UPDATE: JUST TO BE CLEAR, THERE IS NOTHING WRONG WITH SONY’S 12 BIT LINEAR RAW. BUT YOU REALLY SHOULD BE AWARE OF IT’S LIMITATIONS COMPARED TO 16 BIT RAW OR POSSIBLY EVEN 10 BIT LOG.
This came up in the comments today and it’s something that I get asked about quite a lot.
Sony’s high end cameras, designed for raw – F5, F55, F65 all use 16 bit linear data. This linear data contains an impressively large amount of picture information across the entire range from the darkest shadows to the brightest highlights. This huge amount of data gives footage that can be pushed and pulled in post all over the place. 16 bit raw gives you 65,536 discreet values.
The FS7 and FS5 use 12 bit linear raw. 12 bit data gives you 4096 discreet values, 1/15th of the values, a small fraction of what 16 bit has. This presents a problem as to record 14 stops with linear data you need more than 12 bits.
Not Enough Code Values.
There just aren’t enough code values with only 12 bits (which is why no one else does it). So Sony do some clever math to make it workable. This reduces the amount of tonal steps in in the shadows. On it’s own this isn’t a huge problem, just make sure you expose brightly to avoid trying to pull to much info out of the shadows and definitely don’t use it for low light. On high key scenes 12 bit raw is very nice indeed, this is where it excels. On low key scenes it can appear very grainy, noisy and shadows often look coarse and lack smooth textures. Expose nice and bright and you will get great highly gradable footage. Expose dark and you will have big problems.
Transcoding can add to the problems.
Where you really can run into problems is if you take 12 bit raw (with it’s reduced shadow data) and convert that to 10 bit log (which has reduced highlight data relative to the scene you are shooting).
What you end up with is 10 bit log with reduced shadow data compared to a straight 10 bit log recording. If you compare the direct 10 bit S-log from an FS7 (or F5/F55) to 10 bit S-log derived from 12 bit raw from an FS5, the FS7 picture will have a little more shadow information while the highlights from both will be similar. So the 10 bit direct log recording from an FS7 will typically be a little better than the 12 bit raw derived log from an FS5 in the shadows (but the raw derived 10bit log will always be better than 8 bit log).
If 4K S-Log is really important to you – get an FS7, F5 or F55.
So I’d much rather have an FS7 (F5 or F55) if I want to shoot UHD or 4K S-log. That’s what these cameras are designed for. But, if you only have an FS5, the raw to log workflow will outperform the limited 8 bit UHD log, so it is still definitely beneficial for FS5 owners to shoot raw and convert it to 10 bit S-Log with an external recorder. But better still record raw, then you really will have a better image.
Raw with the FS7.
On the FS7 the benefits of recording 12 bit raw over 10 bit S-Log are less clear. For bright, well exposed scenes the 12 bit raw will have a definite edge. But the files you will create if using an Atomos or Convergent Design recorder are huge compared to the compressed internal recording and that needs to be considered. For low key or under exposed scenes there is no benefit to shooting 12 bit raw you will get nothing extra.
On the FS7 it is not a good idea to take the 12 bit raw output and record it as 10 bit S-Log on an external recorder. While you may have a less compressed codec, you will be compromising the shadows compared to the cameras own internally generated 10 bit log recordings. In most cases you would be better off simply taking the HDMI output and recording that as it avoids the 12 bit linear shadow bottleneck.
Again though – exposing nice and bright is the key to a good result. Get the data up into the brighter parts of the recording and the raw can be fantastic.
Internal and external log brightness shifts.
When you record S-Log internally on the Sony cameras the recordings use full range data levels to maximise the codec performance. You can use data range (which exceeds the normal video range) as it is assumed the data will be graded and as part of this process restored to video range data for viewing. However when recording on an external recorder the recordings sometimes use full video range rather than data range or if it’s data range don’t have the right metadata. This shouldn’t be a huge problem if the grading software behaves itself and treats each type of content correctly, shifting each to one unified range, but sadly this is rarely the case (especially with Adobe). So not only do the internal and externally recorded images come out with different brightness and contrast, but also LUT’s designed for one don’t work the same with the other. It’s a bit of a minefield to be honest and one of the reason why I prefer to always grade with dedicated grading software like resolve which handles the levels conversions properly (most of the time at least).