FS5 - Raw sensor data and "native" ISO

[Cliff Totten]

We all know the FS5 will get it's optional raw sensor data output upgrade very soon. This leads me to think about a couple of questions:

We know Sony sets a minimum ISO setting for all it's cameras with SLOG. Example; RX10-II = 800ISO, A7s-II = 1600 ISO, FS7 = 2000 ISO and the FS5 = 3200 ISO.

We also know that Sony adjusts and changes these ISO values as they see fit: Example: A7s originally at 3200 was lowered to 1600 ISO in the MK II and the FS700's original minimum was changed to 2000 ISO with its firmware upgrade.

But what about raw sensor data output? If the FS5 will output "true" raw sensor data, this will bypass any and all down stream Sony gain processing. If this is the case, wont this mean that the "RECORDER" will control the "minimum ISO" mapping from raw sensor data? (Especially if you are recording raw into ProRes)

Is that right? The recorder takes in the raw pixel voltage readings, deBayers the RGGB sensor matrix and adds the color. And, of course, then applies whatever gain/ISO that YOU want.

Wouldn't you then be able to process an image that is LOWER than what Sony forces on it's in-camera image processing?

So, would it be possible to operate the FS5, recording ProRes from raw into a Shogun's SLOG gamma curve at just 1600 ISO if you want?

Interesting.

[Chad Johnson]

It's all too complicated to think about on the FS5 when using ISO. On this camera the base ISO is just an arbitrary (maybe not exactly arbitrary) number used to denote 0db gain. If you set the camera to display gain rather than ISO, your camera will always be at the lowest possible ISO when you set 0db gain, without varying numbers to confuse us. This is how big boy cameras display their gain - with gain readings. So you can set the 3 gain level switches to 0db, 6db, and 12,db so you can quickly switch up an extra stop or two of gain on the fly, and know that no matter what PP you are using, when you see "0db gain" you are at the lowest possible gain/ISO for the cleanest image as far as gain goes.

[Doug Jensen]

+1
Well said.

[Cliff Totten]

Originally Posted by Chad Johnson

It's all too complicated to think about on the FS5 when using ISO. On this camera the base ISO is just an arbitrary (maybe not exactly arbitrary) number used to denote 0db gain. If you set the camera to display gain rather than ISO, your camera will always be at the lowest possible ISO when you set 0db gain, without varying numbers to confuse us. This is how big boy cameras display their gain - with gain readings. So you can set the 3 gain level switches to 0db, 6db, and 12,db so you can quickly switch up an extra stop or two of gain on the fly, and know that no matter what PP you are using, when you see "0db gain" you are at the lowest possible gain/ISO for the cleanest image as far as gain goes.

Agreed.

I have an FS700R that I use with my Shogun. (12bit raw sensor data converted to 10bit ProRes)

As I understand it, "true" raw sensor data basically contains this data for each pixel:

1.) Pixel geographical position on the sensor. (The spot where that pixel data is recorded)
2.) The color value that needs to be assigned by the receiving devise. (A Red, Blue or Green value for that particular pixel)
3.) Voltage reading for that pixel. (It's 12bit depth intensity amount measured by the camera's read out circuit)

I was told by Atomos Support that when you add "gain" on the camera, that doesn't change any of the actual pixel data sent into the raw stream. However, it WILL send a meta-data "flag" to tell the receiving devise to add gain to the image AFTER it gets there.

So, the Atomos Shogun will take this 12bit raw data, drop 2 bits, process the RGGB data, de-Bayer it, convert it into to actual "VIDEO" and then add whatever gain amount that the stream flags tell the recorder to do then send that into ProRes.

If this really is the case, than the Shogun will need to amplify the signal to whatever it needs to create an SLOG-3 gamma curve into ProRes and it's display monitor.

I "think" it's the raw data receiver/ recorder's job to add enough gain to "bend" the linear data and "squeeze" it into a "curve". (I "think" you need to amplify it to make that happen)

I'm not 100% sure about the exact process I have stated above. However, I'd LOVE for someone with more intimate knowledge of Sony FS raw sensor data to chime in on this. Any Atomos or Convergent Design people out there?

Wow,..that is pretty intense!