The first digital acquisition format most of us worked with was DV, which has a resolution of 720x480 and uses a discrete cosine transform-based codec. This means that each frame is compressed without reference to any other frame. Also known as intraframe compression, this type of codec is very simple for NLEs and computer processors to decompress and edit. It offers high video quality, though at 25Mbps, and the data rate requirements relatively hefty. After DV came several professional formats that used the same underlying compression technology but different tape formats. Sony’s DVCAM uses the same DV codec and MiniDV tape, but it spins the tape faster and records a wider track, reducing the potential for dropout and allowing the tape to be edited using a frame-accurate deck.
DVCPRO is Panasonic’s professional DV format. It uses proprietary tapes with an even wider track pitch, metal particle rather than metal evaporated composition, and some additional technical features to simplify deck-based editing. DVCPRO records at the same 25Mbps data rate as DV, though Panasonic also offers DVCPRO 50, which doubles the data rate, increasing quality.
Those were our choices until HDV arrived. In contrast to DV, HDV uses MPEG-2 compression, a so-called LongGOP format. This means that each group of pictures (GOP) is comprised of I, B, and P Frames. The I Frames are complete frames (like the DCT-compressed frames used in DV), while the B and P Frames reference redundancies in other frames to increase compression. Using MPEG-2 allows camera vendors to squeeze HDV video into the same data rate (and MiniDV tape format) as DV, yet have a frame size containing roughly 4.5 times the pixel area. However, MPEG-2’s LongGOP structure complicates editing, since you may have to decode multiple frames to display a single frame. You probably know that 1080i HDV has a display resolution of 1920x1080 but a capture resolution of 1440x1080, for a pixel aspect ratio of 1.33 to 1. This means that during HDV editing and playback, the horizontal pixels are stretched by 33%.
AVCHD is a lot like HDV, except it’s more freeform. For compression, it uses AVC (aka H.264), another LongGOP format, with data rates ranging up to 24Mbps, about the same as HDV, though most current devices store video at much lower data rates. Since AVC is a more sophisticated algorithm than MPEG-2, it’s tougher for most computers to decode, which is why Premiere Pro doesn’t edit AVCHD today and why Final Cut Pro converts AVCHD footage to ProRes422 during ingest.
I call AVCHD "freeform" because although it displays at up to 1920x1080, it supports capture at a variety of resolutions. My Panasonic AG-HSC1U shoots 1440x1080, while Sony’s new HDR-SR10 captures at true 1920x1080. This is the first true pixel-for-pixel HD camera I’ve seen.
Now let’s have a look at Panasonic’s AG-HVX200, which captures in DVCPRO HD, Panasonic’s high-definition format. At 1080i, DVCPRO HD uses four parallel DV codecs (I Frame only) and has a video data rate of 100Mbps. However, the underlying video has a resolution of 1280x1080, for a pixel aspect ratio of 1.5, compared to 1.33 for HDV and 1 for some AVCHD solutions.
At NAB this year, Panasonic will reportedly show a camcorder with AVC-Intra technology (AVC-I) that enables full-raster high-definition recording—1920x1080 pixels for a 1920x1080 display image, with the AVC codec enabling the higher resolution at the same 100Gbps data rate as DVCPRO HD. NAB will also see the introduction of 32GB P2 cartridges with the promise of 64GB cartridges shipping in the fall, which will translate to a little more than a minute of AVC-I (or DVCPRO HD) per GB per cartridge at the highest resolutions.
Up through HDV, most editors worked pretty similarly with all formats because the DV and HDV formats were simple streams of audio and video. With P2, however, Panasonic stores data on P2 cards in MXF (the Material eXchange Format), an XML-based structure that separates out audio, video, thumbnail images, and metadata. Not all NLEs handle MXF similarly, and some can’t even write MXF data without third-party plugins, which is critical if the metadata is important to you.
Back in the DV days, life was simpler. Now you need to be conversant in LongGOP, MPEG-2, AVCHD, and AVC-I, and you must know the raster and display resolutions (not to mention bit depths) of all of the candidates to choose your HD format. If you believe that tape is on the way out, as I do, MXF should be on your radar screen. You should also make sure your NLE provides the optimal platform for editing MXF content.
Jan Ozer (jan at doceo.com) is a contributing editor to EventDV and Streaming Media.