BrianMurphy
Veteran
Barry was there any discussion of backwards compatibility such as firmware upgrades for legacy AVC gear?
AVC-Ultra - revamped
- Thread starter Barry_Green
- Start date
Barry_Green
Moderator
Not that I know of; I can't think of any gear they've made that's been able to be retrofitted to accomodate newer hardware. Basically the cameras and gear that you have now, are what they are. Newer formats can be added to upcoming new products.
Barry_Green
Moderator
The times changed back in 2008 when they introduced AVCCAM.
They still advocate that intraframe is the highest-quality, most robust way to record. But the fact of the matter is that there exists a market for low-bandwidth codecs too, and the only way to get reasonable performance in low bandwidth is with long-GoP. The distinction that they draw is that they didn't do MPEG-2 long-GoP, they waited for h.264 and its greater efficiency, which makes long-GoP a more viable technique because it's not a case of the codec failing due to it just not being able to encode what's being thrown at it. Since h.264 is 2x to 3x more efficient than MPEG-2, the impact of it being long-GoP is a lot less visible; in fact at 50mbps of AVC it should be extremely robust. Already the 21mbps of AVCHD is reasonably solid, only softening up under the worst of conditions.
combatentropy
Veteran
Thank you, Barry. I'm pleased that the camera companies continue to improve things.
But consumer technology is moving faster. Sandisk just announced SDXC cards that handle 90 MB/sec. That's enough for 1920x1080 10-bit RAW at 36 frames per second.
AVCIntra, at 100 megabits per second, remember is only 12.5 MB/sec --- pretty paltry compared to consumer technology like the SDXC cards announced, as well as SSDs that handle 300-500 MB/sec.
But consumer technology is moving faster. Sandisk just announced SDXC cards that handle 90 MB/sec. That's enough for 1920x1080 10-bit RAW at 36 frames per second.
AVCIntra, at 100 megabits per second, remember is only 12.5 MB/sec --- pretty paltry compared to consumer technology like the SDXC cards announced, as well as SSDs that handle 300-500 MB/sec.
Barry_Green
Moderator
Because the manufacturers make claims about what their cards are capable of going "up to", but that doesn't say what their minimum sustained guaranteed write speed is. There's all these "super fast" cards out in all different modes, but look what actually gets recorded -- Red basically only works with their one compactflash card, the AJA Ki Pro Mini only works with a couple of cards on the market... it just isn't all that widespread that there's high-bandwidth stuff being recorded on inexpensive cards. The only device that I can think of that can store high bandwidth footage onto relatively cheap cards is the nanoFlash, and it's really different from everything else. I wonder if it doesn't have some sort of substantial buffer in it that gets it past the hiccups that the cards are likely giving.
combatentropy
Veteran
Despite my earlier complaint, this is definitely something to get excited about.
Everyone wants AVCIntra 100, but an MPEG-4 GOP codec at 50 might actually have better PSNR.*
Although, typically, intraframe-only compression is considered better than interframe --- and often really is --- it's not necessarily true. It's technically possible to have a lossless interframe codec. Interframe, or GOP, compression just means that for some frames only the differences from an earlier one are saved. It's possible to save all the differences, not just most of them. If you did that, it would, indeed, be mathematically lossless.
Interframe compression is so much more efficient than merely intraframe, that again, 50 Mbps inter- could be better than 100 Mbps intra-.
----
* PSNR (Peak Signal-to-Noise Ratio) is an interesting concept. There are codecs that are lossy, ones that are lossless, and ones that are "visually lossless." I used to think "visually lossless" was a squishy term. Maybe it's visually lossless to you but not to me. But there is some reason to it: If the noise introduced by the codec is less than the noise already in the uncompressed signal (from the image sensor, signal-processing chips, etc.), then you could call it visually lossless. (http://cineform.blogspot.com/2007/09/visual-lossless-and-how-to-back-it-up.html)
AVCIntra is considered visually lossless. I bet that AVC-Long-Gop would be also.
Last edited:
themagickite
Well-known member
that's pretty interesting, I'd always thought of long-gop as being a compromise...
Ok, so this is getting a bit hypothetical but... Would it be possible for a long-gop codec to compress colour information temporally instead of spatially?
So the A-frame contains a 4:4:4 16 bit whatever super colour sample, and the subsequent frames save on bandwidth by referring to the A-frame while adding any necessary changes.
Does panasonic have any interesting colour profiles to work with?
Sony seem to be leaps and bounds ahead with S-log
I can imagine canon/technicolor coming out with something unreal
As far as I can recall, panasonic were the first to employ anything like that with their cine-gamma, surely the team that thunk that one up has been busy doing something cool?
The more compressed your codec is, the more the in-camera "look" colour/contrast settings are baked into the footage... right? Or would it be theoretically possible to have a highly compressed codec which also allows for a lot of post work?
Ok, so this is getting a bit hypothetical but... Would it be possible for a long-gop codec to compress colour information temporally instead of spatially?
