Computers derived from the IBM-compatible PC standard are the world's most popular desktop fixtures, and the main reason for this popularity is the relatively low cost of PC hardware and software, as Janet Cook and Paul White have already discussed elsewhere in SOS this month. Another PC plus is the ability to accomodate extra hardware to perform specialised tasks -- in the case of musicians, adding MIDI and/or audio capability is a matter of buying the right kind of soundcard. However, this presents the musical PC user with a problem: picking out the 'right' soundcard from a large number of alternatives. (Another difficulty associated with soundcards is what to do if you are unlucky enough to have hardware compatibility problems, which, as Paul White has already explained, will be dealt with in a forthcoming issue of SOS). The purpose of this article is to simplify the process of choosing a soundcard by explaining the different types available (an overview of the actual soundcards currently on the market is also in the pipeline). But first, as they say...
The idea of having slots for I/O (Input/Output) cards was not new when the IBM PC was designed at the beginning of the '80s: the Apple II had offered expansion slots in 1977. Almost as long as there have been PCs, there have been PC-compatible cards available for the computer musician, the most significant of which were the MPU401 series of MIDI interfaces from Roland, which allowed the PC to act as the heart of a MIDI system. The ability to record and play back audio on a PC took a little longer to develop, and didn't really take off until a small Singapore firm called Creative Labs released a games soundcard called the GameBlaster, which later grew into the SoundBlaster.
However, it was the introduction of Windows 3.1 which really revolutionised PC music-making. This not only integrated the PC's music and sound capabilities into its operating system, but brought down the prices of suitable hardware to within the reach of impoverished musician types. The main technical upshot of Windows 3.1 and the Multimedia PC (MPC) standard was that the external hardware no longer had to be compatible with the ageing Roland MPU401 standard (for MIDI), or the rather limited SoundBlaster audio specification. This resulted in an explosion in the number of competing products attempting to fulfil PC audio needs. A wealth of confusing, conflicting hardware is now available, accompanied, naturally, by a host of puzzled buyers wondering what choice to make.
The functions of a modern PC audio card can be divided into three broad categories; MIDI control, sound synthesis and digital audio replay (more on the last of these next month). A soundcard will offer at least one of these facilities, and possibly all three, along with some 'glue' electronics, such as a simple onboard mixer for combining the various audio signals, or a SMPTE/EBU timecode input for synchronisation.
The simplest and oldest type of PC music card is the dedicated MIDI interface. This type of card is technically not a lot different from a serial interface: it provides a way for the PC to exchange MIDI messages with external keyboards, sound boards, lighting controllers, and various audio-visual devices. This interface-only type of card places the least stress on your PC, as processing a MIDI data stream -- even from a multi-port interface -- is quite tame compared to shifting around digital audio data.
The Roland MPU401 has always been the de facto standard for MIDI cards due to historical reasons. Compatibility with this standard is only really important if you want to run software that directly accesses the MIDI hardware -- for instance DOS-based games or MIDI software. There are a small number of Windows-based applications that need to directly control a MIDI interface, due to performance considerations, but these are pretty few and far between. In general, MIDI cards perform better than serial or parallel port MIDI interfaces, since the PC's processor can more directly access the hardware.
Another important feature of PC soundcards is onboard synthesizers. These come in a variety of flavours, and you have to be careful to ensure that you get what you expect -- both in terms of quality and facilities. All MPC soundcards include a synth, as it is part of the MPC specification, but the quality can vary greatly. There are a variety of available sound-producing technologies, ranging from extremely basic FM sounds to sophisticated sample-based systems. Set out below are the basic technologies, so that you can get a feel for what's on offer and decode jargon-ridden sales blurb.
OPL3 (FM) SYNTHESIS
This is the oldest -- and perhaps still the most common -- form of synth available on PC soundcards. The OPL2 and 3 chipsets were originally developed by Yamaha for use in arcade games machines, but were adopted by early PC games cards developers as an easy way of getting music into a PC. The sounds are extremely basic, and don't compare well with even the cheapest home keyboard available in your local High Street electrical store. If a soundcard doesn't define its method of synthesis, it probably uses these chips, which are still standard in the PC-based games world. There's not a lot you can do with these onboard synths, except make music that sounds like it has been created on a Stylophone.
The latest OPL development is the OPL4 chipset, which combines the earlier FM technology with Yamaha's AWM (Advanced Wave Modulation) techniques to give the advantages of sample replay technology (see below). OPL4 gives backwards compatibility with the earlier FM sounds, and could be a good choice if you want to be able to test your sound files on older soundcards -- say for multimedia soundtrack composition. Sometimes, more advanced cards aimed at the games market -- like the Creative Labs AWE32 and the Turtle Beach TBS2000 -- will incorporate the OPL3 chipset for the same reason.
