S-log, cheap 'n easy?
 

Can I shoot S-log video, record it with a Ki-pro mini, and apply a LUT in post? I'd like to be able to shoot a "digital negative", but I won't have access to Sony's new digital recorder. I also won't have access to a post facility with a professional colorist.

No, that's not possible. The S-Log signal feeds out of the Dual HD-SDI ports and is 4:4:4 RGB. The Ki Pro, Nanoflash, and similar devices can only record 4:2:2, and should appropriately be attached to the single SDI spigot on the top right rear of the camera. That port is designed to be used for onset monitoring during S-Log recording, so the LUT can either be displayed or turned off, but the recorded result is baked in.

To use S-Log you would need to purchase the firmware upgrade, rent or buy an appropriate recorder, and then, (after any SxS offline editing) move to a suitable online edit suite.

I suspect if you only occasionally have use for S-Log that renting a recorder may be the most economical option.

I understand--with S-log, there's so much data that it takes dual SDI and a special recorder to handle it. (Under most circumstances, I don't think my clients will want to take that route. Nice to know it's there, though.)

I'm curious: what would constitute a "suitable online edit suite"? What software do you use to apply the LUTs?

And, back to the more primitive world that I live in: even without the S-log feature, a Pro Res file recorded from the SDI output of the camera is going to allow quite a bit of room for color correction, right?

Lingo?
 

One more thing: I'm a little confused about this term, "Baked In". Suddenly, I see it everywhere.
Can anyone offer an exact, technical definition of what this means? "Adjustments applied, file size reduced"? Right now, every time I hear that term, "Baked", I sneeze.

Camera sensors collect a lot more data than is generally possible (or sometimes even desirable) to record. In camcorder mode white balance, knee, gamma curves, gain - these are all internal adjustments to the data after it has been collected by the sensor. For any camera output the data has to be conformed to the specifications of that signal type. For any internal recording media the data has to be compressed to that codec.

Basically, you tell the camera which sensor data you wish to retain, and which to discard. What's recorded to the SxS cards is only a fraction of what the camera sensor saw.

As Alister Chapman has previously pointed out, S Log is not a true uncompressed copy of the camera sensor data - I believe he said there is some highlight compression applied (among other things), but it's the closest this camera can get to it.

So... for your SxS or Nanoflash, you can consider what it records to be "baked" because there is no going back and unapplying the gamma or LUT once it's been recorded to that compressed media.


I own an EX1 which uses XDCAM EX. The footage has an impressive amount of lattitude for it's size. I've also recorded ProRes footage from the 10bit HD-SDI, and that definitely gives you even more room to make adjustments in post. Just not as much as S Log would.

Hi Jim,
The way I think of it, with the HyperGamma settings you can essentially compress about two stops of (overexposed) highlights down into your recordable range. (that is where the hypergamma curves bring 400%+ of video down to 100% or 109% video level.) This can then be color-corrected to massage the data a little more, or you may find you like the look. Remember, though, that anything recorded at above that range is completely gone. High quality pro-res should be fine for this, and presumably SR would be even better for basic color correction.
With S-Log you're getting the full range of the chip, which I interpret to mean up to around 800% video (or one extra f/stop of highlight detail to work with), crushed down into your recording range - using a logarithmic system so that you still have excellent color-correctability in the blacks and mid-tones where it really counts. Obviously, because S-Log gives you a larger range of recorded exposure to work from, you have more room for color correction.
But with S-Log you will absolutely have to do a post-color-correct on the footage, whereas with the Hypergammas you may like the look (or not need to do as comprehensive a color correct.)
If you don't have access to a professional colorist, you will need to do a lot of learning/testing/practicing to get yourself up to speed if you want to use S-Log.
I'm sure that once the F3 camera and dual link option become available and popular, especially since Sony seems to be starting to open up the SR codec at the software level (based on what Peter Crithary said in his Rule presentation, available elsewhere in this forum), we will start to see even more recording options become available.
'Baked In' -- to me that just means that a LUT has been applied to the image, and the image is recorded post-LUT, so there's no way to remove the LUT and get the image recorded on the SxS cards back to the way the S-Log data looked. That's why the image on the SxS cards doesn't look like the S-log data.

Feel free to correct me if I'm not getting it straight...

Thanks for the information.

As a guy who started out shooting large-format Ektachrome, all of this image control is and exciting prospect.

Dave, I think you have summed it up perfectly.

The big difference between 4:4:4 and 4:2:2 is that 4:4:4 is RGB so you have the full S-Log data from each sensor so you can adjust the individual colours to get the exact tonal response that you want, so in effect you can colour balance in post. With 4:2:2 YUV your colour balance is pretty much baked in when you shoot. You can correct and adjust it but there just isn't as much lee-way as with RGB.

