USB & MP3

The leading USB developers announce USB 2.0 on the Net.

In the second instalment of his regular column looking at the latest developments in digital technology, Dave Shapton considers the future of the troubled Universal Serial bus, and the direction in which the MP3 debate is heading.

I'm a great fan of the concept of improved, high‑speed digital data transfer protocols, such as USB and its faster relative, Firewire (or, to give it its prosaic official title, IEEE 1394). And they're no longer just concepts, either — at last products have been released that support both standards, offering genuinely straightforward installation with dramatic improvements in performance over the serial and parallel ports they are intended to replace. What I like most is not having to open up my computer to add devices to it. In fact, USB and Firewire make it practical to have a PC that has no expansion slots at all, and will work better for it! We are already seeing digital audio and MIDI USB devices, and I hope that these will be the first of many.

I don't want to detract in any way from the manufacturers' efforts to bring us new USB products, therefore, nor from our enthusiasm to buy them — but a word of caution is appropriate.

I've been hearing from several sources (including product reviews in SOS, such as Martin Walker's review of the Opcode DATport in SOS July '99) that USB is not the plug‑and‑play panacea it is meant to be — there are problems with corrupted data, sluggish transfers, and so on. However, I don't think the problems necessarily lie with USB itself, or with the products that support it: it's more likely that the software drivers are at fault, possibly even at the operating system level. After all, Windows 98 is the first version of the Microsoft OS that supports USB, and the iMac is the first mass‑market computer from the Apple stable to support it. No‑one these days expects the first release of a radically new piece of software to be anything other than a late beta release (software that is for sale is not 'bug‑free': in reality, it's a version that is considered to have few enough bugs for it to be acceptable to paying users, which is quite different from being 'bug‑free').

I have every sympathy for the software developers who have to make these things work. And it's even harder with multi‑device interfaces like USB. When you consider what it's supposed to do, it's not surprising that it's difficult to make it reliable. USB lets you hot‑connect devices without rebooting, supports dozens of simultaneous connections, and has a hub‑based architecture to avoid unwieldy daisy‑chaining of cables. You only have to try installing a PCI card in a computer to know how many times you have to re‑boot, install software and re‑boot again. The promise of USB was that it would avoid all this, but the current reality is different.

For the time being, I'd say it's best not to hot‑plug USB devices, and never connect anything to USB that doesn't first have its own driver installed. Ignore this and you'll find it could be difficult to install the correct drivers when you do get round to it.

In addition to problems over auto‑configuration, I've also heard that some USB audio devices are not working as they should. Although USB's available bandwidth is easily adequate for uncompressed real‑time digital audio (CD‑quality audio needs 1.4 Mbits/sec, and USB offers a bandwidth of 12 Mbits/sec), any interruptions or delays would quickly become apparent. The DATport S/PDIF interface review in SOS mentioned that the product was prone to clicks, and that the manufacturer was working to correct the problem. Maybe the driver was not correctly handling the multi‑device aspect of the buss. Who knows — but the moral of this is that, at least for the time being, you should test any USB device thoroughly, and make sure that your supplier will take it back if neither you nor they can make it work on your computer. I've no doubt that software support for USB will improve rapidly, and that the forthcoming USB 2 specification (see below) will help as well.

Incidentally, I've found that the best way to test digital audio interfaces, in the absence of specialised test equipment, is to record a 1KHz tone, and listen for glitches (and not just for a few seconds, either — basically, the longer you can spare, the better). Single‑sample clicks can sometimes pass unheard because they represent — at an instant — a very high frequency, which your ears might not pick up, whereas we seem to be more sensitive to interruptions to continuous tones.

USB 2

The RIAA's attempt to sue Diamond over the Rio MP3 player, as reported on CMPnet. The lawsuit was finally dropped in August.

So what of USB 2? Today's USB peripherals have a bandwidth of 12 Mbit/sec. That's useful enough for undemanding or 'bursty' data transfer, but the trouble is that if you add devices, they have to share the data bandwidth between them, and you quickly get a log jam. At a developers' conference this month, attended by Intel and its partners, the revised specification for USB 2 was announced (you can read more about it on the USB developers' site at www.usb.org — see screenshot, page 28). If you were expecting USB 2 to be more of the same but a bit faster, think again: it's going to have a maximum data rate of 480Mbit/second. That's 40 times faster than USB 1

So USB 2 has the same kind of data rate as SCSI and Firewire, which means that it could be used both for real‑time I/O devices and hard disks. It remains to be seen how suitable it is for either task, because, ultimately, performance depends on how well it can arbitrate between conflicting requests from devices it is hosting, and what sort of load this puts on the CPU (the main processor in the PC).

Now, maybe I'm missing something here, but I don't see why we need USB 2 when we have Firewire. The data rates are similar, but Firewire is already established, is likely to be the standard for connecting domestic entertainment devices, and has special data modes that guarantee bandwidth for real‑time media transfer. It would be ridiculous to have Firewire and USB 2 on the same motherboard, and it would be ludicrous to have to convert a Firewire datastream to USB 2. I'll give you more on this as details emerge.

Off The Mpeg

Handspring's Visor palm‑top. An MP3 playback module is planned as an optional accessory.

A lot has also been written about MP3 and, whichever way you look at it, MP3 is already a major media phenomenon. The question I find intriguing is how important it will prove in the long term. One thing is certain: it has got the attention of the record companies and copyright organisations.

In some ways, it reminds me of the rather paranoid frenzy that accompanied the arrival of DAT, over 10 years ago. The issue with DAT hinged on its ability to produce a clone of a CD. Everyone, it was claimed by DAT's opponents, would clone their friends' CD collections onto DAT and would never buy a CD from a shop again (an argument as plausible as saying 'nobody shops at IKEA because of the queues'). As it happened, DAT bombed as a consumer medium: but the point was that, for most people, cassette tape was simply good enough. Very few cassette decks are found in quality‑critical locations; most of them are situated in radio‑cassette recorders, cars, or cheap‑and‑nasty lo‑fi systems. So concerns about the ability of DAT, or any other digital medium, to clone copyright material, rather than make a good but imperfect copy of it, are simply irrelevant. MP3 files are certainly not clones of the original: the degree of compression needed to send digital audio files over an Internet dial‑up connection is so severe that, to my ears at least, significant quality is lost.

But the question that is worrying the copyright organisations isn't about quality. It's about availability. Suddenly, a significant proportion of the world's recorded music repertoire is available for nothing from the Internet. But is this really anything new? Most of the world's recorded music always been available for nothing from your friends, and from the radio. How is the Internet different to this? It isn't. You listen to a track, like it, and either make an illegal copy or you go out and buy it.

The knee‑jerk reaction of some music industry organisations (especially in America) has been to call for a ban on devices that can upload and play MP3 files (witness the attempt by the Recording Industry Association Of America to sue Diamond Multimedia over production of their Rio hardware MP3 player — a lawsuit which, interestingly, has now been dropped). Now, some Windows CE devices and at least one 'palm‑top' computer just announced can also play MP3 files (see www.handspring.com). Should the copyright organisations ban them? Should they ban computers connected on a network (including the Internet)? Why not ban white vans because they tend to be used in bank robberies?

The closer you look at this issue, the more it evaporates into nonsense. You can't make laws to stop people knowing how to do things. The MP3 algorithm is in the public domain and always will be. To me, the arguments about MP3 will stagnate until there is widespread take‑up of online music delivery. It will get interesting again when broad‑band access comes along, because then you'll be able to copy your friend's entire CD collection faster than in real time. Oh, and the connection will be permanent, not dial‑up. At this point, as a copyright owner, I'd get worried.