Market Overview - Audio Cards: Beep and Booster

Audio cards have turned the PC from a pathetic squeaking mouse into a full-throated musical monster. Chris Long explains the score

The PC, IBM?s wayward child, obviously suffered a Victorian upbringing: for it, to be seen and not heard was paramount. Although its parents must have balked at the idea of ?not a peep? simply because that was the one noise it was allowed to make.

This was very much unlike its smooth-talking, charming and highly irritating competitor, the Mac. The Mac would ? and still does ? witter on at the drop of a hat, creating more noise than a crowd of very noisy things indeed. This was the perfect sop for Mac users who inevitably use their computers as very expensive playback devices for strange noises and confusing clips from films, while watching their machine?s dwindling market share.

Of course, the PC?s respectful silence was perfectly acceptable for the first years of its life, people preferred it that way ? being seen and not heard was a good thing.

Then things started to change, and as the market grew up it decided ? much to the amusement of Mac users ? that limited conversational ability from a computer was a good thing. And it was around this point that the extent of the PC?s lack of verbal ability was discovered: it wasn?t quiet by keeping a respectful silence, it was quiet because it was mute.

This meant that the PC had to be modified to make any noise beyond the plaintive beep that was its trademark. And this meant the sound card. When IBM designed the PC, it did so in a hurry. The motherboard didn?t have that much built on to it, and it relied on add-on cards to give it the functionality that the motherboard lacked. Thus the bus which the add-on cards plugged into became the bottleneck for the data splashing around inside ? but at least it meant that if you wanted audio you could have it.

Adding a sound card was fraught with danger, or at least fraught with opportunities to make you pull your hair out. Invariably, the process entailed wrestling with interrupts, I/O addresses, DMA channels, and driver conflicts. Then there were the sound cards clashing with other cards or, as was most common, the application not working with the sound card.

This was because most of the applications using a sound card were games, which were a nightmare in their own right as games programmers took all sorts of liberties with their programming code. Simply running some of this software could hang a perfectly good PC, never mind getting it to work with a sound card.

Enter the Creative Labs Sound Blaster, not the first sound card available but easily the most successful. The Sound Blaster, simply because it worked with the software, helped make games on the PC usable. Launched in November 1989, the Sound Blaster was an 8bit sound card with an 11-voice FM synthesiser and Midi port. And by late 1990, due to the burgeoning multimedia market, it was the best selling add-in card for the PC ? and there were a lot of cards.

Nowadays it isn?t just games that are demanding more sophisticated sound handling, the internet is pushing growth along with business applications like voice to text, voice activation and ? the big one ? video conferencing (which isn?t just pictures, unless you are particularly good at lip reading).

Getting sound on to the PC in the first place takes some fiddling. As the sound comes into the computer it is sliced up many times a second and each one of these slices is digitised and stored on the computer. Because the bits between the slices of sound are thrown away, it is logical that the more slices, the better the digital recorded sound represents the original sound.

There are various levels of quality that can be applied to the recording ? in effect how close together the slices are. For example, to get CD quality audio you slice the sound up 44,100 times a second, and this is the sampling rate. Top-end broadcast quality is 48KHz and the figure can go right down to 8KHz where there is much less detail recorded. Most systems record stereo at 44.1KHz, which is CD quality.

The first generation of sound cards were 8bit, so the information in the audio sample was squeezed into eight bits. A couple of years ago that changed and it is now 16 bits. Newer cards support 20 and 24 bits, working on the principle that it is better to over sample than to under sample.

Interestingly, it won?t be until the death of the ISA bus that everyone will have ?more than 16bit quality? on their PC. The bus width only being 16 bits means that the audio signal coming into the card may be turned into 20 or more bits but that has to be reduced to 16 to be manipulated by the PC?s bus. Despite this, due to some judicious juggling, a lot of the extra information in the 20bit words is retained. Obviously PCI bus machines won?t have this problem and to speed the conversion most cards coming out these days are PCI.

The actual conversion from analogue to digital is done by an analogue to digital converter (ADC). It will come as no surprise to learn that the reverse system is a digital to analogue converter (DAC) and, just to make life interesting, these can also be called the codec (converter/deconverter).

The quality of the codec is also paramount in the quality of the recorded sound. Once converted, the digital signal is sent to a DSP (digital signal processor) that does all the wild and wacky effects and then the sound is either played back or written to the hard disk as a WAV file.

Some cards have more effects chips or split the use of the DSP and add other effects chips, but these are the two major parts of the sound card.

Despite the improvements in the sampling systems, the big problem with audio cards is noise. Not the noise of the squeaky fan or hard disk, but radio frequency (RF) noise generated by all the different bits inside the case. Combine this with the fact that over the years it has been thought perfectly acceptable to build an amplifier (to drive speakers) and a CD-Rom interface (IDE, proprietary or Scsi) on the card, making even more noise.

This then affects the signal to noise ratio ? that is, how much hiss you get along with how much Led Zeppelin?s Stairway To Heaven. The ideal is no hiss at all; this isn?t possible on a #25 card that has all but the kitchen sink built into it ? in this area kitchen sinks are hissy. Lately, in an attempt to cut down the extraneous noise, there has been a move to drop the CD-Rom interface and limit the other bits and bobs.

