stays cool and runs forever on its internal battery (and that’s why I used it for the standing-up interview with PianoCraft’s co-owner, too).

On resolution, naturally this film, designed for online streaming, would have made no sense delivered in Ultra-HD (basically, 4K). The lowest common denominator of the BMPCC shooting in 1080p brings everything down to its level anyway, but I didn’t downgrade the shooting resolution to HD on the a7R II and GH4 even though it would have saved lots of storage space. Punching into a 4K frame at times really saved my ass, cropping out things I didn’t want visible, while curing transitions during interviews that would have otherwise looked like jump cuts only Werner Herzog can pull off. And when Olivier Cave played the piano, I could set-and-forget the GH4 on a tripod while roaming around with the BMPCC, pointing it into his hands on the keyboard, saving framing and panning and zooming for later in post.

In the end, I felt really gratified by the production experience, Sony meltdown aside. To say something positive, the a7R II delivers extraordinary portability combined with cinematic fidelity, closing the gap between my world and the Arri Alexa. As far as I’m concerned, there’s no meaningful difference anymore. And if this is a classic “camera shoot-out,” of course the a7R II wins against the BMPCC and GH4! I can’t even begin to count the ways, but for each: the BMPCC introduces a shocking amount of aliasing, more than I had ever realized before, especially in the last few minutes of the film when you see berber-carpeted steps in the background, and piano strings even at medium focal lengths. And the GH4’s Achille’s heel – literally the reason I’m heading back to Sony after a year of sadly building my killer Micro Four Thirds (MFT) lens collection – is its small sensor size which will never change, resulting in a high noise floor and minimal focus isolation capability. That forced my hand into procuring crazy wide MFT distorted glass, like the Voigtländer 10.5mm f/0.95 hyperprime lens used on the hands of the pianist, just to blur out the background.

As for the film itself, it feels like nothing ever created before. (Note by Note and Pianomania are showing their age while never having been cinematic to begin with, and American Grand was literally not meant to be a comedy, even though it’s hilarious.) Not being a braggart, it all just shows how indie filmmaking (empowered with this technology) can bypass conventional legacies of production, from fundraising to crew-building to union restrictions. You can just do it yourself. One of my greatest joys running these FocusPulling communities as a hobby is to see all the amazing work that you send into the user group streams. I can’t wait to see what you create with these cameras, and whatever new toys we’ll get conned into buying. I’m looking forward to all of it!