Reader Peter Greed emailed me recently, drawing my attention to a Highdefdigest forum discussion on the value of 1080p24. This thread hangs off a pretty good article on 1080p24 from Joshua Zyber on the subject, entitled ‘What’s the Big Deal About 1080p24?‘.
I recommend this article to anyone wanting to understand 1080p24. This can be a difficult issue to get your head around, so reading a few different takes on it is generally useful.
However, I differ on one point made in the article:
In a best case scenario, when a player that offers 1080p24 output transmits the signal over HDMI to an HDTV that can accept and sync with the signal, the improvement over standard 1080p60 will still be very subtle. You may have to strain to find it, or do comparison tests of certain movie scenes at both frame rates. If you don’t have equipment that will support 1080p24 video and you’ve gone this far in your life without ever noticing 3:2 pulldown judder, this may one of those things better not to worry too much about. [my emphasis]
This probably is the case with North American and Japanese viewers. After all, they have lived all their lives with the strange video cadences necessitated by the 60 hertz frequency of their TV systems. However, we here in Australia (along with Europe, India and plenty of other places) are used to 50 hertz systems. Due to odd sourcing of some video material, this has generated its own weird set of problems (ie. conversion from natively NTSC to PAL, and in the case of some older movies, our PAL versions were converted from NTSC, creating the worst possible image!)
For good quality film transfers, the technique has been simply to speed up the film by about four per cent so that each film frame (which runs at 24 frames per second in the cinema) is shown at the native PAL rate (ie: 50 fields per second interlaced, which translates to 25 frames per second). Consequently we are used to — have been raised our whole lives on — smooth camera pans and action.
When I first wrote about 1080p60 vs 1080p24, it was just after I had seen Blu-ray in action for the first time. That was in September last year. I wrote:
But there was one major problem with the picture on both display devices: the pans were jerky.
At the time, I didn’t even know if we would be getting special 50 hertz version of Blu-ray discs, or the same 60 hertz versions as supplied in the US (it is the latter). But this jerkiness was really, really obvious to one used to smooth PAL motion.
UPDATE (ten minutes later): I just realised that Peter was actually after some comments on a specific post on that thread. In that post the author argues that 1080p24 is overhyped, suggesting that it is unnecessary because of advanced motion compensation techniques employed in many modern display systems.
Actually, this is irrelevant to the issue.
A 1080p60 HD DVD or Blu-ray player takes the four progressive film frames recorded on disc — A B C D — and outputs them as A A B B B C C D D D. That is what causes the jerkiness: every second frame is shown for 50% longer than the intermediate ones. There is no reason why a high end display should not process these incoming frames by going back to 24fps (dropping all the repeated frames, bringing the video back to A B C D) or, say, going back to 48fps (dropping the third frame in all triple repeats, so the video becomes A A B B C C D D), or even increasing it to 72fps (all frames repeated three times). That is what my DVDO iScan VP50PRO can do. But this is nothing to do with motion compensation. Instead, this involves robust cadence detection to determine which frames are being repeated and how often.
Motion compensation is a subset of the better processes used for deinterlacing interlaced material, not correcting uneven frame cadences. With interlaced source material, a good motion adaptive deinterlacer will determine which bits of the picture are in motion, and which are static between fields. It can then simply weave the static portions together to increase the resolution of those portions of the image. This post, and the one immediately below it, show the marked difference in picture resolution provided by good motion compensation.
However, the moving portions of the picture (ie. moving between fields, not frames) will still need to rely on traditional deinterlacing techniques, such as bobbing. Good circuits use edge adaptive algorithms to eliminate jaggies in the display of diagonal edges in the moving portion of the image.
The best deinterlacers can even detect the cadence of 1080i60 material (in which the frames are split up into fields and some of the fields are repeated more often than others) and, using motion detection, reconstruct excellent 1080p60, and in some cases, 1080p24. The problem with this is that from time to time there is ambiguity in images.
In the PAL DVD world, I use a few test clips, gathered by me over the years, not from any particular test DVD, to check out deinterlacer performance. One of these is incredibly ambiguous and tricks the automatic deinterlacers from DVDO, from HQV, from Gennum, from Pixelworks and from Faroudja (plus, of course, the many unbranded deinterlacers). That’s why I like to insist that all deinterlacers should have a ‘force film mode’ option, although few do.
By outputting the video as 1080p60 a player should, if the video is flagged correctly on the disc, produce perfect results (aside from jerkiness). By outputting the video as 1080p24, there is no qualification on the perfection of the results.