So there I was with my brand new FS5. The size and ergonomics of the camera are fantastic, the fact that it uses inexpensive media was more than welcome, I really enjoy working with it but…
…yes, there’s a but:
In spite of the 4K logo on the side, it’s actually a way better HD camera.
“No….” you will try to argue, “it has that one UHD mode…”.
Yes, it has a UHD mode, but it’s 100 mbit/sec and limited to 4:2:0 color subsampling at 8 bits color depth. At least from my point of view, this is a restriction that jettisons the FS5 out of the UHD video camera class. As long as we stick to Rec709 or Cinegamma at 3840 x 2160, the internally recorded image is bearable, but once we switch to a “full dynamic range gamma” (S-Log), this is no longer the case.
At the moment, the only way to output a halfway decent, grade-able image is to use an external recorder. Even though the bit depth of the FS5 HDMI output is limited to 8-bit (regardless the resolution), the higher color subsampling of 4:2:2 and the lack of image compression delivers a recognizably better image that also offers more options for grading.
Image quality advantage:
The Atomos Shogun can be connected to the FS5s uncompressed HDMI port and records the signal in the significantly less compressed PRORES HQ format. 850 mbit/sec 4:2:2 vs. 100 mbit/sec 4:2:0 makes a bit of of a difference. Not that it changes the image of the camera significantly at first sight, but the material just looks more vivid and “real”. A bit of pixel-peeping shows that even unimportant things in the background are just represented with a lot more image information:
The measurement advantage:
The FS5 offers zebra and histogram to help us evaluate the image. That’s really not a lot, considering my old Canon XF305 from 2010 was already able to display additional measurement tools like waveforms, RGB parade or a vectorscope.
With the Shogun, all of those tools are available at the touch of the screen (yes, it has a touchscreen). Even if we – for whatever reason – prefer to record internally on the FS5, we still have better tools for image evaluation available.
The LUT advantage:
From the F55 down to the FS7, it’s possible to load custom LUTs (look-up tables), which are a tremendous help with cameras recording in a log format. LUTs are dropped onto the (recorded) S-Log image in our viewfinder to show us how the image could look once it’s transformed back into rec709 space. While colorists surely have a lot more options in their digital toolbox than just slapping a simple LUT on the S-Log image in post, those on-camera LUTs help us to evaluate the “milky-shallow” S-Log image we’re recording.
In comparison to its bigger brothers, the FS5 only has a feature called “gamma display assist” that adds contrast to the viewed S-Log image in order to give us a bit of “wysiwyg”-direction.
Via the Shogun, we have the ability to add (up to eight) real LUTs to the on-screen Log image. Apart from “general purpose” factory-installed Log-LUTs, any .cube LUT can be transferred by simply copying it from the SSD to the Shoguns internal memory.
That way, we can also “rate” the camera at ISOs other than the standard ISO 3200 sensitivity: For example, by applying the “XDCAM_Slog3-Film1-2over.cube” LUT that can be downloaded on the xdcam-user website, we can rate the camera at ISO 800 and get away with less image noise. And via such a LUT, we also have a much better interpretation of a possible final image than just guessing that the shallow mush we’re viewing on-camera is actually what we want.
The audio advantage:
The FS5 has 2 audio inputs via XLR for line / mic / mic +48V phantom power (PP). That’s quite a difference to the usual mini-jack input we can find on the good old 5D Mark III or similar DSLR or mirrorless variants. Well, this one’s a real camera.
The FS5 also has a built-in “quasi” stereo-microphone (I’m also coming from a solid audio engineering background so I know what a true stereo microphone setup should be). Even though it’s quasi-stereo, it would be absolutely sufficient for some “effects” (let’s call it “ambience”) pickup to underline b-roll images.
My preferred setup for documentary audio should be something like that (in the best of all cases including an audio assistant with 2-channel mixer and a long break-away cable, but also works in single operator mode):
- One lavalier microphone or shotgun pickup of any talking person / interview partner to have a really close sound source for the mix.
- One shotgun-microphone (on- or off-camera) for safety reasons or second lavalier mic for conversations.
