Welcome to our community

Be a part of something great, join today!

  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

4K DI options

Blair S. Paulsen

Well-known member
Joined
Dec 28, 2006
Messages
5,503
Reaction score
34
Points
48
Location
San Diego, CA
Website
www.alacritymedia.com
I think we would all love to have a real time 4K pipeline for DI and if you have enough money you can do that today. What if you want to do it at a realistic price? Somewhere north of shoestring indie territory but south of a heavy iron post facility.

For the sake of this discussion lets set a timeline of a year or so. We want to grade native RedCode36 and see the results of our adjustments in real time on the display. Yes, this is a lot to ask - but we take it for granted that we can correct and view SD footage this way and with a little budget you can do the same in HD right now.

Note: due to the quality of modern compression schemes and the bandwidth/cost issues of uncompressed 4K this theoretical exercise posits "visually lossless" compression will be used as needed.

Parameters/Issues

Display topology:

AFAIK at NAB 2007 Red Digital used a DVS Clipster serving a Sony SXRD 4K Cinema sized projector via 4 dual link HD-SDI cables to show Peter Jackson's CTL short. Looked great but too expensive, too big and overkill for DI.

Jim has a 4K delivery system in development and has promised to show it at NAB in April 2008. The basic components are likely a server, a monitor or projector and a way to connect them. This is a key piece.

Grading:

Assimilate's Scratch system, well equipped, can grade native RedCode at 2K in mostly real time (depending on what operations you are doing). Perhaps with enough horsepower on the computing side this can be done at 4K. (Lucas, Mark or someone with more hands on Scratch time can jump in and clarify this as I don't wish to misrepresent their product.)

IMHO the best path to make this happen before 2011 is a dedicated acceleration module, PCIe2.0 card comes to mind, that can decode RedCode in real time. The alternative is to leverage the tech that Red uses in their 4K delivery system by integrating it with post tools. Of course I have no idea if Red is developing a stand alone server like the Clipster or something that runs on a MacPro tower with a hardware interface, or something else my pea brain can't conceive of.

Would there be enough market for someone like AJA to build such an accelerator card? What is the quantity/price break between FPGAs and dedicated silicon? What kind of CPU/GPU horsepower and bus interface would be required to decode on the fly without an special purpose accelerator?

Yes, before Mike Most and others jump on me, I happily concede that a talented colorist is a huge factor in the mix. I just want to give them the tools to see what they are doing without giving Discreet the deed to my house (or Ketch's :biggrin: ). In any case, I am out waving my sign and it says "4K power to the people", rock on.
 
Display topology:

AFAIK at NAB 2007 Red Digital used a DVS Clipster serving a Sony SXRD 4K Cinema sized projector via 4 dual link HD-SDI cables to show Peter Jackson's CTL short.

Small correction... At NAB, the SXRD was fed from a Keisoku Giken 4K DDR, not a Clipster.

Lucas
-----
ASSIMILATE, Inc.
LA, CA, USA
 
I think we would all love to have a real time 4K pipeline for DI and if you have enough money you can do that today. What if you want to do it at a realistic price? Somewhere north of shoestring indie territory but south of a heavy iron post facility.

For the sake of this discussion lets set a timeline of a year or so. We want to grade native RedCode36 and see the results of our adjustments in real time on the display. Yes, this is a lot to ask - but we take it for granted that we can correct and view SD footage this way and with a little budget you can do the same in HD right now.

Note: due to the quality of modern compression schemes and the bandwidth/cost issues of uncompressed 4K this theoretical exercise posits "visually lossless" compression will be used as needed.

Parameters/Issues

Display topology:

AFAIK at NAB 2007 Red Digital used a DVS Clipster serving a Sony SXRD 4K Cinema sized projector via 4 dual link HD-SDI cables to show Peter Jackson's CTL short. Looked great but too expensive, too big and overkill for DI.

Not suited for DI. Good luck calibrating that projector.

Watch JVC take the 4K projector lead at NAB.

Jim has a 4K delivery system in development and has promised to show it at NAB in April 2008. The basic components are likely a server, a monitor or projector and a way to connect them. This is a key piece.

Nope.

Jim did NOT say he would SHOW anything. He said he would announce details. I HOPE it's a server/decoder that would play out REDCODE RAW and REDCODE RGB (it's coming because it MUST for delivery) to QUAD DVI and QUAD HD-SDI. I would like to just PLUG a third party drive into the server without transferring files.