I wouldn't describe it like that. While Sony has bundled S-Log and 4:4:4 together (as of now), they are useful independently. The only dependency I can think of that hasn't been discussed is that S-Log isn't that useful if it's recorded to an 8-bit codec. Since S-Log needs to be manipulated later, you need those 10 bits to keep the image from posterizing as you "stretch it back out" with contrast later.

It's going to allow more, yes, but understand a good portion of that magic is in recording to a 10-bit format as opposed to an 8-bit format. This means KiMini as opposed to Nanoflash, at the moment.

S-Log at 8 bit will look terrible with lots of banding. As you have to convert the log recording to near linear for a sensible looking image there will be a lot of pulling and pushing of the image and you must have 10 bit for it to look decent.

So, with a standard F3, I'll be able to record a low contrast scene -- broad tonal scale-- with my Ki Pro Mini through the SDI output, at 10 bits. Apple Final Cut Studio can handle 10-bit?

The key to all of this is the 10-bit SDI output? Right now, if I record a flat, low con, scene through the SDI output of my HPX-500, it would fall apart if I try to correct it much, correct?

As a matter of course, I usually use a "low-contrast" style with my Canon 5D. But I can't seem to correct it much in post, because the video's already been through the meat grinder.

Yes - together with a low noise camera - and that is the bit that's all too often forgotten.

There's no point in 10 bit recording for most cameras because even if the bit depth will permit pushing and pulling without banding, the noise will stop it instead for other reasons. The indications seem to be that the F3 will have a low enough noise level to make 10 bit worthwhile.

I understand. Thanks for clarifying this.

As a part-time colorist, I disagree. Colorists need images not to posterize as we manipulate, no matter where the noise floor/level is from the camera. And it can't be fixed by transcoding 8-bit to 10-bit. No matter how flawed the original signal is, capturing it in 10-bits helps later if you color.

With a noisy camera there is no advantage to 10 bit over 8 bit for acquisition as all the extra bits don't contain any useful extra data. The posterisation is caused by the sensor noise itself or by quantisation artefacts. Increasing the bit-rate to reduce quantisation is certainly beneficial and often the advantage seen from a 10 bit over 8 bit codec may be due to a significant bit rate increase (provide you factor in the extra data required to record 10 bits and any codec efficiency factors), thus a reduction in quantisation. You need to ensure you have a noise figure for the sensor that is better than 58db for a traditional gamma curve, with log it is a bit different because of the way the samples are spread further apart as you go higher up the curve.

Editing is different as when you change or manipulate the image you will then be creating new samples and it is beneficial to be able to record those new samples with as little truncation as possible.

AFAIK the F3 will only have Cinegammas unless you get the Dual Link upgrade.

According to Peter Crithary from the Rule Boston Camera presentation a couple of the Cinegamma settings (that ship standard) are hypergammas:

Cine 1 (Hypergamma 4)
460%, 109% output

Cine 2 (Hypergamma 2)
460%, 100% output

Cine 3 & 4
(Same as EX1 & EX3

(Watch 12 minutes, 53 seconds in)

This is nothing new. All PDW-F3** and EX's have the 4 Cinegammas and CG 1 has always been the same as HG4 and CG 2 has always been the same as HG 2. I've written about this many many times. It is not unique or new to the F3 as claimed by Peter, they have been around for at least 5 years. A bit like his statement that S-Log is like Raw, it is not anything like Raw, S-log is a processed signal and is not straight from the sensor.

AFAIK the standard F3 will have the same 4 CG's as an EX1/3, it just so happens that 2 of these are the same as hypergammas.

"Originally Posted by Jim Tittle
I understand--with S-log, there's so much data that it takes dual SDI and a special recorder to handle it."

I am currently shooting 4:4:4 RGB C-Log (same as S-Log) on an Arri Alexa for film-out on a Dreamworks Animation project. We are recording onto 32Gb SxS cards.

Glad to know SxS cards are fast enough for 4:4:4 footage, but you still need a recording device with the right hardware.

The PMW-F3 may have SxS card slots, but that doesn't mean it has the hardware to actually feed a 4:4:4 signal to those slots.

Even if it costs 20k to activate S-Log output on the camera and buy a field recorder (likely less using CineDeck) that's going to clock in at half the price of an Alexa.

The SxS cards should be fast enough for the compressed pro-res that the Alexa records on them. The new SxS cards have a data rate around 1200 Mbps as I recall, while the prores is compressing the 444 at about 330 Mbps. Of course with the Alexa as well, to record at a higher data rate you'll need to do that to an outboard record system.
I'd be interested to learn more about the specifics of the Dreamworks project - is this live action footage that's being used as an element in animation? / what kind of downstream manipulation is the footage getting between the camera record and the eventual filmout?