Panasonic 4K discussion at InterBee 2013

It's worth mentioning that in terms of basic datarate, RAW is actually LESS than coded. (Depending on bitdepths and subsampling.)

For RAW, no of bytes per frame is no of pixels x bitdepth / 8 - so for quad-HD and 16 bit RAW - 3840x2160x16/8 = 16,588,800

For component it will be no of luminance pixels x bitdepth / 8 x 2 (x2 as in a 4:2:2 system there are as many chroma pixels as luminance) - so for quad-HD, 10 bit and 4:2:2 - 3840x2160x10/8x2 = 20,736,000.

So final filesize will be largely determined by the amount of compression used - whether RAW or coded. An AVC-Ultra or XAVC file is smaller than a RAW one normally not because of intrinsic factors, rather due to higher compression rates.
 
samsara said:
the new Varicam is probably the one with S35 and 2/3 inch mode
Click to expand...
There are a lot of s35 cameras already making use of B4 glass via adaptors - Alexas, F5/55s etc. Basic principle is that it spreads the image to cover a larger area, so no need for sensor cropping and no change in angle of view or zoom range. If you want to see a photo of just such being used on an Alexa in errr, rather difficult conditions by a BBC crew then go to http://www.bbc.co.uk/programmes/galleries/p01fy89z - image no 3! :)

Optically, as well as fully covering the larger sensor, then for any given setting of the lens, it will behave like a lens of about 2.7x the focal length, and with an f stop about 2.7 stops less than indicated. Hence the depth of field characteristic for any setting will be the same as when used on a 2/3" camera without the adaptor.

Normal first reaction is "I can't afford to lose 2.7 stops!! What about low light!?"

Don't worry. Bigger sensor means more sensitive so the effects cancel out. Move the lens from a 2/3" camera to an s35 without touching the iris ring and the exposure should be the same. (Assuming all else equal.) If you get correct exposure at ISO 500 and f2, you'll also get it at ISO 4000 and f5.6, if you want to look at it that way.

They're already being extensively used - http://www.lensadaptor.com/b4-23-sony-fz-package and http://www.sony.co.uk/pro/product/broadcast-products-system-cameras-lens-adaptors/la-fzb2/overview if you want two examples. There are others at various prices, some need you to also engage the lens doubler to work properly.

The point about s35 is that it gives more options for greater depth of field - without sacrificing deep depth of field. (You can always stop down.) And with 2/3" at 4K there is also the problem of diffraction limiting severely cutting back the available range of usable iris adjustment. Typically, you can only stop down a couple of stops from wide open before starting to get diffraction softening. So there are very sound reasons for Panasonic to have gone s35 for 4K, just the same as the other manufacturers - the laws of physics apply equally to all of them! :)
 
Last edited:
harddrive said:
I got myself curious and looked up the numbers for this. At 24 fps the Sony version of 4K RAW takes up about 430GB per hour, and I think RedRAW is somewhat less?

Yes, it's a lot more than the 110GB per hour that AVC-Ultra or XAVC would use, but a lot less than the 1.3TB he refers to that would be needed if no form of RAW compression was used. Assuming the same transfer rates, even the Sony form should then take around 45 minutes to transfer, instead of 2.5 hours.

RAW may not be for everyone, but with compression the systems in use don't take up the 1.3 TB/hr that uncompressed RAW would.
Click to expand...

4k (3840x2160p24) at 14bits per pixel would give you 1.3TB/h of uncompressed raw. So there is lots of latitude before even going to compression, simply going down to 10bit raw would get you down to 895GB/h, or 1.074TB/h for 12bit. Then a 3:1 compression ratio could let you hit 300GB/h, which isn't too shabby at all, especially if you can use a lossless compression scheme, and with CinemaDNG, you can use whatever compression scheme you want, even Zip or Rar if you want! JPEG compression is built in though.
 
vanvideo said:
A 4K Varicam for less than $10,000? I don't see that happening, as appealing as that sounds. I'm guessing more like in the $20K range, much like the F55. A regular HD Varicam is pushing into the $30K range.
Click to expand...

These are based on cameras announced/released in 2012. We are discussing about a 2014 release. So we should expect a lower price.
 