A high-end professional audio card should sport a signal noise ratio of better than 97db; midrange cards start at about 90db. To aid clean recording, cards are now sporting better connectors like the RCA phono connector in preference to the standard 3.5mm plug ? some are even producing gold connectors (the gold inhibits oxidisation of the metal, which in turn causes interference).

Some mid-range cards have SP/DIF (Sony Philips digital interface) plugs which allow digital dumping direct to digital devices like DAT recorders. Top-of-the-range cards have break-out boxes where the input and output connectors are outboard of the computer. This also allows the use of professional XLR connectors improving the quality of the input sound even further.

These days, most mid-range boards support full duplex audio, allowing you to play back sounds and record at the same time. But as technology fast forwards, this feature becomes even more important as it also allows users at both ends of a connection to speak simultaneously ? an important feature for video conferencing and for telephony applications. It?s a sign of the times, and a reminder that sound cards aren?t just for games any more.

This throws up the problem of the size of audio files ? like video they can get very large indeed. For example, a 4.75 minute stereo track recorded at 44.1K takes up over 51Mb. Hard disks have therefore become an important adjunct to an audio system. So much so that storage companies like Iomega are getting in on the act. Its 1Gb Jaz removable disk drive is fast enough to handle audio files and it has announced software that manipulates audio files and allows recording and play back.

Iomega is also highlighting the movement in this market by launching the Buz Multimedia Producer, an audio video capture system. This part of the market is growing big time, with professional hardware manufacturers bringing low-cost, easy-to-use products to the enthusiast PC user.

Many lower end audio cards have Midi built into them; semi pro and pro audio cards tend not to. The big advantage of Midi is that it is a compact media type, it deals with the data behind the note rather than the note itself. So if you wanted to play, say, Stairway to Heaven on a Midi system, it doesn?t play a CD-quality digitised version of the guitar introduction, it sends data which tells the card to play notes that are built into it.

For this reason, Midi files are far smaller than audio files because they store playback instructions (the pitch of the note and its duration) rather than digitised audio. The way Midi works is fairly simple too: the card will have samples of sounds built in, generally standard instruments such as a grand piano, church organ and classical guitar.

The sounds can be achieved in one of two ways: FM synthesis or wavetable synthesis. In FM synthesis, the chips on the card emulate the sound of the instrument ? this can sound anything from cheap and nasty to even cheaper and nastier.

In wavetable synthesis a real pre-recorded sample of the sound is placed in memory and used. A lot of cards have this built in these days, but the system is memory hungry, so much so that top-end cards like the Creative Labs? AWE64 Gold can take an extra 28Mb on top of its standard 4Mb.

By using a program called a sequencer, you can record and play back multiple tracks, building up complete arrangements one instrument at a time. These sequences are then saved as standard Midi files, and the beauty of the Midi format is that it will play the same tune on any machine equipped with a Midi system. The only difference in the sound will be down to the different instrument sounds on the Midi card.

All of these hardware and software options adhere to a relatively recent variant of the Midi standard called General Midi (GM), which sets out a library of 128 instruments (plus a slightly lesser known requirement to play at least 24 notes simultaneously). Thus, a Midi file prepared for a GM card will play back the correct sounds. The downside is that the sound quality varies dramatically ? high-end cards sound stunning, but older or low-end cards can sound downright nasty.

Of course, the internet has a role in this too. Both Microsoft Internet Explorer and Netscape Navigator support Midi playback ? Navigator through its Live Audio technology, Internet Explorer via Microsoft?s Active X.

Just like the non-internet world, the browsers use whatever Midi system is in the computer at the time, with the subsequent variations in quality.

High-end boards now have much more powerful DSPs, which are used to provide real-time effects such as reverb and pitch-shifting. Other cards from the likes of Creative Labs, Mediatrix Peripherals and Turtle Beach can be configured with onboard sample Ram, which lets you use your own WAV files as synthesiser instruments. DSPs also help the multimedia experience like 3D sound, Dolby Surround AC-3, and digital simultaneous voice and data (DSVD).

Also referred to as positional audio, 3D sound is supposed to simulate natural everyday 360 degree sound and thus fool the ear into thinking sounds are coming from beyond the points where the speakers are located. Microsoft?s Direct Sound and Direct X use 3D sound to correspond to the movements of objects in the 3D visual space. This will create a whole new generation of applications, ranging from virtual training to 3D simulations. Of course, the term 3D is used and abused by many companies and the results can be variable. It?s another case of which standard will out.

As the millennium gets nearer, the PC continues on its equally inexorable evolution into the centre of our world. The current trend in PC audio technology will not be bucked and despite the movement we have seen in the past 15 or so years, there will be more. With this in mind, a bunch of computer companies have got together to define the future of audio on the PC and produced the audio specification called Audio Codec 97 (AC97).

It is the brainchild of an unlikely alliance of Intel, Analogue Devices, Creative Labs, National Semiconductor and Yamaha, and is a specification for a programmable DSP-based audio system. One of the committee?s first actions was to increase the current audio sample rate from 44.1KHz to 48KHz and make 20bit resolution standard. Another point addressed is noise; the card is split in to two ? the input (ADC/DAC) and output (DSP) stages are separated to stop interference. The first samples of the AC97 card are already here, but the rush should occur in the fourth quarter of 1997.

As ever, the march of technology is happening right in front of our ears; the only problem we have now is to improve the noises that appear on our computers ? that may be even more difficult than working out the technology.