- Some type of quasi-stereo pick-up for room / ambience / “effect” sound that can be used to add “reality” to an interview or a b-roll cut.
1..2…3…4… it’s clear to see that this setup requires a minimum of four individual audio tracks. While offering four audio sources (XLR 1+2 & built in mic L+R) and also recording in a format (XAVC-L) that offers 4 audio tracks, the FS5 can only allocate two audio sources at the same time.
The Atomos Shogun features 2 additional XLR audio inputs, accepting line and mic signals and even offering +48V PP (globally, meaning for both inputs at the same time). These two tracks will be recorded alongside whatever you are picking up via the cameras own inputs. And there are also two outputs for the audio guy with the break-away cable:
The timecode sync advantage:
The FS5 has no external LTC timecode input. But I go by the opinion that solid timecode sync is way more reliable than syncing via audio in a multi-camera setup.
The reason is simple: When syncing by audio, the sound sources will often be located at a different distance to the individual cameras and sound travels rather slow in comparison to light (340.29 m/sec at sea level). Let’s say we’re filming at 25 fps and the cameras are more than 14 meters apart, the offset between them is already more than one frame (unless the sound source is straight in the centre between the two cams).
The only chance to sync timecode between the FS5 and other cameras is to gather all of them cam-ops together and let everyone reset their free-run TC at the count of three. A common practice that will result in 1-3 frames offset in the beginning and will deviate more during a long day of filming.
The only way to lock timecode between multiple independent systems is to distribute a master timecode (or even better – master timecode together with a master clock / genlock sync). In the old days, this was done via BNC cable, in modern times, devices like the incredibly versatile Timecode Systems miniTRX+ distribute wireless timecode and sync.
I have been using this system on a multi-cam documentary production with external sound recently and experience shows that this is one of the most reliable systems I have ever seen.
Shifting our focus back on the Shogun, I guess it’s worth mentioning that the device “talks LTC timecode” since firmware 6.5 via the genlock input (right next to the SDI I/O) at the back. If you connect any LTC SMPTE timecode signal and choose the right options in the timecode menu, the clips recorded via the Shogun will be stamped with the incoming timecode.
From my point of view, that is one of the most radical features that open new production possibilities to the little Sony FS5 (or any other camera without timecode I/O).
The Shogun accepts regular Sony NP-F type batteries and the boxed set even includes one. I prefer to power the Shogun and the camera from the same battery. With a custom cable like my Hirose > Hirose & barrel connector it’s not much of a problem to power my audio receiver and the Shogun together with the camera via Sony BP-U60T or the Hawk Woods BP-75UX. I tested the performance and after 2 hours of continuous recording, there was still 10% charge left on the BP-75UX.
The Shogun package comes with convenient cradles to house SSDs. The Atomos website hosts a document that lists a number of compatible models. My choice of SSDs are the SanDisk Ultra II 950 GB models that come for a price less than 200 € each, giving us more than 2 hours of 850 mbps PRORES HQ UHD recording (in comparison, a 128GB Sony XQD card will cost about double the price). So far, these SSDs delivered a problem-free, rock-solid performance.
SDI for 10-bit HD & future FS-RAW:
As mentioned in the previous blog post, the FS5 HDMI out is limited to 8-bit color depth at the moment. Luckily, the camera also features a 3G SDI output via BNC that delivers uncompressed 4:2:2 10-bit (only in HD resolution).
The Shogun features both SDI & HDMI connectivity and it already supports FS-RAW via SDI. Sony promised to enable FS-RAW via SDI on the FS5 in a future (paid?) firmware upgrade, so the Shogun can also be seen as a future-proof option for RAW recording.
Nothing’s perfect or “details to consider”:
Fan noise: The Shogun has a built-in fan for active cooling that is about 4-6 dB louder than the camera fan (measured directly on top of the FS5). While it was never an issue on my recordings, fan noise might be noticeable in very silent environments.
