FS7: 2 Days with the Sony PXW-FS7: Early Impressions

[Lenilenapi]

There is a learning curve to using CINE EI. I suggest reading Alister Cghapman's tutorial on his website and probably Doug's manual/video is worth getting. A few points re your issues:

-FS7 settings for gain or ISO are actually independent of each other which took me a while to grasp. The ISO settings also vary with each gamma. So you may have jot set and then you switch gammas and the ISO switches now read different. If you stick with gain it stays the same was I recall .
- CINE EI will always record at 2000 ISO regardless of what the gain switch says - so that's confusing. It is only active if you are using a LUT in which case it only affects the LUT itself and not the underlying SLOG recording - UNLESS - you have it set to record the LUT itself in which case the ISO setting is what you will record.
- SLOG is slow . Traditional 709 video has lots of gain and noise reduction that enables expanding a smaller DR to cover 0-100 IRE. So you can get clean images with fill exposure at 3200 ISO. However in SLOG that lower exposure is no longer being expanded to 100 IRE and so your only getting a partial exposure at low levels. Compounding that Sony in its wisdom has turned off noise reduction in CINE EI and 2000 ISO is noisy unless you use noise reduction in POST. Most people therefore shoot at 1000 ISO or lower. Hopefully they'll fix that in a firmware upgrade.

Lenny

 

[basspig]

It IS very confusing, this dichotomy between ISO and gain. I, too, thought they were the same thing by two different names. But maybe it's more akin to F-stop versus T-stop on a lens, where the first is the aperture and the second is the light transmission factor. Hopefully, someone may explain that in a youtube or a blog site.

I have Alister's page on CINE EI captured to disc and am slowly digesting it. It is confusing to look at from the context of a REC 709 shooter. It's almost a necessity to forget what I've learned about the old and don't add confusion by referencing everything to the old way of thinking.

I think I understand why S-Log 3 is said to look noisy. I record the 8-bit HDMI to my Shogun (to get Premiere compatible ProRes) and I also ready Sony's whitepaper on S-Log. S-Log 3 encompasses 16 f-stops and with a 14-stop sensor, it only uses about 2/3 of the grey levels. In 10-bit that's not a problem, but in 8-bit, the number of steps is only about 140, as the upper steps are not reached by the sensor output. That means each step is coarser than if the signal were filling the whole 8-bit 256 level range. Any noise in the signal will be a greater percentage of the luma and chroma values available and thus appear accentuated as compared with 10-bit.

S-Log 2 has an advantage if you're limited to 8-bit recording, because it uses the entire range of digital levels. Yes, you lose 1/2 and F-stop because the curve is clipped at 13.5 f-stops at the 1024 bit level, but the noise is not nearly so accentuated because there are more grey steps to define it and its contrast is not accentuated. A good corollary to this would be looking at noisy footage at 1/2 resolution vs. full resolution in the NLE playback. The 1/2 resolution really brings out the noise and makes it looks horrendous, but at full resolution, it takes on a fine, filmic grain appearance. I find that my S-Log 2 footage has far less noise--in many cases, no visible noise at all. I shot some scenes yesterday where the sun was reflected off a white wall and some white gravel on a walkway. A few feet away was brown earth and green moss, under the shade of the trees. Above is blue sky with white, billowy clouds. Everything in the picture is properly exposed with 3 dimensional detail in the clouds, texture of the white gravel fully rendered, while the brown earth and green moss are clean and free of visible noise. The caveate is that I have to turn up the monitor backlight to 100 (I normally use it at 10) to begin to see this scene in any facsimile of near reality. I can imagine it would only be even better with 10-bit recording.

In short, S-log 3 almost requires 10-bit recording, because it uses a limited part of the grey levels so the full resolution of the encoding hardware can never be used (short of having a 16 f-stop sensor). S-Log 2 hits the 1024th grey level at 13.5 stops and thus fills the entire range of the encoding hardware.