NEW more detailed info on the DVX200 (4/24/2015)
- Thread starter atroa
- Start date
There has been a change - directly from R118 is now:
In the past the corresponding line simply read "4:2:2" (for both interlace and progressive)
joe 1008
Veteran
I do very much hope this is also true for the german and austrian mainstream channels ARD, ZDF and ORF as the last specs I could find were from may 2014 an claimed the contrary.
But yes, 4:2:0 should be OK, as we know that in progressive compression there's little difference compared to 4:2:2. But do THEY know? I mean, those structures are sometimes caught in a very conservative pattern.
But yes, 4:2:0 should be OK, as we know that in progressive compression there's little difference compared to 4:2:2. But do THEY know? I mean, those structures are sometimes caught in a very conservative pattern.
Thomas Smet
Veteran
Well now that really depends on the sensor creating the image. In no way shape or form can a 4k sensor create a 4:4:4 4k image. Everything in that image is interpolated and the chroma has the detail that is smoothed compared to how it should really be at 4:4:4. Red pixel #2 is a mathematical interpretation of red pixel #1 and red pixel #3. Same if you reinterpolate 4:2:0. There is of course rounding error and bayer is a bit more sophisticated but the reality is there just isn't 4:4:4 in a 4k single sensor camera. To truly get a 4:4:4 image you either need three chips or shoot at 8k.
Yes you have unique samples but my point is they are fake samples. All the 4:2:0 conversion does it throw out the in between fake pixel. Since it was created from an average chroma interpolation can get a new value back that is very close.
Thomas Smet
Veteran
Don't you find it a little odd that one of the best cinema cameras out there isn't even considered tier 1. I don't know about everybody else but I'm still waiting to see this magical broadcast that looks better than a movie like Guardians of the Galaxy or the Avengers. I love engineers and work with a lot of engineers and while they approach these standards with the best intentions and logic they usually lack common sense and a greater understanding of what content actually is.
It is typically my experience that these standards exist so not every Tom Dick and Harry start submitting content to TV stations just because they were told by Apple that their IPhone is a HD camera. By making up these standards they really filter out the amateurs. Yes it is about quality but quality can't always be judged by a piece of paper and published specs. This is a crude blanket policy with little to no thought or reasoning and nothing more than that. No different than IT departments in companies making up draconian rules that have no logic other than because they said so.
If you're talking about the Alexa, then in one sense I see where you're coming from..... but don't forget that it's only considered tier 2 for 4K - it's tier 1 as an HD camera. And if you're defining best practice for 4K acquisition, then surely truly being 4K must be a very important attribute!?!
It's like a butcher selling meat as organic when it isn't - it doesn't matter if it's wholesome, has great taste and texture........ if it wasn't farmed to organic standards, it ain't organic and shouldn't be sold as such. (I'd still be happy to buy it, but that's not the point.)
I disagree about it being a "crude blanket policy". As said previously, productions are allowed a certain percentage of non-conforming material anyway. They may be allowed a far higher percentage if circumstances justify it. How can such be described as a "blanket policy"?
To say nothing of lower tier cameras being considered fine for certain genres, if not for others. (XDCAM EX doesn't meet the full spec for general acquisition, but is heavily used for news throughout the EU.)
Yes, any policy will inevitably bring up inconsistencies, but at least with this it gives an excellent starting point. And don't forget we're talking about guidelines anyway. If Tom Dick and Harry really did make something fantastic on their I-phones nothing in the guidelines prohibits any broadcaster from showing such if they want to - and similar has happened.
I think the change is pretty recent, and remember it's a change in EBU policy so I'd expect it to take a while to filter into the policies of various channels anyway. And as said before, R118 is a set of guidelines, not rules. There's nothing to enforce any broadcaster to enforce them - but equally nothing to stop any broadcaster from having even more rigid guidelines!
And don't forget that even if origination was in 1080p/25, it originally had to be transmitted as 1080psf/25, and then all the interlace and desirability factors in favour of 4:2:2 apply just as much as for 1080i/25. Don't think it's just an organisation acting conservatively. It's my understanding that in the UK, 1080p/25 does get transmitted as such (not via psf) - but that may not be true throughout Europe. Any such broadcasters may therefore still have legitimate reasons for insisting on 4:2:2 throughout - at least at the moment.
