Why Digital TV Won't Kill Discs Soon

It has been theorized that digital TV and downloadable video will rapidly destroy the DVD market, to the extent everyone has broadband. I don't think this is going to happen as rapidly as people thing, because even ignoring some of the other economic advantages of physical media (like the sense of ownership), the economics of the bits that make up the video are different for discs and broadcasting/downloading in some important ways, resulting in a quality advantage for the discs.

There are two primary advantages of digital video over analog video:

1. As it is digital, you can layer a wide variety of error correction, variable formats, metadata, alternate streams, and other such things into the signal, and it can be perfectly stored and re-transmitted without loss.
2. Digital data can be compressed with far more sophisticated algorithms than you could ever hope to compress an analog signal with.

The primary benefits touted for the consumer comes from the first point, but that's a cost center for content producers, who need to produce the extra content (be it higher-resolution source video, extra audio streams, etc.), and a cost center for the broadcaster, that needs to allocate scarce bandwidth for that content.

The primary benefits for the broadcaster are the second point; the same amount of analog bandwidth that previously carried one television station can carry several standard-def stations, or a high-def station, or a whole bunch of audio streams, or whatever else suits their fancy. (Digital bandwidth is fungible; as long as you stay under your bandwidth limit you can send 1000 audio channels or 20 video channels or anything in-between.)

Careful readers will note that the content producers don't get benefits either way; they just get higher costs. In a way, though, the game doesn't change much; profitable shows and network will survive and unprofitable ones won't. (The theoretical advantage of getting more viewers with a higher quality signal is nullified once everybody is offering that higher quality.)

Digital video, like digital pictures, can effectively be compressed to an arbitrary degree. For instance, see this famous nebula picture. I took the 6.51MB original TIFF image, which is uncompressed, and squeezed it down to 25 kilobytes, which is about as small as I could get it before it appeared to consist of nothing but blocks. (As it is, there's still a lot of blocks, but the main detail of the nebula has at least some definition...) Your browser may scale that image, be sure to look at it in its full-resolution glory. That's a compression factor of around 260. For reference, that's within 25% of the size of the HTML text comprising this webpage.

On the Hubblesite.org page, they offer you a 180.6KB JPEG that looks a lot better. Gimp, the program I used to create my JPEG, allows me to offer you up to a 1,230 KB JPEG image, which is virtually indistinguishable from the source image to the naked eye. A JPEG is obviously not an MPEG, but the compression technique of a standard JPEG and the compression used in MPEG 2 work in very similar ways.

Obviously, the harder you compress an image, the worse it looks, but nothing else stops you from compressing the video ever tighter.

And this is where the economics of discs vs. the economics of using scarce bandwidth in realtime leaves discs a niche. When using bandwidth to transmit video, there is an extremely strong incentive to compress a channel as much as possible, then a little bit more, so you can stuff as many channels down the wire as possible. As a result, the general quality of the image is usually significantly degraded; even though "normal people" can't say why the digital channels seem to lack the pop of the analog channels, the feeling is still there.

Despite the promise of higher quality and higher resolutions available with digital, so far, the exact opposite has occurred. Many signals are clearly degraded and inferior to their best analog equivalents. Those that are not, probably will be very soon. (I don't know anyone with an HD TV and HD service, but my guess is that while HDTV looks pretty good now, while there aren't all that many channels competing for the bandwidth and the vendors need it to look good to attract customers, that HDTV signals look very good. As HDTV goes from the exception to the norm, I expect these economics to play out in the HDTV arena, if they aren't already.)

Further evidence of this logic can be seen in the digital radio domain. I've heard both of the digital radio services and to my ears both of them have strong digital artifacts in all channels. The speech-only channels are total garbage, although at least the content doesn't suffer, but all the music channels sound pretty bad too; the claims of CD quality are unmitigated bullshit, it's much closer to audio-tape quality. Over-the-air HDRadio, which is what seems to be getting deployed at least in my local area, transmits FM at 96kbps, and does horrible, horrible things to the signal:

The HD Radio codec, responsible for transforming the analog audio to and from the digital domain, is a hybrid codec similar to the mp3pro low bitrate codec. All audio below about *5Khz is encoded in the usual way to be reproduced on the receiver end. Higher frequency audio is artifically reproduced by the receiver with the help of some additional information encoded into the audio bitstream. Together these two techniques provide the listener with the perception of a fairly wide audio bandwidth at the expense of accuracy on the higher frequencies.

Emphasis mine. You would not be too far wrong to read that as "when it comes to high frequencies, this technology just makes stuff up." I'll admit I haven't heard the local HD Radio, but between that description and the fact that I've never really heard a 96Kbps compression technique of any kind that didn't suck the life out of music, I'm not terribly excited about it.

Contrast the bit economics of broadcasting with the bit economics of discs. Once you have chosen which kind of disc to create, that is, single-layer, dual-layer, etc., you have a fixed number of bits to work with, and you might as well use them all. At least based on my collection, dual-layer DVDs, which offer around 9GB, are pretty standard now, with movies clocking in at around 5-8 gigabytes (leaving some room for the extras). Despite the fact that DVDs only offer a resolution of around 720x480 (there are some variants but nothing radically higher than that standard resolution), which is merely 345,600 pixels, they are 345,600 good pixels. I often watch DVDs full screen on my laptop, which has a resolution of 1680x1050, or just over five times more pixels than a DVD, and I tend to only really notice the video artifacts if I really look for them, or in a few rare situations. (MPEG really doesn't like water or static.)

If you blow a DishTV channel up to the same size, the compression artifacts are readily visible to the naked eye. In fact, you can't hardly help but see them once you know what you are looking for. I'm not exactly sure what resolution DishTV transmits at, but the pixels they transmit aren't good pixels, they're just barely adequate if you don't push them.

Thus: Discs will continue to have a niche for home theaters, and people that perceive that broadcast digital video is inferior to DVDs/HD-DVDs/Blu-Rays, even if they can't explain it in technical terms. I call this a "niche", but it will probably be a very large one.

Meanwhile, if you don't really need high quality, or you don't care about it, and you prefer quantity, the digital broadcast services aren't going anywhere. And for as bitchy as I may sound here, that's a viable solution. What's bothering me isn't so much the squeezing of channels, but the marketing that claims this is done without cost.

An interesting middle case: there are the on-demand or download services, which fit in between these two extremes, especially if you don't implement instant movie startup. Because they are still using scarce bandwidth they are still somewhat subject to the economics of broadcast video, but because they don't necessarily have to be in real-time, it is possible to download a higher quality video stream. Unfortunately, services that start instantly and are thus basically bound to real-time bandwidth constraints will probably win out over services that require a start-up delay, so over time I expect the on-demand services to look more like the broadcast case than the disc case.

As a result of these pressures, I find myself pretty unexcited about the digital revolution in broadcasting. Marketers talk pretty about how the digital revolution will increase quality, but they don't seem to have gotten the memo from the engineers that due to the decisions by the business unit, digital is actually degrading quality to increase quantity. We're going to have more TV stations and radio stations than ever before, but rather than subjectively sounding like crap because we don't like the music or shows they are playing, they will objectively sound like crap due to overcompression. We truly will have hundreds of channels of garbage.

As a final note, many people are not bothered by the digital degradation that I see. If you don't know what I'm talking about, I'd encourage you not to try to find out too hard, because once you start to see and hear the digital artifacts, you can't stop hearing and seeing them.