Will data rate vary with image size?

[Brook Willard]

Is it just that the developement time is prohibitive, or is it a decision based on something else?

It would require considerable additional development time, cost, complexity and power to do so, and frankly there is no point - all of these codecs leave image quality on the table v's REDCODE RGB and especially REDCODE RAW.

There you go.

 

[Stokestack]

OK, thanks for all the feedback. First, let me say I am aware of the current limitations in the known camera specs and limitations of RGB and so forth.

So the remaining questions, with a qualifier or two:

(If the camera DOES reduce the data rate for smaller images): Why not scale the raw data in camera? I did ask this before and got an appreciated but inconclusive answer (as I remember, since I'm too lazy to look for it). Is it really impossible or too processor-intensive? You could scale the image down, keep your 35mm lens view, keep the full chip latitude, and save storage space. Yes, there'd be additional loss of resolution given that raw is still "Bayered", but you're going to have that shooting windowed 2K also, without the benefit of keeping your 35mm lens FOV.

Given the compromises (and minimal storage savings) of RGB, a lot of my question is really only interesting in the context of the scaled-raw concept, I suppose.

 

[Michael Ragen]

I believe scaling raw is actually impossible. Maybe there would be some way of only reading a 2k amount of pixels out of the full 4k sensor to maintain the depth of field?

 

[Brook Willard]

I fail to see the issue with just scaling in post?

 

[Evin Grant]

I believe the reason lies in the chain of image processing, when you capture a RAW image off the sensor one of two things happens next, you can either write the info down, either with compression or not, or you can process the image which requires you to debayer it. Becasue if you scaled the RAW image down before you debayerd it then the color filters would all be in the wrong place. You would end up with a very strange looking image. You can address a particular group of pixels in a CMOS sensor which is how 2K RAW is achieved but adressing individual pixels in a larger array seems to me to require a higher level of complexity to the chip archetecture, I'm not an engineer but this seems logical.

 

[Stokestack]

I fail to see the issue with just scaling in post?

The issue is the lack of storage savings. As far as I can see, there are only two reasons to shoot anything less than 4K:

1. Storage savings
2. You want to use 16mm lenses (only applies to windowed mode, of course).

So given that the Red offers frames smaller than 4K, somebody must assume there's a point.

As far as scaling raw goes: You have discrete photosites for red, green, & blue, right? So I'm curious as to why you couldn't average clusters of red, clusters of green, and clusters of blue sites separately and record them in a raw layout as if the data were coming from fewer photosites. I can see where this requires (as Evin noted) addressing individual sites, where it might be computationally too demanding, or where the distance between like-colored sites might be too great to interpolate properly without involving the other-colored neighbors... but really it'd be interesting to hear exactly what the hurdles are.

 

[Michael Ragen]

Don't forget you can shoot 60fps at 2kraw or 120fps 720rgb scaled from the 2k window. 4k only does 30fps so far.

And you can use B4 lenses with the 2k window.

I recall a long time ago Graeme saying that even in RGB white balance would possibly be a post decision as opposed to being baked in in-camera. Anyone have any info on this? I could be wrong in my recollection.

 

[I Bloom]

As mentioned, you could get a DVCPRO HD deck to go with the RED, run single-link HD-SDI out to it. Here is a $12,000 product:
http://catalog2.panasonic.com/webap...d=103037&catGroupId=34402&surfModel=AJ-HPM100

Unfortunately it's way overpriced. They have entire cameras that cost half that. I think that would be a great solution for this problem. I'd actually be kind of excited to see it. But I think coming from Panasonic it's unfortunately cost prohibitive. We need something like the FS100, but unfortunately that device lacks the necessary encoder. That's how they get you. If it's possible to find an encoder that transfers HD-SDI to compressed DVCProHD over Firewire then we might be in bussiness.

I fail to see the issue with just scaling in post?

The issue is the lack of storage savings. As far as I can see, there are only two reasons to shoot anything less than 4K:

1. Storage savings
2. You want to use 16mm lenses (only applies to windowed mode, of course).

3. Time x labor = cost. The main issue for many things.

This is a big variable right now, as its hard to say how much time a REDCine session will add.

So given that the Red offers frames smaller than 4K, somebody must assume there's a point.

As far as scaling raw goes: You have discrete photosites for red, green, & blue, right? So I'm curious as to why you couldn't average clusters of red, clusters of green, and clusters of blue sites separately and record them in a raw layout as if the data were coming from fewer photosites. I can see where this requires (as Evin noted) addressing individual sites, where it might be computationally too demanding, or where the distance between like-colored sites might be too great to interpolate properly without involving the other-colored neighbors... but really it'd be interesting to hear exactly what the hurdles are.

I unfortunately I think you'd be playing leapfrog spatially with pixels by averaging them in this way. Bayer works in part because spatially the pixels correspond to the color readings at that position. Once you start transforming the image this correspondence disappears (AFAIK).

IBloom

 

[Gavin Greenwalt]

It's not impossible to scale a RAW image, just impractical you have to:

Debayer to RGB -> Scale -> apply OLPF simulation (aka .xx pixel blur) -> Discard color channels in a bayer pattern.

If you *realllly* want a 25% bandwidth saving, you would be better off just decreasing the wavelett quality IMO.

But 2k Bayer != 2k. 4k bayer > 2k RGB. So like you said, there is no real reason to use 2k RGB unless you really need that extra 25% recording time.