Thomas Smet
Veteran
Exactly my point about the blanket policy. It is a starting point guideline and really nothing more. If one has to pick a standard they have to start somewhere so they pick what made the most sense when one factors in a lot of unknowns.
By setting standards at a certain level the one thing that typically will be gained is production standards and high production value. By creating standards at a certain level requiring a certain level of gear you tend to attract those who focus a lot more time, money and effort into the production value. It is typically only the seasoned pros who have access to certain gear or at least willing to invest in such gear. I really do feel like it has more to do with production values that come with a certain level of gear than anything else.
joe 1008
Veteran
They were hinting that this issue might occur in the transmission and storage line. At some point the progressive material might end up interlaced:
https://www.fktg.org/node/6982/neuueberarbeitetete_technischerichtliniehdtv
This was the newest brochure I could find edited by german broadcasters (may, 2014). It might interest some of the german speaking fellows here.
Thomas Smet
Veteran
the 4:2:2 gets converted to 4:2:0 interlaced as well during broadcast.
Where the 4:2:2 does have a slight advantage and I mean slight advantage is when you start with 4:2:2 you have a full vertical resolution in the video. Because 4:2:0 interlaced is a really messed up format in how it alternates which vertical samples get tossed out based on the field there is a quality advantage with starting from 4:2:2. those Alternative field color samples will be slightly more accurate since it has the full vertical height of samples to work with. 4:2:0 progressive is missing the alternating samples entirely so the color can take a small hit. This is only for stations that don't have a way to broadcast progressive.
This is of course for HD. For 4k once again there really are not enough true samples from a single sensor camera to really make a big enough difference. I completely understand not everybody will always have the option to down convert in post but shooting 4k and down converting will get you video better than what a lot of internal 4:2:2 codecs Panasonic would consider using on this camera would give you. I feel your pain that you want an internal HD codec that can record 4:2:2 so you can pass off footage right away. It isn't as if shooters will not have any options however.
1. Record 4k and down convert for really good looking 4:4:4 that can then be saved to a 4:2:2 format of your choice and likely fool just about any engineer out there. The addition of 4k 60p on the DVX200 means you can down convert to 1280x720p 60p or 1920x1080i 60i with 4:4:4 color and deliver an optimum video to the broadcaster of your choice.
2. Use an external recorder that would likely be higher quality anyway since it could record directly to ProRes for the least amount of compression. External HD recorders are pretty darn cheap now and a lot easier to use now than ever before. Adding the new BlackMagic 1920x1080 video assist monitor with HD ProRes recording would be one such amazing tool to use with this camera.
If you are delivering 4k for 4k broadcast just up convert the chroma and export as a 4:2:2 file. I doubt most engineers will be able to tell the difference from a single sensor camera.
joe 1008
Veteran
Valid points. Bu why do all that trickery, if a camera for 2.000€, the PXW-X70 can do it all out of the box? It annoyes me, that the DVX200 which is potentially a much better camera, has such limitations. (Just to make it clear: I'm interested in the DVC200 as a HD camera. 4K for me can wait one or even two more years.)
Thomas Smet
Veteran
True but that codec is also limited to 50mbps IPB for HD. The DVX200 while 4:2:0 and 8bit can record up to 100mbps IPB or 200mbps I frame only HD. It may be lacking in color detail a bit but it makes up for it in a minimal amount of overall compression artifacts. Sony cameras with the higher end versions of XAVC tend to cost a lot more.
50mbps isn't bad for HD and used it a lot with my GH3 and it can look great. Sony is however packing in 4:2:2 and 10 bit into that same space which really eats into that 50mbps. Still a great format but 200mbps Intra from the DVX200 will crush it in terms of artifacts and clean motion. There are a lot of shooters, especially the run & gun types who move the camera around a lot and shoot things that move unpredictably, who will appreciate Panasonic taking the path of higher quality overall image compression as opposed to something they may not actually see unless they do heavy color processing. Different needs for different people and a different perspective on what a good video recording format is.
