Software Synthesizers (Excerpt): Top Music Technology Pages Return to MusicWords Home

[The material below is excerpted from Software Synthesizers, published in 2003 by Backbeat Books.]

The first necessity, if you want to use a software synthesizer, is a decently fast computer -- the faster the better. The synth developer's website will list the minimum and/or recommended system requirements, but I wouldn't suggest settling for a minimum system. If you do, you may be limited to as little as three notes of polyphony and no effects -- pretty much a musical disaster if you're planning to do any serious work. Get the fastest machine you can afford. On the PC side, nothing under 800MHz is even worth talking about. If you're a professional or aspiring professional, 1.5GHz is probably a reasonable minimum. On a Macintosh, you'll get equivalent power with numbers that are somewhat lower.

Don't skimp on RAM, either. Even if the website says you can get away with 64MB of RAM, get at least 256MB. (Note: These figures are reasonable in the autumn of 2002 -- if you're reading this book in 2007, you'll need to do your own research to learn what's currently in the ballpark.)

What computer operating system should you go for? The Mac/Windows choice is still largely a tossup. It seems to me Windows offers some small advantages in terms of the amount of software available, but possibly I'm prejudiced. There's some great music software that's Mac-only.

The question of which flavor of Windows or Mac OS to choose is slightly more awkward. Most Windows developers are supporting XP these days. Support for Mac OS X has been slower to arrive, but seems to be picking up momentum as I write this (in October 2002). If I were buying a Mac today, I'd insist on a machine equipped with both OS 9 and OS X. Six months from now, that precaution may no longer be necessary. In every case, though, the time to check with the software manufacturer and find out what OS variations they support is before you buy the computer. This caveat applies just as strongly to OS upgrades. If you upgrade from Windows 98 to Windows XP, for instance, you might easily find that one of your favorite music applications no longer runs -- or that the application runs, but the manufacturer of your audio interface is not supporting the interface with a driver for that OS. List all of the components in your computer music system and check with all of the manufacturers before upgrading your OS.

Note, also, that in some cases the OS requirements listed in the reviews in this book were current at the time when the review was originally written. I've updated the information as well as I could, but some bad data may have flown in under the radar, so check the review date, then check with the manufacturer. If you want to run the synth under a newer OS, you may be able to, even though it isn't listed.

For reasons mentioned above, I recommend using plug-in softsynths with a sequencer/multitrack recorder as the host application. (If you're using one of the virtual rack systems discussed in Chapter 4, this advice is less applicable.) Some multitrack recorders, such as Digidesign Pro Tools, Syntrillium Cool Edit Pro, and Magix Samplitude, provide MIDI support (which you'll need in order to use a softsynth) only grudgingly, if at all. A better bet would be a full-featured sequencer/recorder, such as Steinberg Cubase (on either Mac or PC), Mark of the Unicorn Digital Performer (on the Mac), or Cakewalk Sonar (on the PC). You don't necessarily need to purchase the high-end version: An entry-level version, such as Cubasis VST, will be more affordable, and will most likely support softsynths. Before you buy any host application, though, check to make sure it's compatible with the software instrument(s) you're planning to use.

Next, you'll need some type of MIDI keyboard. For a desktop music setup, a compact unit like the Midiman Oxygen 8 would be hard to beat, but any MIDI keyboard at all will work. You may be able to pick up a used synth for a few bucks that's still perfectly good as a MIDI master keyboard, even though its sound-producing side is lame by today's standards.

If you're just getting started, you may be tempted to make do with your computer's off-the-shelf audio output (a consumer soundcard on the PC side, or the built-in minijack on a Mac), but there are reasons why you probably won't be satisfied for very long. Here's why:

Computers typically spend some time doing lots of different "housekeeping" tasks. Though these are invisible to you when you're writing an email or whatever, they're going on in the background every second. But as we noted earlier, in order for audio to sound good, it has to be output smoothly, with no interruptions. The way software developers get around this is with output buffering. Essentially, the softsynth sends a chunk of audio data to the soundcard a little ahead of time. The soundcard then pulls this data out of the buffer as needed and sends it to the DAC. Before the buffer is empty (one hopes), the CPU will get around to sending another chunk of data to fill it up again.

