PTravel

The problem isn't one of compatibility -- mpeg2 and mpeg4 are both standards. The problem is one of temporal compression -- temporally-compressed video (at least standard-definition video -- more on that later) is intended as a delivery medium, not an acquisition medium. Temporal compression works by identifying a reference frame and then storing only calculated changes between the reference frame and subsequent frame, as contrasted with non-temporally-compressed video codecs, e.g. DV-25 as used in miniDV, which compresses only within a frame, but not between frames. Temporal compression is very efficient as, unless a camera is panning, many elements within a scene will not change from frame to frame.

The problem, though, is when you want to edit temporally-compressed video. Because most frames are stored as the accumulated delta of the frames before it and the reference frame, it is very difficult to make frame-accurate cuts, i.e. either the cut must be made exactly on a reference frame, or all frames prior to the desired cut have to be uncompressed back to the most recent reference frame. This takes a fair amount of computing horsepower, and is also complex enough that many editing programs don't bother supporting temporal compression (there are, however, some entry-level programs that do).

There is an additional problem when you want to do true editing, e.g. adding titles, transitions, color correction, effects, etc. As with simple cuts, the effected frames have to be uncompressed. The effect or transition is then applied, and then the frames must be recompressed. Mpeg2/4, like many video codecs, is a lossy compression format, meaning information is lost through each transcoding. Editing mpeg in this fashion results introducing an extraneous uncompress/transcode step each time an edit is made, resulting in data loss each time. The net result is degraded video that can only be competently edited on very powerful machines. I edit in Adobe Premiere Pro on my editing computer (3GHz P4). This setup can do anything the studios can do when I'm working with DV-25 material. Premiere Pro can edit mpeg2 with the proper plugins but, when I have, it's dragged performance down to the the pull-out-your-hair level.

There are additional reasons why I recommend against hard-disk camcorders. These machines, at least in the consumer standard-definition lines, use DVD-compliant mpeg2 (or its mpeg4 equivalent) for their "high quality" setting. DVD-compliant mpeg2 is limited to a maximum bandwidth of 10 megabits per second (mbps). Lower quality settings have even lower bandwidth. MiniDV, which uses the DV-25 non-temporal standard, has a bandwidth of 25 mbps. Though mpeg is somewhat more efficient at compression than DV-25, the latter still has 250% more data throughput, resulting in significantly higher-quality video than mpeg. This difference is easily seen when you burn a DVD from a DV-25 video source. I use software transcoders that do multi-pass analysis to very high precision when I prepare DVDs from my DV-25 material. Running on a 3 GHz P4 with 1 gig of RAM, it can take as much as 24 hours to produce 2 hours of transcoded mpeg2 at the highest quality settings. Even then, the difference between the final DVD and the miniDV original is significant and easily seen. Consumer hard-drive (and DVD) camcorders do not do multi-pass analysis -- they're restricted to on-the-fly single-pass transcoding and, of course, are using far less powerful processors. The net result is video of much lower quality than can be obtained from a miniDV machine. This isn't to say that there aren't lousy miniDV machines out there -- video quality is also a function of lens quality, sensor size, electronics, etc. However, all things being equal, miniDV will always produce dramatically better video than hard-disk or DVD camcorders.

All of this changes, however, with respect to HD machines. Consumer HD is all temporally-compressed, so there's no codec-based advantage to one storage format over another. However, which codec is employed, as well as at what bandwidth, is critical. Sony makes consumer and prosumer machines that use the HDV codec, which is an mpeg2-based codec that has a 25 mbps bandwidth, just like DV-25 (though it has to store considerably more data, being HD). However, Sony also makes AVCHD-based machines which use an mpeg4-based codec. Though, theoretically, these machines are more efficient and should produce better video, Sony has arbitrarily limited the bandwidth of AVCHD to between 12 and 17 mbps. This is simply not enough for good HD reproduction, and the AVCHD machines demonstrate severe motion artifacts as a result.

For HD machines, I still prefer those based on miniDV because tape is self-archiving. However, I wouldn't consider an AVCHD machine, and I'd be very selective about HDV machines -- too many have small sensors and abysmal low-light performance.