This technology gives much better results than FM-based OPL synthesis, as it uses samples of real instruments to produce its sounds. The idea is that you store a number of samples in memory on the soundcard itself, and a very simple DSP plays back the data when the PC's main processor demands it. The audio data can be stored in either Read Only Memory (ROM) or Random Access Memory (RAM). ROM-based cards have fixed instrument lists (usually based on the General MIDI sound specification) while RAM-based cards need to have the sounds downloaded to them by the computer before they can be used. Some cards, like the AWE32, have both ROM and RAM wavetables, which gives you the best of both worlds.
The audio quality of these cards depends on a number of factors. The first is the quality of the samples -- or the amount of memory available to them. ROM-based cards tend to have memory capacities ranging from 1Mb to 4Mb of sample data, with the lower end of the range sounding pretty awful, even when compared to quite modest home keyboards. The cards that have more memory produce better sounds, which can be of comparable quality to stand-alone boxes such as the Roland Sound Canvas or Yamaha MU80. Some cards even have effects such as reverb and chorus, which sweeten their sounds considerably.
RAM-based wavetable soundcards use exactly the same technology as the ROM-based cards, but the sample data needs to be downloaded from the PC's hard disk before the card can make any sound. This gives RAM-based cards some significant advantages over their ROM-based siblings. For a start, sample memory on a RAM-based card can be used much more efficiently, since you only need to download the sounds you are going to use. With a ROM-based card, you need to have all the sounds -- for example 128 GM instruments plus 47 drums -- available at all times, even though you are unlikely to use more than 16 instruments at once. For a 4Mb ROM-based soundcard, this works out at an average of just under 23K per instrument, or 250 milliseconds per sample at CD quality (44.1 kHz, 16-bit stereo samples). Doing the same sum for a RAM-based card with 1Mb of memory, you could get almost three times the previous figures, allowing longer or more complex multisamples to be used. RAM-based soundcards suffer the slight disadvantage of a short delay whenever a sample needs to be downloaded, but this can be handled by the Windows operating system.
The other big advantage of a RAM-based wavetable card is that you can edit or replace the basic samples that you use, allowing you to customise the sounds. ROM-based cards have a fixed sound set, and since they are usually GM-compatible, they all sound more or less the same. This is great if you are playing back a commercial MIDI file or a multimedia presentation, but it can get pretty limiting if you are trying to be creative. A RAM-based card will be supplied with a sample/program editor and Bank Manager utility that will allow you to create your own distinctive sounds when used in conjunction with a recording facility on your soundcard. Essentially, you can look at a RAM-based wavetable card as a sampler inside your PC, with all the benefits that implies.
'PURE' SYNTH CARDS
Roland's LAPC1 (pictured left) is an example of this type of card -- it's essentially a complete Roland MT32 synth on a PC expansion card. Unlike the wavetable cards mentioned above, the soundcard synthesizes its sounds from scratch, rather than simply replaying a sample. The advantage of this is that you can be even more creative, not being limited by the sounds you can record, or the finite nature of a sample. Just as with a stand-alone synth, you can create evolving pads, lead sounds and creative effects by programming the sound generator. I know of no new cards currently available that give you this type of synth, but you may pick up a second-hand LAPC1 if you shop around.
While 'pure' synth cards like the LAPC1 are not currently in fashion, synthesizer technology is moving on with the development of new products based on the Virtual Acoustics (VA) algorithms developed at Stanford University. Yamaha have been especially active in this field, producing stand-alone synths like the VL1 and now the new VL70m (reviewed in last month's SOS). VA (sometimes referred to as Wave Guide) is probably the most advanced synthesis technique currently available, employing powerful DSP technology to model real instruments rather than simply replaying 'snapshot' samples of their sounds. This is the most realistic form of synthesis yet devised, but the synths incorporating it are also the hardest to program. There have been persistent rumours about the imminent launch of PC soundcards incorporating this technology: my latest information seems to point to something happening around Christmas 1996, or early next year.
Actually pioneered by Creative Labs when they introduced their first 16-bit SoundBlaster, the concept of a daughter board has been adopted by a number of synth manufacturers -- including Yamaha and Roland. The idea is that you can attach an additional synth to your main soundcard using the connector provided for the purpose, as Paul White has already explained on page 162. This allows you to buy a low-cost soundcard and then upgrade it later, should you find that you need more and/or better-quality sounds. If you want to go along this route, you should check that your soundcard has a WaveBlaster-compatible connector. Yamaha's XG and Roland's SoundCanvas sounds are now available on daughter boards, and Creative Labs offer Emu sounds on their WaveBlaster.