If this doesn't make sense, think of the audio output as a bucket with a little hole in the bottom. The bucket leaks water at a constant rate. Every once in a while, a guy comes along with a pitcher and pours some water into the bucket. Then the guy wanders away to do something else. If he comes back with the pitcher before the bucket is empty, all is well -- but if he doesn't show up, the bucket empties out, and water stops flowing out of the hole.

If the guy only shows up once every four or five minutes, you need a pretty big bucket to make sure the water keeps flowing without any interruptions. But if he shows up every four or five seconds, you can use a much smaller bucket.

In first-generation computers, the bucket had to be pretty big, because the guy couldn't walk very fast. Translation: The soundcard needed a big audio buffer in order to prevent dropouts, because the OS couldn't be relied on to fill the buffer in a timely manner.

When you play a note on the MIDI keyboard, the softsynth starts generating a note almost immediately -- but the sound doesn't go directly to the audio output. It goes into the output audio buffer. If the buffer contains a full second of audio (176,400 bytes of 16-bit stereo at 44,100 samples per second), you won't hear the note you've just played until a full second has passed.

The time lag between when you play a note and when you hear it is called latency.

Latency is less of an issue today in software synthesis than it was a couple of years ago, but it's not an issue you can afford to ignore. With an acoustic instrument -- a piano, guitar, or flute -- there's essentially no latency. You play a note, you hear it. Musicians rely on that instantaneous aural feedback in order to shape their performances in an expressive manner. Playing an instrument in which there is significant latency is possible, but it's very difficult and unsatisfying. (An organist playing a pipe organ in a large cathedral may have to deal with more than 1/4 second of latency. But the rich echoes will blur the sound, masking any unevenness in the performance.)

The built-in soundcard in your computer probably has significant latency. In order to use a softsynth in performance, or even to record with it and not get intensely frustrated by the sluggish response, you need a low-latency soundcard.

Various manufacturers have developed ways of getting low-latency performance out of soundcards. The most widely used protocol is called ASIO. If both your soundcard and your softsynth (or its host application) are ASIO-compatible, you're good to go. In Windows, Microsoft's WDM protocol also provides very low latency. And beginning in the newer releases of OS X, the Macintosh offers low-latency audio as part of the system, eliminating the need for an ASIO audio interface.

Another way around the latency logjam is to use a softsynth that runs on its own dedicated PCI board, which is tucked into a slot inside the computer. The synths available for the Creamware Pulsar card, for instance, are inherently low latency, because the Pulsar hardware is not part of the computer's operating system, and is not subject to OS-type interruptions. Hardware-dependent softsynths are not covered in this book, however.

If you hear dropouts when using your softsynth, check to see whether the soundcard's buffer setting can be increased. Many cards let you adjust the buffer size and other settings. Conversely, if you're hearing an unacceptable amount of delay (latency), try decreasing the buffer size. When you reach a point where the audio starts to break up, increase it again just slightly, and the latency characteristics of your system should be optimized.

Finally, let's not ignore the characteristics of the amp and speakers through which you're listening to your computer. Consumer multimedia speakers (or even headphones) will do in a pinch -- but you can make a violin out of a cigar box too, in a pinch. If you want to hear the full-bodied lows and crisp highs your synthesizer is pumping out, set aside enough in your budget for a decent pair of monitor speakers. Studio-quality nearfield monitors are available from Event, Fostex, Mackie, Roland, Yamaha, and many other companies. Having said that, "studio quality" is a marketing term (that is, it's all but meaningless). Quite often, it's the cheap gear that's advertised as "studio quality." Speakers that are advertised as "studio reference monitors" have at least some pretensions of being decent. If someone tried to advertise a cheap speaker as a studio reference monitor, their friends would throw scraps of food at them and put deceased rodents in their mailbox. Or at least we can hope.

If at all possible, avoid using your home stereo for monitoring. Home stereos often enhance the highs and lows in order to make music sound better. For music production, you want a monitor system whose frequency response is "flat" (that is, the output is as close to the input as possible, with no artificial enhancements). "Reference monitors," assuming they come from a reputable company, will usually have a more or less flat frequency response.