So the A-frame contains a 4:4:4 16 bit whatever super colour sample, and the subsequent frames save on bandwidth by referring to the A-frame while adding any necessary changes.
Does panasonic have any interesting colour profiles to work with?
Sony seem to be leaps and bounds ahead with S-log
I can imagine canon/technicolor coming out with something unreal
As far as I can recall, panasonic were the first to employ anything like that with their cine-gamma, surely the team that thunk that one up has been busy doing something cool?
The more compressed your codec is, the more the in-camera "look" colour/contrast settings are baked into the footage... right? Or would it be theoretically possible to have a highly compressed codec which also allows for a lot of post work?
It's 42 MB / 336 Mb on R1 MX, Epic can do ~3 times that.
Still, P2 is definitely fast enough.
I wish they (and Red, btw) would use non-proprietary storage, though. But I can understand why they don't. Too much garbage out there.
combatentropy
Veteran
Yes, a long-GOP 4:4:4 codec is technically possible. But to clarify, current long-GOP codecs compress both ways, temporally and spatially. So, your question should be, "Would it be possible for a long-GOP codec to compress color information only temporally, instead of both temporally and spatially?"
B
Barry_Green
Moderator
That's because initially, it was. With HDV and XDCAM, it was the compromise they had to use to fit HD footage onto a bandwidth-limited recording medium. HDV was designed to work on a DV tape, with a maximum transfer rate of 25 megabits per second, and there's no way you could even get in the ballpark with MPEG-2 without using serious long-GoP compression. With XDCAM, they had a limit of 36 megabits of write speed on the disc, so they had to fit it all in that limited bandwidth. 36 was better than 24, but still...
The thing is, long-GoP is a valid technique, but it should be used to enhance quality to an adequate level, not as a Hail Mary to squeeze a video signal down to an inappropriate/unsustainably tiny size. The NanoFlash uses long-GoP at up to 180mbps -- and it's superb at that rate. But at 18mbps, it's awful. There's a minimum quality level you just can't go below; long-GoP can improve video quality within reason but it can't be used beyond that. That's why HDV kind of stunk.
AVC is a lot more efficient than MPEG-2 ever was. Even Sony says that you only need 9 megabits of AVC to match the video quality of MPEG-2 HDV running at 25 megabits. So you're much, much closer to quality footage with AVC, and adding long-GoP on top of it can help you reach a satisfactory level of video quality.
I'd still take intraframe whenever possible, but long-GoP is a valid technique as long as it's used reasonably. 25mbps long-GoP is certainly going to look better than 25mbps intraframe of the same codec. Even so, the Nano can do both long-GoP (at up to 180mbps) and intraframe (at up to 280mbps) and Convergent Design says that the I-frame 280mbps is the highest-quality mode it offers. Which is about on par with AVC-Intra at 100mbps.
At the same price bracket, Panasonic has PanaLog. If they introduced an F3 competitor, I'd expect to see it equipped with PanaLog.
Well, it depends on what you consider highly compressed. RedCode is very highly compressed, it's 336 mbps of raw 4k intraframe. That's about equivalent to 80 megabits of 1080p. AVC-Intra is 100 megabits of 1080p, so ... if Red can do it with raw, then surely Panasonic could do it with AVC-Intra data rates. It's just that the video companies haven't been convinced that the video industry (note, not the digi-cine industry, but the VIDEO industry) would actually want raw. Video folks are typically working with video signals, not raw data files. If we can convince them that we actually do want raw, it would seem to be well within their capabilities to supply it.
themagickite
Well-known member
cooooooooool
I'd love to see some other companies get on the cinema DNG band wagon
but even if they all come out with their own competing proprietary raw formats, I doubt I'd complain too much.
I'd love to see some other companies get on the cinema DNG band wagon
but even if they all come out with their own competing proprietary raw formats, I doubt I'd complain too much.
Jarek Zabczynski
Veteran
I can snap an SD card in half using three fingers in one hand. I'll take my P2 thanks. Those things can probably take a small bullet.
I don't know if anyone has been fallowing the GH2 hacking development... but long-GOP compression (at least the way it's being done in camera) has pretty much been proven to be the culprit of "video-y" looking footage. The 1-GOP GH2 footage clearly moves more like a real film scan or Red... where each frame has been compressed independently. The motion doesn't have that "sliding" effect that the stock 24mb/s long-GOP codec had. These new codecs will probably be great for broadcast and such... but if you're really into replicating that "digital cinema" look, I'd be pushing Panasonic for all Intra-frame capture.
I agree that RedCode is highly compressed, but it's based on wavelet compression, which is considered by many to be more efficient than MPEG4-based AVC. I'm sure that Panasonic can deliver raw data, but maybe not with 100 Mbps AVC intra.
Barry_Green
Moderator
Doesn't have to be AVC. Panasonic does wavelet JPEG2000 also. They make a JPEG2000 DCI-compliant board for their D5 HD decks. So they could offer something fundamentally equivalent to redcode if they chose to.
Why? Not being sarcastic, I'm just not sure what in real world terms what the difference would be (either in what a camera could record or the ability to push around the image in post).