RAW is not for broadcasting, that's for sure
but RAW is cheaper to implement, just do a 14-bit-linear to 10-bit-log (as atarijedi said), then some simple compression (8:1 CineForm RAW is a lot easier to implement than 2:1 lossless DNG), and you're back at the same bitrate of AVC-intra-100 (110GB/hour)

So this is what I see in the future: broadcast cameras will still be very expensive, cine cameras not nearly as much
 
harddrive said:
It's worth mentioning that in terms of basic datarate, RAW is actually LESS than coded. (Depending on bitdepths and subsampling.)

For RAW, no of bytes per frame is no of pixels x bitdepth / 8 - so for quad-HD and 16 bit RAW - 3840x2160x16/8 = 16,588,800

For component it will be no of luminance pixels x bitdepth / 8 x 2 (x2 as in a 4:2:2 system there are as many chroma pixels as luminance) - so for quad-HD, 10 bit and 4:2:2 - 3840x2160x10/8x2 = 20,736,000.
Click to expand...

I have a question about raw - for example C500 raw (4K raw interpolated from 4x1920x1080 pixels) is stored before debayering and needs all effective pixels from CMOS chip, it is not 2x 1920x1080 but 4x1920x1080 in this case. Sony FS700 has also 8mil ef. pixels for 4K, so I think that storage of raw is always about storing of all effective pixels for video from chip. Is it OK or I'm wrong with this count of pixels for raw?
 
Pave1 said:
I have a question about raw - ......I think that storage of raw is always about storing of all effective pixels for video from chip. Is it OK or I'm wrong with this count of pixels for raw?
Click to expand...
That's correct - and is the figure I used in the section you quote - "For RAW, no of bytes per frame is no of pixels x bitdepth / 8 - so for quad-HD and 16 bit RAW - 3840x2160x16/8 = 16,588,800"

So in the above it's using 3840x2160 as the number of sensor photosites - (2x1920)x(2x1080) or 4x1920x1080, as you say.
"
 
Thanks, counting is not my the best ability (and my english is poor .-) ). I was confused by the exact count cca 8,5Mp for 4K, because cameras have different count of the effective pixels for 4K video (and this number is very low). Ideal is 3x4096x2160 (26,5M pixels, *16/8=53MB for one frame) for 4K 444 in RGB (no debayering - interpolation). So it is right that for the 4K 422 result it is 16,5MB as you wrote if you count only one 4096x2160=8,5Mp in 16bit (and for YCC422 10bit video storage it is cca 4096x2160x10x2/8 = 22,1MB without any compression from the same 8,5Mp).

4K 16bit raw (for 422 result) is 16,5MB and 10bit YCC422 storage is 22,1MB (without compression).
C500 and FS700 make 4K (26,5Mp for 3x 4096x2160) by interpolation from cca 8,5Mp (1x 4096x2160).

It looks as really low count of pixels for 4K, it is more like 420 than 422 chroma subsampling.
I think that "middle class" 4K must have about 2x4096x2160 pixels on the chip (16Mp) for good 422 4K,
in this case raw is about 35MB for one frame (16bit) vs. 22MB YCC422 (10bit), this has more sense for me.

So this kind of math is based on the very low input count of pixels.

Thanks for answer, it is clear for me now.
 
Last edited:
Samuel H said:
RAW is not for broadcasting, that's for sure. ......

So this is what I see in the future: broadcast cameras will still be very expensive, cine cameras not nearly as much
Click to expand...
By "broadcasting" I think it's important to distinguish between pre-recorded and live events.

And it's the latter where 4K is likely to be most important first. (A live 4K broadcast of rugby was shown at the recent IBC, and a lot has been said about 4K live broadcasts from next years World Cup, certainly the final.)

Consequently, the camera body is one thing - just as important are how the signal is relayed from camera head to OB truck, and such a system must also deal with talkback, video return, tally etc etc. We don't know what Panasonic have planned in this respect yet, but it's likely to be similar to the products that Sony are marketing to allow the F55 to be used as an OB capable camera - see http://www.sony.co.uk/pro/product/b...camcorder-adapters/bpu-4000/overview#overview

That system allows remote camera operation over fibre cables up to 2,000 m away, and I believe that what travels down the fibre cable is the RAW signal. There would certainly be advantages in doing it that way - gain, colour control, etc would all be done at the base end, rather than sending control signals up the fibre. That should help reduce latency in the camera controls. They also say that the base end will output 4K and HD signals simultaneously - again, more satisfactory if they can both be derived directly from RAW in the truck.