Buying a soundcard that offers you the option of upgrading the synth is a neat way of insuring against the future, whilst making the most of a limited budget, but there are some points to be aware of. One of these, concerning the loss of audio quality when the output of the daughter board is channelled through the main soundcard, has already been raised by Paul White on page 162, but there are other problems too. Most soundcards come fitted with a built-in MIDI interface which is accessed by purchasing a special cable incorporating the appropriate buffering. The software drivers supplied with the card will show this port labelled as External MIDI. As far as the computer is concerned, the synth on the daughter board is connected to the main soundcard's external MIDI port, which means that you can't then use this port for independently driving external MIDI devices.
Another, more mundane consideration is whether there is enough space in the PC to physically accomodate the additional width of the daughter board. I have come across situations where the second card can't be used on a particular PC because there isn't enough space between the expansion cards! While this shouldn't be a issue with most PCs, some of the smaller desktop cases can be problematic.
Next month, Brian will be looking at soundcards that handle digital audio recording to hard disk.
I/O (INPUT/OUTPUT) ADDRESSES
PCs which are based on Intel processor chips (the x86 family and Pentiums) access their internal memory (RAM) and external devices (disk drives, MIDI ports, and so on) in different ways. This shouldn't have caused a problem, but unfortunately, when the PC was originally designed, IBM cut a few corners, and thus the amount of memory allowed for external peripherals is quite limited. Since the more recent ISA buss had to be compatible with the original PC buss when it was developed, this design flaw is still with us to this day. Most PC cards give you a number of choices for their I/O memory address, so this feature is only a mild irritant. Only change the I/O address from the default if you have to.
IRQ (INTERRUPT REQUEST) LINES
The Interrupt lines allow PC hardware to alert the processor chip that a high-priority event has occurred and needs attention. This means that unpredictable events -- like MIDI bytes from a live performance -- can be read by the PC without tying it up completely. These interrupt lines are a far more precious resource than the I/O addresses, because the PC only has 16 of them. Theoretically, the IRQs can be shared between a number of soundcards, but in practice, this almost never happens.
DMA (DIRECT MEMORY ACCESS)
Though mentioned by Janet Cook in her feature on PC anatomy starting on page 164 this month, Direct Memory Access deserves further explanation. When large amounts of data need to be shifted around -- say when the PC is playing a sound file -- a soundcard can take control of the PC's data buss and transfer the data directly to or from memory without getting the main processor involved -- hence direct memory access.
DSP (DIGITAL SIGNAL PROCESSOR)
This is a computer processor specifically designed to handle the type of maths required for processing sampled analogue signals. Originally developed for radar and radio signal processing, DSPs have been applied with great effect to digital audio signals. In fact, you find some sort of DSP in all digital effects units. The addition of a DSP to a soundcard can greatly enhance its performance -- but only if the software you are using supports the DSP!
MIDI VS DIGITAL AUDIO
This is bit of an old chestnut, but is worth explaining one more time. MIDI actually doesn't have much to do with sound. What the MIDI protocol does is describe a performance: what notes you hit, the amount of pitch bend or volume level to be applied at a given moment, and so on. This information can then be used to create music with a MIDI-compatible sound module, but MIDI doesn't directly define the timbre of the produced sound itself. This means that the MIDI data for a given song is very compact compared to the same song stored as digital audio data -- usually by a ratio of more than 100 to one.
THE MIDI BOTTLENECK
The relative compactness of MIDI data over digital audio mentioned above means that MIDI information can be transferred in real time over a comparatively slow data link. MIDI data is transmitted at 31,250 bits per second, or around 1,000 MIDI note messages per second (in a worst-case scenario). To put this in perspective, it would take just under an hour to transmit one minute of CD-quality audio over a MIDI link.
In live and sequenced performances, the speed at which MIDI data is transmitted is less important than the amount of time it takes the computer (or sound module) to recognise and respond to the information. This time period is called the Latency, and depends to a large extent on the power of your PC or the DSP in your sound module. If the latency is constant, it won't affect matters very much, but a variable latency will cause timing inaccuracies in the playback of a recorded MIDI track. Basically, the faster a PC is, the smaller its latency will be.
For more information on PCs and soundcards, check out the following books, all of which are available from the SOS Bookshop. See the Mail Order pages at the back of this month's issue for further details.
The Complete SoundBlaster by Howard Massey (order code B313, £10.95).
Multimedia On The PC by Ian R Sinclair (order code B272, £11.95).
PC Music Handbook by Brian Heywood and Roger Evan (order code B332, £10.95).