RED down-conversion issue...anybody else had any luck?

[Joe Walker]

I'll write a small review later of my first time working with a RED One camera later but right now I have a more pressing issue:

 

So after shooting on a RED cam, we take the footage into REDCINE and convert it to 1080 23.98 Uncompressed 10 bit and it looks great. Trouble is we need to downconvert to SD. We've tried everything. Using software we've used After Effects, Shake, Final Cut, and all of our SD footage is showing mad aliasing...I mean bad! The only way that's "working" for us right now is to route the HD footage out of one edit suite and into a Panasonic AJ-HD150 deck and then route that into another edit suite to be captured as SD. That's the only way to get a clean SD copy with no aliasing. Trouble is that ties up two edit suites just to do a downconvert. My question to all those that have used the camera and successfully made an SD downconvert (in one edit suite), how did you do it? Any help would be appreciated.

[John Sprung]

To do a broadcast quality down conversion, you need to route it thru a Teranex or Snell & Wilcox box. That's why good downconversions are expensive. Back in the tape world, there's an offline quality downconversion built into the HDCam SR machines.

 

If you just want offline quality making do with what you have, you need to soften the HD to make the Nyquist limit of your SD format, and then downconvert. See what kind of blur or soft focus options you have in the various software packages you have, and test a bunch of options on the stuff that aliases the worst.

 

 

 

-- J.S.

[Joe Walker]

John,

 

I sort of figured it had to be a hardware fix. The blur options have helped somewhat, and we demoed a Teranex Mini. It had acceptable results and next we're going to demo the higher priced model. Our Kona 3 is....well....not the best at downconverting this stuff. Thanks for your help.

[Scott Fritzshall]

To do a broadcast quality down conversion, you need to route it thru a Teranex or Snell & Wilcox box. That's why good downconversions are expensive. Back in the tape world, there's an offline quality downconversion built into the HDCam SR machines.

 

If you just want offline quality making do with what you have, you need to soften the HD to make the Nyquist limit of your SD format, and then downconvert. See what kind of blur or soft focus options you have in the various software packages you have, and test a bunch of options on the stuff that aliases the worst.

What is the difference between how a hardware box and software packages do it? Because I resize stuff in software all the time and have never had any problems, even for broadcast. Just as a test, I just resized a frame of RED footage I had lying around from 1080p to NTSC in Nuke, and it looked totally acceptable. Is there something I'm missing here?

[Gary McClurg]

Here is what a friend did... Here are his tests he did... based upon the first post that I sent him through an email...

 

Just exported a 4K RED clip to uncompressed SD with REDCINE and all is well. No visible aliasing.

 

Maybe a codec problem?

 

What's the goal here? Do they just want to edit in SD or are we talking about a deliverable? If DVD, why not do the edit with the 1080 files and let Compressor do the down convert?

 

Just after sending that I realized I didn't follow their workflow. I went 4k to SD. They went 4k -> HD -> SD. Trying that now.

 

Stay tuned.

 

Next email from him...

 

Ok, did 4k -> Uncomp HD -> Uncomp SD.

 

Originally I did the 4k to HD downconvert with REDCINE. And the HD to SD with Shake. I got really weird results that I can't explain.

 

Then I started with REDCINE to go to HD, and then in FCP put that HD clip on an Uncomp SD timeline and let FCP handle the downconvert. Exported a QT from FCP and all was fine. No aliasing, just a nice SD clip.

 

Last thing. The footage I have was shot in September which would have been a way-old firmware build, so my test isn't exactly comprehensive.

[Joe Walker]

Just as a test, I just resized a frame of RED footage I had lying around from 1080p to NTSC in Nuke, and it looked totally acceptable. Is there something I'm missing here?

 

The problem we've had is only visible in moving footage, when the playheads are parked on still frames, everything seems fine, its when we hit play we're seeing all sorts of jagged edges. :rolleyes:

[DJ Joofa]

Joe,

 

Can you make some portion of that clip available in a standard format (not in R3D)? I want to have a look at that. In case you can, please take it offline, my contact is djjoofa (at the rate of) yahoo (dot) com

Edited May 29, 2008 by DJ Joofa

[John Sprung]

... when the playheads are parked on still frames, everything seems fine, its when we hit play we're seeing all sorts of jagged edges. :

Try moving forward thru the tape field by field with the jog knob. It sounds like the frame buffers may be getting read out to tape in the wrong order. Look around for field1/field2 dominance settings in the software. NTSC also has that four field color frame sequence, see if there's any reference to that, too.

 

 

-- J.S.

[John Sprung]

What is the difference between how a hardware box and software packages do it?

In theory, there's no reason that software couldn't do downconversions the same way the dedicated boxes do it. In practice, the vast majority of the need for broadcast quality downconversion exists at the network and station level, where they really don't need anything else that a computer could give them, so a dedicated box makes the most sense. I'm not sure, but another possibility is that they (Teranex, S&W, etc) figure it's harder to steal the secret sauce out of a box. ;-)

 

 

 

-- J.S.

[Joe Walker]

Try moving forward thru the tape field by field with the jog knob. It sounds like the frame buffers may be getting read out to tape in the wrong order. Look around for field1/field2 dominance settings in the software. NTSC also has that four field color frame sequence, see if there's any reference to that, too.

 

 

-- J.S.

 

 

Sorry John, I was being facetious, I was referring to the clip on the Final Cut timeline, watching the preview out on a high def monitor via the Kona 3 card. Going to tape isn't really an issue, so long as we route to the AJ-HD150 first, from there we can pretty much go anywhere, its going the software route of down-conversion that's killing us.

 

DJ, I'll try and make something available to you guys here soon, so you all can take a look at it.

Edited May 29, 2008 by Joe Walker

[Shawn Booth]

It has also been suggested to go from 1080 > 720 before dropping down to SD.

[Michael Most]

What is the difference between how a hardware box and software packages do it?

 

Sophisticated. multi-pass, adaptive motion estimation and anti alias technology. Software could do such things, but the rendering time would be exorbitant, and the software is usually not that sophisticated as it is usually just one feature of a product that is built for a different purpose. Dedicated hardware boxes, like the Snell and Wilcox and Teranex products, are specifically built to do what we're talking about, and have much more sophisticated processing for things like pulldown detection, deinterlacing, and antialiasing.

[Shawn Booth]

I just burned a DVD of a music video we shot on RED. I used RedCine to create UC10bitters, edited and then exported from FCS using Compressor, burned the disc in DVDSP.

 

Zero problems. This time. First time I tried it (on another video) I have artifacts all over the place, especially in the blacks.