Shooting Raw With The FS5 And Ninja V+

This came up in one of the user groups today and I thought I would repeat the information here.

One issue when using the Atomos Ninja V+ rather than an older Atomos Shogun or Inferno is that the Ninja V+ doesn’t have an internal S-Log2 option. This seems to cause some users a bit of confusion as most are aware that for the best results the FS5 MUST to be set to PP7 and S-Log2 as this is the only setting that fully optimises the sensor setup.
 
When you shoot raw, you are recording linear raw, the recordings don’t actually have any gamma as such and they are not S-log2 or S-Log3, they are raw. The S-Log2 setting in the FS5 just ensures the camera is optimised correctly. If you use the S-Log3 settings, what you record is exactly the same – linear raw, just with more noise because the camera isn’t as well optimised.
 
Any monitor or post production S-Log2 or S-Log3 settings are simply selecting the intermediate gamma that the raw will be converted to for convenience. So when the Ninja V+ states S-Log3 this is simply what the Ninja converts the raw to, before applying any LUT’s. It doesn’t matter that this is not S-Log2 because you didn’t record S-Log2, you recorded linear raw. This is simply what the Ninja V+ will use internally when processing the raw.
 
You have to convert the raw to some sort of gamma so that you can add LUT’s to it and as S-Log3 LUT’s are commonly available S-Log3 is a reasonable choice. With earlier recorders you had the option to choose S-Log2 so that when viewing the native S-Log2 output from the camera, what you saw on the monitors screen looked similar to what you saw on the FS5’s LCD screen when the FS5 was set to S-Log2. But S-Log2 is no longer included in the latest monitors, so now you only have the option to use S-Log3. But from an image quality point of view this monitor setting makes no difference and has no effect on what is recorded (the FS5 should still be set to PP7 and S-Log2).
 
In post production in the past, for consistency it would have been normal to decode the raw to S-Log2 so that everything match throughout your production pipeline from camera to post. But again, it doesn’t really matter if you now decode the raw to S-Log3 instead if you wish. There will be no significant quality difference and there is a wider range of S-Log3 LUT’s to choose from.
 
If the footage is too noisy then it is under exposed, it’s the only reason why the footage will be excessively noisy. It is true that raw bypasses the majority of the cameras internal noise reduction processes, but this only makes a small difference to the overall noise levels. 

Even with the latest Ninja V+ what is recorded when outputting raw from the FS5 is 12 bit linear raw.
 
12 bit Linear raw is a somewhat restricted format. 12 bits is not a lot of code values to record a high dynamic range linear signal. This is why most cameras use log recording for wide dynamic ranges, log is much more efficient and distributes the available recording data in a way very sympathetic to the way human vision works.
 
In practice what this means is that the 12 bit linear raw has LOTS of data and image information in the upper mid range and the very brightest  highlights. But relatively very little picture information in the lower mid range and shadows. So if it is even the slightest bit under exposed the image will degrade very quickly as for each stop yu go down in brightness you halve the amount of image information you have.
 
In an underexposed image the noise will be very coarse in appearance and the image will be difficult to grade. You really do need to expose the raw nice and bright and because of the way the data is distributed, the brighter you can get away with the better. Never be afraid of exposing linear raw “just a little bit brighter”. It is unlikely to severely bite you if you are over exposed but highly likely to bite you if it is even a fraction under.
 
The 12 bit linear raw from the FS5 is not going to be good in low light or when shooting scenes with large shadow areas unless you can expose bright so that you are bring your levels down in post. If you have to bring any levels up in post the image will not be good.
 
Raw is not a magic bullet that makes everything look great. Just as with S-Log it must be exposed carefully and 12 bit linear raw is particularly unforgiving – but when exposed well it is much better than the internal 8 bit log recordings of the FS5 and can be a fantastic format to work with, especially given the low cost of an FS5.
 
I recommend going back to basics and using a white card to measure the exposure. If monitoring the raw via S-Log3 the white card needs to be exposed around 70%. If using a light meter with the FS5 set the light meter to 640 ISO.
 
If you do want to use a LUT on the Ninja to judge your exposure use a LUT with a -1.5 stop offset. The darker LUT will encourage you to expose the raw brighter and you will find the footage much easier to grade. But it should also be considered that it is also quite normal to add a small amount of selective noise reduction in post production when shooting raw. 
460x150_xdcam_150dpi Shooting Raw With The FS5 And Ninja V+

15 thoughts on “Shooting Raw With The FS5 And Ninja V+”

  1. Trying to understand the logic to the Ninja V scales with my Sony devices.
    Related to PQ mode on the Ninja V why is the top of the scale set to 1600nits specifically? Using RAW with A7SIII/A1 clipping shows up around the 1300nit mark according to the waveform (which matches up with Zebras clipping at 94+ on the camera). This being the limit of the sensors I assume.
    It get wonky when using parade waveform. The Ninja will show Red clipping at 800nits with green and blue now extending to 1600nits.

    Sometimes I see the Ninja waveforms (in parade especially) start from below 0 which seems broken? Is there accuracy in what I am seeing?

    Thanks for any insight.