Reflective screen: The LCD touchscreen of the Shogun has a remarkable image quality, but in case we’re planning to use this device for exterior shots on a bright, sunny day, it’s almost impossible to use it without appropriate shades. The Porta Brace rain / dust protective cover may not have the perfect fit, but definitely offers decent protection:
Audio latency: I mentioned that the Shogun adds two additional inputs of audio to any camera. Unfortunately, it can’t be avoided that it takes more time for the camera to convert the signal to SDI / HDMI and pass it to the recorder than it takes for the Shogun A/D-converters to process the audio signals going directly into the unit.
As a result, a 2 (HD) or 3 (UHD) frame offset exists between the Shogun audio and the camera stream (video & audio).
The Atomos developers did in fact consider this unavoidable fact and provided the ability to add an incremental frame offset to the Shogun menu options. This way, the audio signals going into the Shogun can be aligned with the video delivered by the camera. So far, so good…
Unfortunately, the offset affects ALL audio signals (including the ones from the camera) and not only the ones going directly into the Shogun at the moment.
Workaround: If we want to record four channels of audio with the FS5 / Shogun combo, we should skip the audio offset function and shift the Shogun audio tracks in post. If, for any reason, we want to skip camera audio and only use the Shogun inputs, the audio offset option should be used of course.
Video / Audio latency & timecode: The same as above also applies when using the Shogun timecode sync option. The incoming timecode will be as exact as possible, but the signal from the camera will lag behind 2 – 3 frames. Thankfully, this is a constant and linear offset that shouldn’t get smaller or bigger over time.
Workaround: Shift the camera track accordingly on the timeline. It would be really nice to have a frame delay for the timecode in order to align it with the SDI / HDMI conversion delay. I have already contacted both Timecode Systems and Atomos and they passed this feature request on to the devs.
Especially for the new (lowered) price, introduced at the beginning of February, the Atomos Shogun is an incredible accessory for the Sony FS5 and adds a ton of professional features that the camera is missing without it. Mounted on the back of the camera body with a Manfrotto 244MICRO arm, I never had the feeling that it added up too much to the cameras sleek appearance or hinder its performance in the field.
Disclaimer: Please note that I am not affiliated with Atomos in any way and I am also aware that other recorder/monitors like the Convergent Design Odyssey 7Q+ exist. I just happen to own and use a Shogun to overcome FS5 limitations.
It’s been over a month now since the purchase of my FS5. In the meantime, I spent hours on testing the camera and I had the possibility to film a number of commercial productions with it.
Every day, I read a number of posts on forums and it seems that the same topics appear over and over again. So I thought it might be a good moment to clear up a few things about this camera:
Bugs that will hopefully be addressed in the next firmware update:
(scheduled end of February 2016)
- Due to the temporal internal noise reduction that enables us to use Cine1 even at ISO 6400, ripped edges and “noise lines” that make the image fall apart may occur with the camera in motion, especially in UHD mode.
Possible Workaround: Avoid filming scenarios with a higher gain than +9 dB.
- When using the kit lens (Sony SELP18105G), zooming in or out using the zoom rocker of the smart grip may knock the image out of focus. Note that the SELP18105 is “electronically parfocal” (meaning even though it is not optically parfocal we can use the old ENG practice of zooming in, setting focus and then zooming out again because the lens automatically adjusts for focus differences at the individual zoom stages).
Possible workaround: Use the zoom controls on the lens.
- Internal recording capabilities: The camera uses the XAVC-L codec in all modes. This is a inter-frame / Long GOP codec. By the nature of this codec (and in contrast to any intra-frame codec), not every single frame is encoded individually. This means it will put more stress on our editing workstation than any intra-frame codec.
XAVC-L in HD (1920 x 1080) is recorded with a bitrate of 50 mbps, 4:2:2 color subsampling and 10-bit color depth.
XAVC-L in UHD (3840 x 2160) is recorded with a bitrate of 100 mbps, 4:2:0 color subsampling and 8-bit color depth.
Possible workaround: Use an external recorder connected either to the HDMI or SDI output.
- HDMI output: The HDMI output delivers an image at any resolution mode of the camera (meaning up to 3840 x 2160) with a color subsampling of 4:2:2 but only at a color depth of 8-bit. This is resolution-independent (meaning it’s the same 8 bits at HD or UHD), most probably a hardware restriction and thus most probably won’t be fixed in any firmware update. Don’t take my word for it though, in case I’m wrong and Sony offers us 10-bit HDMI out (like the cheapo GH4s and XC10s have) via a firmware update, I’d be the last one to complain.