Grading:

Assimilate's Scratch system, well equipped, can grade native RedCode at 2K in mostly real time (depending on what operations you are doing). Perhaps with enough horsepower on the computing side this can be done at 4K. (Lucas, Mark or someone with more hands on Scratch time can jump in and clarify this as I don't wish to misrepresent their product.)

I am not really all that interested in viewing in 4K while grading COLOR. I'll do it when I can for the "smoke and mirrors" perception thing. The film will look the same in terms of color. 2K is the deal. The size of the screen and distance needed to resolve 4K is not really suit any DI suite I can imagine. What I want in my RED PROJECTOR is actually RED 4K image sampled to 2K - that is the SWEET RED SPOT.

IMHO the best path to make this happen before 2011 is a dedicated acceleration module, PCIe2.0 card comes to mind, that can decode RedCode in real time. The alternative is to leverage the tech that Red uses in their 4K delivery system by integrating it with post tools. Of course I have no idea if Red is developing a stand alone server like the Clipster or something that runs on a MacPro tower with a hardware interface, or something else my pea brain can't conceive of.

Would there be enough market for someone like AJA to build such an accelerator card? What is the quantity/price break between FPGAs and dedicated silicon? What kind of CPU/GPU horsepower and bus interface would be required to decode on the fly without an special purpose accelerator?

Yes, before Mike Most and others jump on me, I happily concede that a talented colorist is a huge factor in the mix. I just want to give them the tools to see what they are doing without giving Discreet the deed to my house (or Ketch's :biggrin: ). In any case, I am out waving my sign and it says "4K power to the people", rock on.

Not sure what you are after Blair. Right now, TODAY, we can grade 4K RED media in 2K realtime on Scratch on a prototype workstation from BOXX. render out 4K DPX or TIFF. Not sure what else you want that isn't already working ... March 7th, 2008.
 
I think we would all love to have a real time 4K pipeline for DI ......

I have to ask...

Why? What exactly are you delivering and to whom that this is important? If we're talking about actual distribution, it should be noted that at least 85% of high end DI work is done in 2K, and to date there have been very, very few digital cinema packages done in anything but 2K. So even in the highest end theatrical distribution chains, there is little demand or even opportunity for 4K finishing or delivery elements. Other than personal ego or a lack of understanding of distribution realities, what are your reasons for feeling that this is needed?
 
I have to ask...

Why? What exactly are you delivering and to whom that this is important? If we're talking about actual distribution, it should be noted that at least 85% of high end DI work is done in 2K, and to date there have been very, very few digital cinema packages done in anything but 2K. So even in the highest end theatrical distribution chains, there is little demand or even opportunity for 4K finishing or delivery elements. Other than personal ego or a lack of understanding of distribution realities, what are your reasons for feeling that this is needed?

Exactly. FIRST you want killer RED 4K to current 2K infrastructure.

But for the record ... personal ego and perception run this industry.

Oh yeah ... and stupid committees.

I am selling "F the DCI" T-shirts at NAB.
 
REDCINE EDL/ FCP XML Import

REDCINE EDL/ FCP XML Import

Why not we have edl/FCP XML import in REDCINE to do a primary color correction of what edited in FCP. Cheapest option.Once done with primary color correction export 4k as dpx file and take to any DI suite you want. Why not RED implement this in REDCINE as of now?

Sorry, if i deviate from the topic.
 
I'd be happy with finally having the R3ds drop right in the timeline, and having an FX plug to manipulate the native files from FCP. There's obviously some reason this hasn't happened yet but once that can be done I'll be pretty happy. Don't get me wrong, the 2K proxy is working great on my machine right now but the round trip to Redcine is workable but really not ideal.
 
I'd be happy with finally having the R3ds drop right in the timeline,
Evin - you mean QT wrappers from R3ds I assume. Which you CAN do now. RT performance is based on system, BUT wrappers do NOT have a true debayer. FCP timeline will NEVER take "r3ds" as FCP is built on QT.

and having an FX plug to manipulate the native files from FCP.

I don't think FXplug is the right architecture for the job.

Don't get me wrong, the 2K proxy is working great on my machine right now but the round trip to Redcine is workable but really not ideal.

Well ... round trippin to REDCINE is about to get a hell of lot more elagent. I'll show you in person next Sat. at the LA User Group breakfast.

