September 04 issue

2 replies [Last post]
Alan Roberts
Alan Roberts's picture
Offline
Joined: May 3 1999

I'd just like to point out that John Ferrick's otherwise splendid back-page piece has a glaring error in it. The comment about the high-end HD cameras is wrong, they are not uncompressed, they are not even full resolution, you have to go much higher than £40k for that.

The Sony HDW900 and 750 (and the 730S for that matter) are all 1920x1080 cameras, but they subsample the image down to 1440x1080 before compression. Not only that, they further subsample the chroma down to 480x1080, so the structure is 3:1:1. That data is then 4.3:1 DCT compressed for tape.

The Panasonic cameras come in two flavours, 720P and 1080. The 720P camera (AJ-HDC27F, known as Varicam) is 1280x720 but subsamples to 960x720 and then further subsample the chroma down to 480x720. The resultant data stream is then 6.7:1 compressed for tape. The 1080 cameras (several models in prototype) are 1920x1080 but subsample down to 1440x1080 for 50Hz or 1280x1080 for 60Hz, then further subsample the chroma down to 720 or 640 before data compression by 6.7:1 for tape.

Hardly uncompressed.

Fot more details on stuff like this, look in the Appendices in this document http://www.bbc.co.uk/rd/pubs/whp/whp034.html .

P.S. The addenda to that document are the menu settings for HD cameras that I derived for BBC use. They're in wide-spread use across the world, I've spent years getting stuff like this right.

Get my test cards document, and cards for 625, 525, 720 and 1080. Thanks to Gavin Gration for hosting them.
Camera settings documents are held by Daniel Browning and at the EBU
My book, 'Circles of Confusion' is available here.
Also EBU Tech.3335 tells how to test cameras, and R.118 tells how to use the results.

DVdoctor
Offline
Joined: Apr 1 1999

Alan of course is quite correct. What I was trying to say but did not translate to the written word correctly was that whe have two emerging standards for HD one as Alan correctly states in infact compressed but is at 100mbs data rate. The other, I probably should have said is Highly compressed since it is at 25mbs data rate, and does so using a relatively standard mpeg-2 compression scheme. The issues that are being debated are whether this highly compressed scheme with its long GOP's Group of Pictures HDV will offer sufficient enought quality for HD that it will be come accepted as a viable acquisition standard moving forward, or whether the difference in HD quality that the 100mbs systems with their newer codecs will be a barrier. My opinion is that just as DV with its initial debates regarding quailty compared to Betacam Sp and DigiBeta, HDV will be successful based on the cost/quality improvements that it will offer, the fact that it can use the existing tape format, and that transfer of the date at the 25mbs data rate is easily achieved.

Perhaps we need to have a new terminology for these two levels of compression .
I do remember people also correcting the misconception that Digibeta also was uncompressed.

John Ferrick

Alan Roberts
Alan Roberts's picture
Offline
Joined: May 3 1999

We do, indeed, need ways to discriminate between compression types. And we already have them.

Intra-field compression. This covers interlaced and PsF (Progressive with segmented Frames); Digibeta (88Mb/s), miniDV (25Mb/s), DVCAM (25Mb/s), HDCAM (144Mb/s), HDD5 (350Mb/s), HDCAM-SR(440Mb/s or 880Mb/s), DVCPro (25/50 for SD, 100Mb/s for HD interlaced). Each field of only half the frame lines is treated as a separate image, so motion portrayal is fine. Compression uses DCT data blocks, 4x4 or 8x8.

Intra-frame compression. This covers Progressive systems like DVCProHD in the Panasonic Varicam (100Mb/s). Each frame is treated as an entity, the camera runs at any speed between 4 and 60 frames/second, the tape runs at 60. DCT coding.

GoP compression. This covers MPEG2 transmission (Long GoP), HDV (25Mb/s short GoP), IMX/XDCAM (30/40/50Mb/s short GoP), MPEG4 consumer cameras (medium GoP), DVD (long GoP), web output. Lines can be taken either interlaced or progressive, but several frames are taken together as a group and much higher compression is possible because it the coder predicts frames using motion detection. GoP length varies; 8 or more for transmission, maybe 4 or even only 2 for acquisition. MPEG4 is taking over from MPEG2 (e.g. WM9) and uses better motion prediction. Wavelet compression is more complex but more efficient, and may replace MPEG eventually.

The issue in hardware isn't the complexity of the codec, none cost much in the global scale of things, it's only silicon. The real issue is media usage. Panasonic are dumping tape and going for solid state (P2), Sony are dumping tape and going for DVD. Both are much more limiting than tape (HDCAM-SR is 1/2" tape at 440Mb/s (4:2:2), and there's an option board to run it at 880Mb/s for stereo HD (2x4:2:2) or uncoded RGBkey (4:4:4:4)) at present, because of media capacity. The other alternative is hard-drive recording; feature films have already been made this way ("Attack of the Clones", "Russian Ark" and many more), and cameras are now being built with detachable hard-drive packs (Kinetta) or even on-board RAM (Tornado). P2 is nice because you can hot swap cards, so you get totally continuous recording; files are effectively already captured for editing, just plug the cards into PCMCIA slots but limited in saize toi about 4Gb at present. Hard-drives are nice because you just plug them into the editing system as extra drives, and size isn't an issue (1TB drives cost less than $1k).

The reason why we're getting so many new formats is because they all make sense in their own niche markets. None is the killer solution, each has merits and problems, the only certainty is that tape will eventually go away in the same way that crts will, but it will take a long time.

Hope that helps.

Get my test cards document, and cards for 625, 525, 720 and 1080. Thanks to Gavin Gration for hosting them.
Camera settings documents are held by Daniel Browning and at the EBU
My book, 'Circles of Confusion' is available here.
Also EBU Tech.3335 tells how to test cameras, and R.118 tells how to use the results.