On the right hand side of the link above there are other links which give information about the experimental FIFA broadcasts. If nothing else, they indicate that once you get into thinking of live broadcasting of 4K, there's a lot more to it than just the camera head!
 
^

I think it is RAW down the fibre as well. From what I read and understand the CA-4000 bolts onto the F55 taking the "RAW" output from it to a CCU, at which point the CCU can provide four streams of QFHD 4k.

RAW sensor data is 100% the better way to do it, I agree.

I hate to give any camera maker any credit, but Sony really thought this one through. A HD/2k/4k internal codec camera, with the option of 16-bit RAW recorder, with the option of 4k fibre broadcasting capabilities.

We can argue all we want about 4k being here or not - but 4k for live events seems to be catching on.
 
Last edited:
@ harddrive & starcentral: that's waaaay beyond me, I have absolutely no idea how that works, I'm just an amateur with way too much time to spend on these things


@Pave1
By recording RAW, you're usually* recording all the information that came out of the sensor. There is no option that will give you any better results than that.
Modern debayering algorithms are very good, a 2K sensor will give you a relatively sharp 2K image, with maybe >800 lines of real resolution.
But yes, since each pixel gets information for one color channel only, there's a lot of interpolation going on, and you can think of it as an in-built color subsampling. It's not the same kind of subsimpling as 422 or 420, though, so, storing your 2K RGB image debayered from a 2K sensor as 444 gives you better color information than if you store it as 422 or 420. But if you want to think of it as somehow-below-422, that would be fine in my book.

* one exception: the 5D3-RAW. It seems it reads one of every three rows, then averages every 3 horizontal pixels of the same color, and this way gets a 1920x1080 Bayer image (a single 14-bit value for each point in that array). If you record RAW, this array of 14-bit values is stored in a very fast CF card, and debayering happens later on, at your computer. If you record H.264, this array gets debayered in-camera, the resulting 1920x1080 YUV image is then compressed and encoded as H.264. I think the softness of the H.264 video comes from the relatively bad debayering that the camera performs.
 
Last edited:
Samuel H - YCC422 color subsampling is created after debayering from raw. And you need one Y 4096x2160 and two color chanel at half resolution (both together have the second 4096x2160).

If camera stored raw only from one 8.5Mp (one 4096x2160) than debayering must to create too much pixels and for example a chroma key may not to work properly. It really isn't 422 signal (after debayering), as we know.

At HD it is like Canon XF100 - one 1/3" chip with 2mil pixels (1920x1080) and 422 record. For me it is not "middle class" HD, but low quality HD (it is ok in this case because XF100 is only B camera for XF300 and 422 record has for the same workflow with XF300).
4K cameras with 8,5Mp is like this one, it is low end 4K, not "middle class" 422 4K. But it is only by product of my answer, I want to point that correct 16bit raw is bigger than YCC422 10bit signal - only case where this is not correct is if the camera has a very low pixel count from the chip. For 422 4K you need 2x4096x2160.
 
Last edited:
Samuel H said:
@ harddrive & starcentral: that's waaaay beyond me, I have absolutely no idea how that works, I'm just an amateur with way too much time to spend on these things.
Click to expand...
I think you're too hard on yourself! :) The principle is not that difficult, you already seem to know the hard bits!

Recording in camera is one thing. Basically either record the data straight from the sensor (RAW) or process it to form a "ready to watch" signal in camera, then record it, nearly always with some compression to reduce the datarate. The point that the Panasonic interviewee in the original link doesn't acknowledge is that in most current systems it's normal to also compress the RAW before recording, so practically, recording RAW is much easier than he implies.

