Highlight Retention with IPP2 on DSMC1 test.

[Alex Lubensky]

When IPP2 was out, I have tried it a couple of times, but didn't see much gain on older Red cams (I use RedCine rarely). In December I was shooting a small spot for BUDxUBER, and there were two gel neon-colored Kino Tubes right in the frame. It was the time Resolve introduced IPP2 support, so I've played with it for some time on grade. I've noticed the KINO tubes acted vastly different on Legacy and IPP2, so I have decided to make some latitude tests later.

Every exposure chunk is 30 seconds long, so if you feel you've got enough of one simply jump 30 seconds.

IPP2 TEST (Medium, Soft):

LEGACY TEST (RC4, RG4):

LEGACY VS IPP2 highlights retention side by side:

EXPOSURE RAMP IPP2:

---

Tech info: Red Scarlet MX (firmware 6.3.102) in 4k DCI, Sigma 18-35 1.8 EF (at 20mm), 25 fps, shutter 1/50 (180). Black shading was done prior test.

Composed together in Davinci Resolve Studio 14.3, decoded in h.264 1080P at 30MB/sec limit (max debayer quality). Red Raw settings: 16 Bit, Full Res. Premium. Decode using: Version 3 (IPP2) RedColor4, RedGamma4. IPP2 set to: medium contast, soft highlights. Legacy Settings were: RedColor 4, RedGamma4.
No grading/denoising was done. WB Is set to 3200K.

The exposure base was ISO800 at F/5.6, shown as EXPOSURE 0. All on-screen marks are relative to base exposure, i.e. F/32 becomes F/64 when +1 setting applied, F/2.8 becomes F/2.0 when -1 setting applied.

The exposure was matched using RED Raw panel in Resolve changing ISO to corresponding level. The only exception was on ISO100 setting - I had to add 0.25 Exposure correction due to glass imperfections to match the base exposure. It didn't change the hilghlight clipping at all though.

Spot measurings were done with Seconic L558 meter on even flat surfaces.

 

[Alex Lubensky]

What I've noticed, is that IPP2 handles highlights somewhat better, like gaining 1/3rd of a stop in highlights retention. It's known, but I wanted to measure the exact limits of possible. Hope this will come in handy for someone.

 

[Shaeden Gallegos]

Thank you Sir! This is great

 

[Jack Gyori]

Thanks so much for this test! Was just about to do it myself.

 

[David Moreno Belmonte]

What about shadows/low lights? would it help to reduce mysterium/dragon sensors noise issue?

 

[Subrata Sen]

What about shadows/low lights? would it help to reduce mysterium/dragon sensors noise issue?

In Davinci Resolve I find better result by setting the Raw tab like this: Version 3, RedwidegamutRGB and Log3G10.
Coming to Color page, click on the clip button, and set highlight Rolloff to "very smooth".
Works beautifully on mysterium and dragon sensors. Less noise, better highlight retention and grades perfectly.

 

[rand thompson]

Subrata,


I think you meant "Very Soft" instead of "Very Smooth".

 

[Subrata Sen]

Rand,

Yes. It should be "very soft". That makes it very smooth.

 

[rand thompson]

Ha Ha, that would make it very smooth.