    1. 1600 NITs is the full brightness range of the screen. Nits is not a good scale to use for exposure as there is no well defined relationship between captured range and display range, so a lot of the numbers are arbitrary based on what Atomos thinks looks good on their screen and how they chose to decode and display the S-Log3 or raw signal.

      Zebras on the camera clip at 94% because this is where +6 stops is on the S-Log3 gamma curve and the A7S3 and A1 both record at up to +6 stops over middle grey.

      When viewing the Luma (Y) channel you are viewing red green and blue combined, so the luma value is the average of the R, G and B values. With a raw workflow some of the white balance adjustment may be done by the recorder so the RGB parade values may be different to the output luma (Y) values depending on the white balance settings.

      1. Alistair, you say 1600nits is the full brightness range of the screen? I have only seen Atomos spec this screen at 1000nits. Or are you saying 1600 is peak brightness but full screen brightness is 1000?

        1. Sorry my mistake. You are correct that the screens maximum brightness is 1000 Nits. 1600 Nits appears to be the range chosen for the waveform as this corresponds to what the very best HDR displays may be able to show. It is still a poor choice for exposure measurements as screen brightness and capture range are two different things, not necessarily directly related.

  2. Hi Alister.

    Hi.

    I have a SONY FS5 with the RAW licence activated going into a NINJA V+ via SDI with the ATOMX SDI module. All licences have been activated on the NINJA V+

    I don’t seem to be able to select anything other than 4K PRORES RAW when bringing in an SDI raw signal from My sony FS5. If I try to select PRORES HQ it reverts back to PRORES RAW every time – With a dialogue box explaining that a RAW signal has been detected and PRORES RAW codec is needed to see it etc. I have tried this on a SHOGAN 7 with no trouble RECORDING 4K PRORES HQ etc. It seems that I only have PRORES as an option when inputting a standard HD signal via the SDI.

    I was hoping to be able to record 4K PRORES as PRORES RAW is a huge file (though the colours are super impressive!)

    Also – am i supposed to be able to record 4k H.265? again this seems only to be a function available in HD.

    Thanks for your help

    Ben

    1. This is how it works on the Ninja V+, there is currently no option to transcode the raw to other formats. But regular ProRes Raw is only marginally larger than ProRes HQ and depending on the scene content can in some cases be smaller.

      You should be able to do H265 in 4K via the HDMI.

      1. Thanks for clearing that up. And yes I feel it’s worth doing on this camera while I’m still using it. It has certainly helped in terms of being able to use the FS5 as a C cam on FX6 shoots (just)

  3. Is it possible a firmware update from atomos will add slog2 support in the V+? I think this is not hardware related.

  4. Could this be possibly solved by a firmware update from atomos? Feels like slog2 support should not be a hard thing.

    1. If you are shooting raw it doesn’t really make any difference as raw is raw, it is neither S-Log2 or S-Log3, it is raw.

  5. Hey Alister, thank you again so much for the very useful information here! You mentioned getting 10 bit prores from 12 bit prores raw, which is compressed from the camera’s 16 bit raw data (does FS5 sensor output 16 bit at the upstream?) can be extremely worse than the internal 10 bit xavc due to the details loss in the shades that happens during both transitions (16 bit raw to 12 bit prores raw, and then 12 bit prores raw to 10 but prores). My question is that is the 12 bit prores raw a log raw instead of a linear raw? And the 10 bit prores is also log encoded from the 12 bit raw? If that’s the case, I can expect there may not be a vast disadvantage to record 10 bit prores from 12 bit prores raw from camera such as sony fs5 through sdi?

    Again, I really appreciate you sharing your invaluable knowledge here so we can all learn from you!

    Best regards,

    Wei

    1. Most of these cameras have a sensor with a 12 bit A to D converter, so the signal starts at 12 bit. BUT whenever you add compression you introduce noise artefacts and other processing errors that will result in new output values that sit in between the original 12 bit values. Recording at a greater bit depth than the sensors A to D helps preserve these new values which lessens any image degradation caused by the added compression.

      The FS5 outputs a 12 bit linear signal and externally this is compressed and recorded as 12 bit linear. This is not an ideal situation as the inefficient compression at 12 bit linear will degrade the image data prior to you having to do all the post production processing to turn it into a colour image. Any compression artefacts will affect the quality of the processing. Then converting this from 12 bit linear raw to 10 bit ProRes log will add another layer of artefacts on top of all the previous processing artefacts. But the FS5 can’t output more than 8 bit in 4K, so this will be as good as can be had from an FS5, but it won’t be better than the 10 bit direct output from an FS7 where all the processing is done before any compression, so will not have all the compression and conversion artefacts.

      The current cameras output a 16 bit signal which is then compressed down to 12 bit log. This is a much more efficient way to compress the data so it will have fewer artefacts that will cause image degradation during the processing needed to create a colour image. This is a far superior pipeline from the sensor to post than the very inefficient 12 bit linear of the FS5. But as the newer cameras can directly output 10 bit 4K log (again all the processing is done prior to compression directly on the original sensor output) the difference between the raw and the XAVC or a normal ProRes recording is so small that most people will never ever see any difference.

Leave a Reply to alisterchapman Cancel reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.