Possible workaround: Use the SDI output to get 2 bits of additional color depth.
- SDI output: The 3G SDI output delivers an image up to HD (1920 x 1080) with a color subsampling of 4:2:2 and a color depth of 10-bit. In case we want to record or monitor externally while in UHD mode the SDI output will still only deliver the image at 1920 x 1080.
A future firmware update should give the FS5 the possibility to output (linear 12-bit) FS-RAW which can be recorded via an external recorder (like the Atomos Shogun or the Convergent Design Odyssey7Q+).
- UHD monitoring choice: Only when recording in UHD mode, due to the FS5 interal processing limitations we can choose to either
monitor the image on the built-in EVF / LCD and record internally
monitor the image externally (via HDMI or SDI) and record both internally and to an external recorder – in this case, the built-in EVF / LCD will only display useful HUD-data like the histogram, time information etc. but no actual image.
Sony promised that the camera will be able to output both internally and externally when outputting FS-RAW.
- Will a Shogun or Odyssey 7Q+ enhance the image quality?
From my experience, yes. Even though the color depth is limited to 8 bits in UHD, the image looks way better than the internal recording, thanks to less compression artifacts and higher color subsampling. I am using the Atomos Shogun which accepts both the SDI signal (via BNC) and the HDMI signal (both up to 60P). It already “talks” FS-RAW, so once the option arrives, I am also ready for that.
Apart from that we also get on-screen waveform monitoring, additional audio inputs and a really great screen.
- Battery issues: Up front, I need to say that we’re best off with original Sony batteries. The company tries to protect their own battery market by finding ways to block out third party manufacturers – it’s a lot of money in that market segment, so I can’t blame them.
Many of the third party batteries (compatible with the FS7!) I tried cause strange problems: Either we’re unable to format the SDXC cards as long as the percentage of a freshly charged battery isn’t shown on the FS5 display, or the camera resets the internal clock and timecode when powering it on after inserting a third party battery. Out of my experience, this doesn’t happen every time (very random), but it happens frequently.
Now I’m that kind of guy that doesn’t like to collect batteries but rather power all my peripherals (like headlight, audio receiver and even the Shogun) from a single battery.
Sony offers the BP-U60T that comes with an additional Hirose output for such a task, but unfortunately, that type of battery is out of stock and new ones are expected to ship in late March 2016.
The only working third party battery that I could find on the market was the Hawk-Woods BP-75UX which also offers two (!) additional D-Tap outputs. The same company also builds D-Tap to single or dual Hirose adapters. Absolutely great!
- Issues with Final Cut Pro X: While I also have Adobe Premiere on my main workstation for compatibility reasons, my personal choice for editing is FCPX. While we can discuss days and nights about which NLE is the best I know I’m way faster there.But two problems exist at the moment (as of beginning of February 2016):- The luma values of the FS5 UHD clips is interpreted wrongly, so directly importing the clips into FCPX will result in a strange luma shift.- With HD clips imported directly from the camera, random frames “block up” during playback. Unfortunately, this also affects the render engine, so even if the rendered timeline plays back without any problems, certain “blocky” frames may creep into our final render. Needless to say, every final render should be checked for errors, but spotting those single, random “blocked up” frames requires a lot of concentration.
By the way, this is no FS5-specific issue. I have seen this issue with other more “recent” Long GOP codecs as well, like the one the Canon XC10 is using.
Apple is aware of the problem and they’re currently working on it – I hope a fix doesn’t take them too long.
Possible workaround: Batch-transcode the clips directly to PRORES before edit. One of the quick and reliable solutions to both of the problems mentioned is Convergent Media EditReady. I use it every day.
- S-Log 2 / 3 in UHD mode: Recording S-Log 2 or 3 internally will most probably give you banding, and there is no way to avoid that. I am afraid it originates more in the heavy image compression than in the lack of 10 bit color depth, even blue sky banding didn’t show up when using the Shogun external recorder: