Any Good Discovery about How Downconvert to SD?

[Gabriele Turchi]

I had to prepare fro an Offline SD Job...and off course the SD Version Looks Really Bad....

The Kona SD Realtime Looks Bad...

I Guess recording into an HD Deck First a 1080 signal , than use the deck to downcorvet and recapture should generate good results ..but is an expansive way to do it...

So , this SD downconversion was a problem a while ago...But i was never really into it until Today....

So I am wondering if now exist some good Solution...

Thanks

G

 

[jimhare]

I think some people apply a slight blur when downconverting.

 

[Gabriele Turchi]

Well i guess that is not the high quality solution...

Also i forgot to say that i am noticing how a very well calibrated HD Image (let say Prores 1920x1080), after the conversion in 720x480 (so SD Timeline) and using an SD signal out Like 525i out from Kona Component ...teh Image Looks Much Darker....

Like If the Downconversion to SD should Change the Black Level...

Any thoughts?

Thanks

G

 

[BigLu]

Can you show some samples?
Id like to look deeper into this.

 

[Gabriele Turchi]

I guess th eonly way to show is to photograph my plasma Tv....And i am sure wiil not be a good reference example...

The point is that i hav ecalibrated and exported some HD clip , than they ask for an SD version because they are going to intercut with Digibeta SD Stuff (so SD timeline) i have exported the Prores HD into prores SD...And playing back the SD timeline using an SD signail the image is really Dark...

People from Aja Told me that if i am using the anolog out (i do) i should set the aja panel black level on 0 IRE fro HD and 0.7 for SD...But actually switching those e setting i can't see any difference...

I am wondering if between SD and HD there is a different black level system so if prepare material in HD that is going to go in SD should have a ""special"" procedure taht take in consideration this black different Level...

PS:They Told me ""if you use analog""...Maybe for SDI does not count...

Any help is appreciated

Thanks

G

 

[michael L]

any updates to this? I too dread going SD from RED because it looks so bad. Looks worse than SD shot on prosumer MiniDV cameras.....honestly!

 

[Gabriele Turchi]

I agree...

Dynamic Range Exclude Of Course!!

I can't find a software that do that Well...But it is weird because the DVD Encoding Looks Good...But a SD Prores...Looks very Bad...

It seems that the only good way to do It is to export on a HDD5 Tape and than dowconvert using the Deck...

Even The Kona Does not a God Job from HD...Otherwise the solution could be playback out from Kona and record Using another Kona (and another computer...)

PSoes exist a better downconverter than Kona?


G

 

[Kevin Wild]

I'm not sure this is really a "RED" issue as much as a normal downconversion issue no matter what camera you use.

We have not had good results using our Kona LHe's to downconvert anything with graphics. On many spots and videos we do, we try to do graphics AFTER downconverting the footage, so we do graphics in SD if going out to tape. Otherwise, the real time Kona output has a ghosting type look to graphics that is not good. (We have 4 rooms all with this and they all show this. I've talked with AJA and Miranda engineers at NAB last year and they basically shrugged their shoulders and said "that's what you get.)

Sorry I can't offer any great solutions...let us know if you find a good way!

Thx.

Kevin

 

[Charles Angus]

What are you doing that isn't working?

c./

 

[Dan Hudgins]

Downsample issues for 4K to SD

Downsample issues for 4K to SD

I am working on one solution to this problem, but the code is not posted yet, you can PM me in maybe 6 weeks or check my "What's New" to see if I posted the new downsize code.

One problem with downsize in many programs is that they do not use 100% of the 4K pixel data when they downsize to gain throughput. If you read 100% of the 4K pixels and do real number adjustments to the brightness in each of the output pixels at SD (what SD size are you using, like 720x480?) it takes more time.

The results I am getting so far look good, but I have to work in some more details before it gets posted.

When a program skips pixels in the 4K data you get more aliasing and noise as well as the loss or flicker of smaller details in the 4K file, so the SD result looks very bad.

If you blur the 4K files before the skip pixel downsize, the pixels that get picked have some data from the skiped pixels, but the bit depth can limit the carryover so you still get aliasing and noise that should not be there.

If you are going to downsize you should not sharpen the 4K files, they have too many sharp edges as is. You can blur the 4K files downsize then sharpen a little at SD resolution.

If you down size in stages you may get better results, 4K > blur > 3K blur > 2k > blur > 1.5K > blur > SD > sharpen.

What I am working on does not use such half measures but does a digital simulation of optical printing, it will be somewhat slow because of the real number processing but should yeld lower noise results, we will see soon, if you want me to process some 4K TIF frames to SD TIF frames for you, you can email me (not sure if Yahoo takes 60MB files?) or give me the URL to download from for the sample file to work test conversion on...

I went to this because I have an even worse issue, going from 4K down to small sizes like 360x180 or 240x120 for playback from slow notebook computers and such for on-set editing.