Move to live work and it's obviously necessary to get the signal from camera to OB truck in order to vision mix it. Original methods would be to send separate R,G,B analogue signals back from the camera head and then do processing and code the PAL or NTSC signal in the truck - in general, this was when the truck-camera head cables were multicore and may have had three or four separate video cables in each cable. [Digression for anyone interested :) In the early colour days, the norm was G101 - cable two inches thick! There's a description of a famous OB of the time at - http://www.vintageradio.co.uk/htm/tvprojects4b.htm - and for just a quick quote: "A 100ft reel of G101 cable weighs over a hundredweight (100Kg approx.) and takes two strong men to lift. Each cable coupling is a potential problem, even in the dry, and this was half way up a mountain." ]

Move on, and there was a desire to use cheaper, simpler, and more lightweight cables which would also permit longer cable runs. Triax was one solution - like coaxial, but with two shields instead of one around the central core, this feature enabled it to also safely carry power up to the camera. An implication was that all the video processing and PAL/NTSC coding happened in camera, and what came back down the cable was a coded signal - not separate R,G,B. Had many advantages - but a disadvantage in having the processing in camera was that control data to adjust camera parameters then had to be sent up the cable, which could lead to a slight (but noticeable) delay before taking effect.

Next came digital - and expect SDI/HD-SDI down the cable instead of PAL or NTSC. Also fibre started to replace triax.

But from what I hear now, if Sony are sending RAW down the fibre, we'll be going back to many of the advantages of the early systems, whilst keeping all the benefits of later ones - that's why it's such a big deal. All the advantages of simple, lightweight cable - but it's the modern equivalent of still being able to access the separate R,G,B signals as in the analogue days.

Having access to the camera RAW in the OB truck opens up all sorts of possibilities - decoding it differently for HD or 4K (or both together) for example. It's also likely to minimise delay between the actual event and the signal arriving at the mixer. That can be important in a live OB. It also opens up the possibility of having 4K cameras in an otherwise 1080 OB and being able to choose (at full 1080 resolution) a portion of the image.

I think starcentral sums the F55 up very well with "A HD/2k/4k internal codec camera, with the option of 16-bit RAW recorder, with the option of 4k fibre broadcasting capabilities." As far as this thread is concerned, then we know we can expect an s35 4K camera from Panasonic sometime in the future, we know it will record HD/2k/4k via internal codecs in camera - but the big questions are whether it will have any RAW recording ability, and what it's fibre broadcasting capabilities will be. And it's because of the latter that RAW capability really matters for a camera geared more towards broadcast than "cinema", at least when used cabled.

It would be possible to process the signal in camera, and send that via fibre - but the datarate is likely to be greater than even uncompressed RAW, and it won't be as satisfactory.

I don't really want to get into the "who will buy it" debate again here, but certainly in the UK 4K is getting serious attention within the industry - http://crave.cnet.co.uk/homecinema/bbc-and-sky-to-promote-4k-tv-with-uhd-ready-sticker-50011876/ . Personally, I think widespread adoption is a while off - but 4K is already being used for a lot of high end prerecorded material, and it's in the live event and sport area that we can expect to see increasing interest, especially for prestige events. A lot has been said about next years football World Cup, but probably it's the 2016 Olympics where it could become very significant. There were limited UHD trials at the 2012 Olympics, I expect to see a lot more in 2016. For example, http://sportsvideo.org/main/blog/20...-at-8k-plans-for-pyeongchang-2018-tokyo-2020/ talks a lot about 4K for the major events, and is reporting keenness for 8K for the 2020 Tokyo Olympics. I'll believe the latter when I see it, but at least for 4k, you can see why manufacturers are being pressured to provide live capable 4K kit sooner rather than later.
 
Last edited:
Look forward to seeing what the new varicam is capable of and also AVCUltra. Also very interested to see where this will all wash out as far as price goes... and of course now that Pana have produced a 4K sensor they “are very proud of”, what that means for the rest of their lineup.

As far as bit rates and compression go - show me results not numbers, personally I couldn’t care less about them.
 
harddrive said:
Personally, I think widespread adoption is a while off - but 4K is already being used for a lot of high end prerecorded material, and it's in the live event and sport area that we can expect to see increasing interest, especially for prestige events. A lot has been said about next years football World Cup, but probably it's the 2016 Olympics where it could become very significant.
Click to expand...

- for country with large population and large TV market is 4K good option for TV sports events, but for smaller countries is problem to make fast replacement of equipment in this digital times. Smaller TV broadcast company must make a lot of investment for making of HD programs, it is not only sports events and live broadcasting, but also many documents, entertainment, educational programs, etc. This mean big amount of storages at postproduction, a lot of quality optics for cameras, etc.