You guys gotta remember that Apple has COLOR now in Final Cut Studio. More and more they will build build up that app - Apple is going to assume you want to EDIT in FCP and COLOR in COLOR.

IMHO - COLOR will be the DIY for grading RED. It will be the "poor mans" DI for RED. And since I have been holding my breath for over 5 years for Apple to solve GAMMA HELL - how long do you think I will have to hold my breath for them to change their GPU driver bottleneck? Until then, best of luck getting anything close to excting in Color's playback speed. I do think that EVENTUALLY Color will be bad ass. I just can't wait for them to do it (2-3 years) and I have already moved past it.
 
FCP timeline will NEVER take "r3ds" as FCP is built on QT.

Nonsense!

As we know for the last 10 years that QT has an open software architecture

AND ACCORDING TO THAT IS POSSIBLE TO INCOPORATE R3D CODEC TO QUICKTIME.

It's just a matter of RED and Apple how and when can do it.

We just have to ask Frank Casanova, Apple's director of QuickTime product marketing more about that.
 
IMHO - COLOR will be the DIY for grading RED. It will be the "poor mans" DI for RED. And since I have been holding my breath for over 5 years for Apple to solve GAMMA HELL - how long do you think I will have to hold my breath for them to change their GPU driver bottleneck? Until then, best of luck getting anything close to excting in Color's playback speed. I do think that EVENTUALLY Color will be bad ass. I just can't wait for them to do it (2-3 years) and I have already moved past it.

I've tested Color with my Nividia 5600fx and it is very fast. RT 1080p up to 30 fps. I am not a IT person but the 5600fx is a new dimension in this regard, IMHO.

Hans
 
Nonsense!

As we know for the last 10 years that QT has an open software architecture

AND ACCORDING TO THAT IS POSSIBLE TO INCOPORATE R3D CODEC TO QUICKTIME.

It's just a matter of RED and Apple how and when can do it.

We just have to ask Frank Casanova, Apple's director of QuickTime product marketing more about that.
You are still talking about EXTRACTING r3ds INTO and/or THROUGH QT.

Would you like to discuss how well FCP handles QT architecture?

Don't get me started ....
 
Would you like to discuss how well FCP handles QT architecture?

Mark,

now from this notion somebody could conclude that Apple did not

make well implementation of QT in FCP or even that QT is

totally wrong digital media platform

and for years they (Apple) bluffed thousands of FCP user base and more.

I think something is wrong here.

Apple, Quicktime or Off...
 
Exemplary for the worldwide 4K DI marketleaders are:

filmlights baselight.
http://www.filmlight.ltd.uk/

dvs clipster
http://www.dvs.de/

Pricing for these systems starts at 50-75 at dvs, ~100 at filmlight. For a lightspeed 4K setup one might want to throw in even more hardware.

The main differences to software basing DI systems using Nvidia (or rarer, ATI) graphics cards, as iridas speedgrade, apple color, assimilate scratch etc are:

- realtime 4K processing with FX as blurs and conform tools as dissolves etc
- 4K display in order to see what you are doing
- 4K i/o, typically via quad DL-HD-SDI
- dedicated hardware (i.e. clipster) or clustered computers (i.e. baselight).

while i agree that we will see mainly 1080/2K masters in 2008, i can only warn out of experiences to master at 4k without 4k monitoring/QA.

While 35mm filmout will be hiding many issues of a 2K for 4K master, in a digital 4K projection you will see lots of details you missed in a 2K DI for a 4K Master, logically, as you only see 25% of the resolution on 2K the master/source have.

Especially noise can be very hard or even impossible to spot on a 2K display when its at fine 4K, but i will be -noticeable- if screened by a 4K projector (or used at a tradeshow to feed several 1080p displays etc).
 
Mark,

now from this notion somebody could conclude that Apple did not

make well implementation of QT in FCP or even that QT is

totally wrong digital media platform

and for years they (Apple) bluffed thousands of FCP user base and more.

I think something is wrong here.

Apple, Quicktime or Off...

Marc is correct, Quicktime has -many- well known issues since -many- years.

Mainly these are painful for professional users, as the notorious crossplatform gammaissues etc.
 