All I'll say is to beware of "headline" specifications. Bitrate only tells you part of the story and comparisons only really have validity if everything else is like for like.
As example I'd quote HDV and XDCAM EX. Both MPEG2, the former 25Mbs, the latter 35Mbs - but the latter gives far, far better results than the fairly small increase would lead you to believe - so what gives? Answer is "MPEG2" only tells a small part of the story, and in this case one big difference I believe is that HDV always allocates the same percentages of bitrate to I,P, and B frames. With XDCAM it's dynamic according to content - so with lots of motion, a bigger percentage gets allocated to difference frames, with static scenes a bigger percentage to I frames. And the extra encoder complexity makes a huge difference.
And XDCAM 422 is 50Mbs and 4:2:2 and is fully approved for broadcast HD acquisition. That's MPEG2 based. I find it quite conceivable that with XAVC being AVC based the extra efficiency will enable comparable quality and an increase to 10 bit.
As far as Panasonic go, then don't assume the results from any H264 codec will correspond to a seemingly equivalent AVC-Ultra spec (at same bitrate, same colour space, same bitdepth) just because both are based on the same underlying technology. I suspect a true AVC-Ultra encoder may have more "tricks" it can play with it's encoding than a H264 encoder found in a stills type camera - so get better quality for the same bitrate.
And the codec is only one spec amongst many. Don't judge any camera solely (or even mainly!) on codec aspects alone. The front end can often influence overall results far more than codec, and by front end I include the lens!
Thomas Smet
Veteran
I am well aware of how camera codecs work. I used to write video encoding software and specialized in live video streaming where we built our own streaming software and hardware. It is true that it is hard to directly compare encoding formats in cameras. The GH4 however uses the same video formats and have been proven many times to be extremely high quality H264 encoders. I agree that comparing one 50mbit H264 encoder to another 50 mbit H264 encoder will not always be equal but we are talking extreme bitrate differences here. It doesn't take a lot of understanding to realize the DVX200 codecs will have less compression artifacts during heavy motion and high detail scenes. You only need to look at how well the GH4 handles encoding with 100mbits and 200mbits intra to know this. I am not basing this on theory but from hard cold facts on how well the GH4 encodes video.
Look I'm not knocking the XAVC codec here. Without a doubt it is a good format and I swear by 50mbit H264 when used on my GH3. My only point is that there are pros and cons to the recording formats from both cameras.
Besides the jump from 25mbits to 35mbits mpeg was massive. Thats a lot of extra bits.
Barry_Green
Moderator
This is indeed a huge advantage and the first camera of this type. While the DOF could be shallower on a DSLR if you were to use a super-fast lens, the DOF on the DVX200 should be plenty shallow enough for most scenarios. Typically 35mm film was shot at generally an aperture of f/2.8 or slower; this camera has a sensor that's almost the same size as 35mm film and a lens that opens up to f/2.8 on the wider end, so yes, it should provide movie-style DOF.
The claim is true, and people have mentioned it. But there are those who think that the only way to get shallow-enough DOF is to use a full-frame sensor at f/1.4... for them, this DVX200 won't count as a shallow-DOF camera. But for those of us who got our start in 35mm film, but have migrated towards shooting events and live productions and corporates, the mix of M43 shallow DOF with an ENG-caliber lens is pretty much an ideal balance.
Barry_Green
Moderator
Depends on the frame rate, surely. In the X1000, it's 100mbps in 24p, and 150 mbps in 60p. But that's the X1000; I don't know what the final specs will be for the DVX200.
Wesley Byram
Well-known member
I really liked the way my old HMC cam operated, and the idea of a camera with similar usability that can achieve a big-sensor aesthetic is intriguing (and in 4k to boot). I'm wondering, how plausible is it that the aperture will stay at f/2.8 for at least some of the zoom range? If it could stay wide open at least until you hit 50mm FF equivalent field-of-view, then that's more than good enough for me.
I am wondering that too, as my HMC150 seems to only noticeably stop down near the end of the zoom range. Seems pretty consistent for the first half at least.