So I really think that you must differ between lowend HD or 2K or 4K programs and programs with broadcast quality. At this level 16bit 4K raw (for 422 color subsampling) never be smaller than YCC422 10bit (if is then there is some quality problem), it can happens only at lowend cameras.

So I understand to market 4K option for manufactures but I also see a lot of problems. For example LCD or plasma TVs are moustly 1meter or more wide now, for 4K or 8K you need bigger screen (2,5m) but a lot of people has small living rooms, where is no needs for this big screen. Broadcast TV on smaller TV market will be much longer at HD production than Sony, Panasonic and others manufactures think.

Give us a cheap and fast storage and possibly that will be the first step for global Europe transition to 4K.
 
I think that things have changed so drastically in the last couple of years that we have all become total camera brats.

"I want RAW"
"I want it to only cost $2,000"
"I want ro record to $5 SD cards"

:):):)
 
I said before that I'm not going to further consider the home viewing aspects behind 4K, either now or in the future, in this thread.

What's relevant to this discussion is what the TRENDS are likely to be in broadcast over the next few years.

That is NOT the same as what will be being used - by and large, such equipment might have a lifespan of typically 10 years on average, so even if all the equipment sold from now on was exclusively 4K, that still predicts that even in 5 years time 50% of material will still being produced in 1080! Even after no 1080 equipment has been sold in all that time!

OK, that's a bit flippant, but it's intended to illustrate that manufacturers and buyers do well to think ahead. Not think of just what is needed now, but have some growing room in any new purchase. If you're operating a 1080 "only" OB at the moment, fine, but if you're about to build a new large OB truck I'd say it's highly sensible to build a level of 4K capability into it from the start - even if it starts life with 1080 only cameras etc at the beginning.

And all the "will we see the difference?" arguments are totally irrelevant when we consider events such as the Olympics, World Cup, even ones such as Wimbledon etc. They have traditionally been where new TV technology has been tested in a real environment well before going into everyday broadcast, let alone being "the norm" viewed in most homes. That's been true of colour television and high definition in the past, was even true of television at all in 1948 for the London Olympics then! A lot was made then of the television broadcasts happening - even though there weren't many receivers in the whole of the UK. Read the link I posted above and the same is true of 4K and the equivalent events of the next few years. There will be significant production of 4K, even 8K - end of story.

Noel Evans said:
.......show me results not numbers....
Click to expand...
As far as this camera is concerned, then I don't have any doubt that the results in terms of picture quality will be very, very good. Same can be said of the F55. Differences are more likely to be down to lineup, than differences in the hardware - not forgetting lighting, operation etc etc. And as far as AVC-Ultra (and XAVC) go then for all practical purposes forget them. They will accurately record and reproduce what the front end of the camera produces. What more is to be said?

So what is left is more likely to be down to features - and pricing. Key amongst the former must be RAW capability - that won't necessarily give any better quality than directly recording to AVC-Ultra or XAVC, but it will give far more grading and post flexibility. And one set of the early 4K adopters ("cinema") are likely to be willing to pay for that and accept higher bitrates etc. And since the other main set of early 4K adopters are likely to be the live event, sport, etc sector the other set of features of most interest must be those to do with live usage capability.

Almost since television first started there has been a divide between cameras used for "live" broadcasting (via CCUs into a vision mixer), and those for prerecording. In the early days the divide was total - electronic for the former, film for the latter. As electronic cameras progressed, the lines got blurred - ENG cameras could internally record, but also had a live capability via a links truck. But there still remained a line between cameras intended primarily for studio/OB, and those for self-contained use. Now we may be seeing that line totally disappearing. For the first time a camera equally at home for live or prerecording usage. The advantages for both manufacturer and buyer (in the broadcast world, at least) are huge in terms of operational flexibility without compromise of features.

The F55 has shown the way - it's as starcentral said earlier "A HD/2k/4k internal codec camera, with the option of 16-bit RAW recorder, with the option of 4k fibre broadcasting capabilities." From what we've already heard, then in certain key features the new Varicam matches it - the huge question remaining is whether it can do the same in the aspects we don't yet know about. Alternatively, if it can't (no RAW or live capability, say) will it significantly undercut the F55 on price?

To R Hudson - I don't think any of the three things you mention are at all unreasonable to ask for........ It's asking for them all at the same time that IS unreasonable!! :) :)
 
You must log in or register to reply here.
Back
Top