I think we would all love to have a real time 4K pipeline for DI and if you have enough money you can do that today. What if you want to do it at a realistic price? Somewhere north of shoestring indie territory but south of a heavy iron post facility.
4K realtime DI systems isnt much different in price than i.e. scratch, dvs is somewhat surprising inexpensive.
4k -dci- compliant displays are still much to expensive, but if you grab one of the classic 3840x2400 pix lcds (they use to flow on ebay at ~1000-3000) you have a pretty good sharpness check
http://en.wikipedia.org/wiki/IBM_T220/T221_LCD_monitors
they are quite angle dependent and not to fast however.

Would there be enough market for someone like AJA to build such an accelerator card? What is the quantity/price break between FPGAs and dedicated silicon? What kind of CPU/GPU horsepower and bus interface would be required to decode on the fly without an special purpose accelerator?
this is above aja or decklink in 2008 - dvs (and others) fill that niche sucessfully since over 20 years. they were doing 1080p/2K when the other were pal/ntsc, now they are doing 4K while the others are doing 1080p/2K.

besides the approach dvs has (dedictaed multistream 4K with colorcorrection, fx, transparency etc), others, as filmlight, used clustered computers instead of a single one.

Whats nice about these systems is that they also allow fully realtime 4K networked storage, excellent if you want to attach some VFX / CGI / graphics systems in the background.

However, so far they were all not allowed to use redcode by red.
In the case of dvs they seem to integrate it since mid 2007, and once red opens the redcode codec at NAB 2008, i suppose they will offer r3dcode support rather soon.

4K RT -debayering- however seems not to be on the very very near horizon. Speaking with -the- DI raw pioneer, iridas, at IBC 07, they said that -maximum quality- 4k debayer with software/gpu basing systems in 2008 might be pretty narrow, and after supporting raw for the s:i 2k, vision phantom 65, cineform neo, arri D20 etc about the time red came to the market, i suppose they know quite a bit about raw.
 
MATRIX from Chrome Imaging, Geneva, Switzerland is coming up today

as the most affordable high-performance post-production solution with

a range of turnkey systems starting from DI, 2K to 4K.

MORE>>>
 
I've tested Color with my Nividia 5600fx and it is very fast. RT 1080p up to 30 fps. I am not a IT person but the 5600fx is a new dimension in this regard, IMHO.

Hans

Hey Hans -

I am impressed!

Can you tell me how fast 1080p ProRes with ONE primary grade and TWO secondaries?

Uncompressed 1080p with ONE primary grade and TWO secondaries?

Thanks on advance!
 
and for years they (Apple) bluffed thousands of FCP user base and more.

I think something is wrong here.

Apple, Quicktime or Off...

LOL!!

No, they just "buffed" the high-end PRO users because we are a NICHE MARKET.

We joke at Offhollywood all the time that they should call change the name Pro Apps to Semi-Pro APPS or Sorta-Pro Apps.
 
LOL!!

We joke at Offhollywood all the time that they should call change the name Pro Apps to Semi-Pro APPS or Sorta-Pro Apps.

QT is nicknamed by several VFX guys here gammagamble and slowtime.
 
4K realtime DI systems isnt much different in price than i.e. scratch, dvs is somewhat surprising inexpensive.

A Clipster at its *base* configuration is not much more expensive than SCRATCH. But it is still more expensive. But if you want the realtime 4K configuration - that is *much* more expensive. If you want multistream 4K, that is exponentially more expensive because you have to own not one, but two Cine4K servers.

Laguun, could you please explain how you can do multistream 4K off one Clipster? Since a Clipster can only connect to one Cine4K server at a time, and one server is only capable of one stream, how is this possible? Even the DVS website claims "up to 4K." with no mention of multistream.

Don't misunderstand me. I think Clipster is a great box, and I have a lot of respect for the LA team - but I think you are overstating their capabilities and I know you are understating their price. :)

A Baselight8 is the only Filmlight product capable of realtime4K. They start at over US$500K.

4k -dci- compliant displays are still much to expensive, but if you grab one of the classic 3840x2400 pix lcds (they use to flow on ebay at ~1000-3000) you have a pretty good sharpness check
http://en.wikipedia.org/wiki/IBM_T220/T221_LCD_monitors
they are quite angle dependent and not to fast however.

But 3840 x 2400 is not 4K, and these monitors have horribly slow refresh at maximum resolution. They were primarily intended for medical applications where refresh isn't a critical gating factor. If you're not going to be displaying true 4K, wouldn't you be just as well off with a decent 23"?

Lucas
-----
ASSIMILATE< Inc.
LA, CA, USA
 
Back
Top