444 RGB S-Log has better highlight handling than 422 S-Log

[Alister Chapman]

IGNORE THIS and skip down to my next post (after Doug's). I've found this is another one of those NLE "Gottcha's". There is NO difference in dynamic range between 444 RGB and 422.

I have been testing various codecs as some of you may know and I came across an unexpected discovery today during an extreme exposure test. I discovered that 444 RGB footage recorded via the Convergent Gemini has improved highlight performance compared to 422 external recordings or the internal recordings. The difference in the test I performed was about an extra half stop.

This came as a surprise as I had expected similar dynamic range from both 444 and 422.I believe that the difference is that with 444 R,G,B you can have discreet peak values of 109ire for each of the R, G and B channels. But 422 is YCbCr where Y is an additive combination of primarily G plus about quarter of red and even less of blue ( 0.2126R, 0.7152G, and 0.0722B). So if the conversion to YCbCr isn't exactly right, Y can reach 109 before each of the discreet R, G and B channels reach 109ire (depending on the colours within the scene and especially with bright white light), thus it's possible in some situations (like a bright sky) that Y will clip before the individual R G and B channels so you end up with less headroom with 422 YCbCr than 444 RGB.

[Doug Jensen]

On a similar note, if anyone is interested in comparing footage from the Gemini vs. five other recorders, Vortex Media has 50GB of raw footage available for ordering. There's really no way everyone could test this many recorders and shooting modes on their own, so I did it for you.

PMW-F3 S-LOG & External Recorders Test Drive

There are examples of S-LOG vs. Cinegammas recorded with Gemini 4:4:4, Cinedeck EX, Pix 240, Ki Pro Mini, NanoFlash, and SxS cards. 8 bit, 10 bit, 4:4:4, 4:2:2, and 4:2:0 examples. MP4, MOV, DPX, DNxHD, ProRes 422, ProRes 422HQ, ProRes444, CineForm Film Scan1, and CineForm Keyscan.

If you want to evaluate S-LOG and several recorders and codecs for yourself, this is your chance -- and for a fraction of the time and trouble it would take to do on your own even if you had the hardware to do it.

[Alister Chapman]

OK. The question over what's happening to illegal levels was raised by a commenter on my blog and I've dug still deeper.

First of all: There is NO difference in the dynamic range of the Gemini recordings and the 422 recordings. This is what I had expected, but not what my NLE's (both FCP and CS5.5) were telling me.

The problem is this. The Gemini is RGB and when you drop RGB into an NLE's time line the levels are mapped to legal levels, i.e. nothing above 100% (unless you have superwhite mapping turned on in FCP). The Samurai and internal F3 recordings however go up to 109%. But the NLE maps the output to the monitor or frame grabs ignoring anything over 100%, so the illegal whites get clipped and the appearance is that the 422 recordings (illegal levels) have less range than the 444 RGB (legal levels).

My bad. Sorry if I've made anyone panic. I guess I'm going to need a decent dual link 444 monitor to really analyse 444 footage. With 422 I can just plug it in to my Transvideo and measure everything with that.

[Nate Weaver]

Hey Allister, you gotta remember too that a lot of folks shooting with the Gemini are going to be sending their final products through color correction.

Personally, I never get nitpicky about what a camera looks like in an NLE because when the original footage is loaded up in the Davinci, it's usually a slightly different ball game.

Say FCP clips those QTs, but in Davinci they're read differently and those problems either are not there, or are trivial to fix. This goes for Scratch, Pablo, Baselight, any of them.

Originally Posted by Alister Chapman

I guess I'm going to need a decent dual link 444 monitor to really analyse 444 footage.

I just sold my Dreamcolor and Pana 1700, and bought a Sony PVM-1741. It's amazing. HIghly recommended, won't break the bank TOO bad.

[Steve Kalle]

From what I have seen and heard, Flanders Scientific produces amazing monitors and some of them come equipped with 444 inputs at no extra cost.

[Nate Weaver]

Originally Posted by Steve Kalle

From what I have seen and heard, Flanders Scientific produces amazing monitors and some of them come equipped with 444 inputs at no extra cost.

For LCDs, I would agree they're some of the best.

OLED is a different ballgame...after buying this one, I won't be buying any more LCDs unless I have no choice.

[Peter Corbett]

Nate, does the 17" size feel large enough in an edit suite (after your Dreamcolor)?

[Nate Weaver]

Originally Posted by Peter Corbett

Nate, does the 17" size feel large enough in an edit suite (after your Dreamcolor)?

I've downscaled my edit rig to a 17" Macbook Pro for the moment. It's big enough for me, but I don't do supervised edits ever, I work alone and do postings.

If I'm ever concerned about focus for a shot, I just put my nose to the monitor to check. The 17" is full raster, 1920x1080 res. Right now a 2 foot distance is working nice. If it had to be farther away, I'd get the 24. But at this point, I'd sell the 17 and gladly pony up the extra $1700 for the 24 if I had to, it's that good.

[Peter Corbett]

Yes I don't have clients in my studio (aka den). The 24" would be the go but it is $2K more. I'm just using a Samsung HDMI domestic 20" and a Sony 9" SDI CRT for colour checking!