 

[Adrian Correia]

Ibloom, it seems to me that a DVCPro HD or HD-CAM deck rental is the best option for your concern. This is the exact reason that dual link HD-SDI outs are included on the camera. Panasonic also has thier new P2 drive (AG-HPG10) with HD-SDI in, battery powered. Although it seems a bit silly you could ostensibly turn your Red into a UberHVX by hooking up this drive and recording the HD out to P2 in the DVCPro-HD codec your clients are demanding, keeping a 4K Redcode Raw master for yourself.

this isn't a bad option if you are shooting for a producer who wants something tangible in his hands at the end of the day. I actually just had this conversation today with a producer. It juat took a little talking and things were clearer - but i definitely think some people are going to need their hand held with this workflow..thankfully P2 already started this trend so I don't think it will be so bad....especially when producers see the quality this camera can produce.

 

[I Bloom]

this isn't a bad option if you are shooting for a producer who wants something tangible in his hands at the end of the day. I actually just had this conversation today with a producer. It juat took a little talking and things were clearer - but i definitely think some people are going to need their hand held with this workflow..thankfully P2 already started this trend so I don't think it will be so bad....especially when producers see the quality this camera can produce.

At that pricepoint though I think a DVCProHD tape would be better. Since people that are ready for P2 are more likely ready for REDCODE. Besides its easy to get caught with you pants down with P2 capacity issues.

The one thing missing from the HD-SDI to external system, is variable frame-rates. Again probably not an issue, since dramatic productions and music videos will want REDCODE. We hope.

Overall, it adds so much cost that it's 6 and half a dozen between a REDCine workflow. Unfortunately for me, I get on a lot of planes with my camera package. I'm worried about not being able to bring enough horsepower computerwise.

Should be fine though. We'll work it out.

IBloom

 

[Stokestack]

I unfortunately I think you'd be playing leapfrog spatially with pixels by averaging them in this way.

That's what I was talking about with, "where the distance between like-colored sites might be too great to interpolate properly without involving the other-colored neighbors".

It's not impossible to scale a RAW image, just impractical you have to:
Debayer to RGB -> Scale -> apply OLPF simulation (aka .xx pixel blur) -> Discard color channels in a bayer pattern.

We're talking about resampling without debayering.

 

[Sebastian Cramer]

Ok, if I'm looking at a windowed sensor of 2 K to shoot Redcode RAW, will my datarate drop from 25MB/s to a quarter? I'm specially interested in the slowmotion features and in a lot of cases I just have to deliver SD. So 2K RAW would absolutely do my job with still a lot of oversampling. I'm just wondering if my asumption the data rate drops down to about 6,3MB/s is right.

Anyone has an idea on this?

 

[Gavin Greenwalt]

Probably not a quarter. I would expect closer to 1/3 at best. (7-9MBs). Again no official word yet to my knowledge.

 

[MikeCurtis]

Stokestack - Graeme, the guy writing the codec, has said it is not that easy to scale RAW. I'll take his word for it, he is scary smart. Gently, I suggest let that idea go, it has been considered and released by folks smarter than us as non-viable at this time. I went through the same logic loop you're discussing many months ago.

RAW is more efficient than RGB. If you want a lower datarate and can live w/Super16 optical characteristics and 2K res, shoot that.

Perhaps there might be a datarate dial to reduce image quality and datarate - that would certainly help. A slightly less well compressed 4K RAW image, scaled to say, 720p, might not show a meaningful difference when compared to a "full-on" datarate 4K RAW scaled to 720p. That isn't a stated feature of the camera, but knowing a bit about wavelets, that seems like it could be a future possibility.

As for having stuff at the end of the day, a couple of options:

1.) If need to be untethered, at end of the day, realtime dump over HD-SDI to a deck of choice. If want data for ingest into NLE, consider something like AJA's IO HD or MOTU's V3HD for ProRes or DVCPRO HD output over FireWire to a laptop - can hand off a FireWire or SATA drive at the end of the day (probably less costly than an HD deck as well)

2.) If you need the fastest possible release to customer, consider a deck, V3HD, or IO HD tethered to the camera and recorded live via HD-SDI.

Red, as Stuart stated, is sticking with their plan of Redcode RAW & Redcode RGB as the only two in-camera codecs. If they are stating that is their plan, and it would be time/money/size/weight/power detrimental to add other codecs, not to mention lesser quality, that makes sense to me and I'm not going to argue with the engineers about how hard something is or isn't - they know this stuff better than I do.

-mike

 

[GlennChan]

We don't and won't offer DV / DVCAM, DVCPRO 50, DVCPRO HD, HDCAM, ProRes or DNxHD video codecs in the camera body.

It would require considerable additional development time, cost, complexity and power to do so, and frankly there is no point - all of these codecs leave image quality on the table v's REDCODE RGB and especially REDCODE RAW.

Hmm this isn't Red's intended market, but if you wanted to make the camera suitable for the news market there might be an argument to implement those other codecs into the camera. In a shared storage environment, low bandwidth and native editing are desirable. Lots of people browsing and working off footage on a server puts demand on storage and networking... which is why I believe DV25 is used in news applications. Global (in Canada) for example shoots DV25 (albeit on tape). HD is definitely on the horizon and news organizations are looking at that... I'm not sure what formats (or format) will get adopted by the various manufacturers. Possibly HDV (one of its variants), possibly AVC. I wouldn't know.

Whatever essence + wrapper (or ingest method) they go with, it might be valuable if Red could shoot that format so it could be used for news. Of course:
A- You need your system to play well with the asset management solution, and the NLE, and the playout system.
B- There are competing products on the market... Sony, Grass Valley, Panasonic, etc. e.g. Sony's XDCAM EX looks like it makes a lot of sense. Tapeless, HDV/MPEG2 currently has lots of support for it relative to other HD formats, and HDV/MPEG2 is low bandwidth.
C- As lame as tape-based DV25 is, people are using it. Tape-based requires capture time, which is kind of lame. And DV25's quality is kind of lame by today's standards. But it works (see A).