Before the advent of affordable digital tape and budget hard disk recording systems, MIDI virtual tracks were a great way to make high-quality recordings. If you had MIDI tone generators, some signal processors, and a mixer, you could compose in a sequencer, send the data to the MIDI gear, and mix down in real time to DAT or analog tape. The name "virtual tracks" came about because the various MIDI sounds acted like tape tracks and were mixed down like tape tracks, but weren't tape tracks -- the audio was derived by playing back data recorded in a sequencer through a tone generator, in real time.

Virtual tracks are still very useful in some situations. But virtual tracking has never been trouble-free. Let's look at how to get more out of virtual tracks by teaming up with digital audio recording. Typical virtual track problems include:

Timing Issues

When a computer has to send significant amounts of data to multiple MIDI instruments over multiple channels and ports, timing "slop" can become audible. In addition, different synths take different amounts of time to respond to MIDI data, so even though they all play back in real time, some notes may play slightly later than others. (The usual fix is to slide a sequencer track forward or backward in time to compensate.) If you're using only one MIDI instrument, it may "choke" during passages when it has to play a lot of voices at once.

Non-Repeatability

Suppose you mix down a tune full of virtual tracks from various synthesizers, then decide later on to make a few tiny tweaks in the mix. Those instruments have to play back through something (normally a mixer), which introduces a variable. In accordance with Murphy's Law, there will usually be some parameters that aren't under MIDI control. These include mixer gain trim (even most digital mixers don't automate mic pre gain), analog mixer EQ, synth volume knobs, etc. As a result, your mix probably won't play back exactly the same way if you've been working on a different project in the interim. And if you haven't been completely faithful about backing up all your synth data, you may have to re-create multitimbral setups from scratch -- or even individual programs if you left them in your synth's memory, then edited them for some other project.

Effects

With audio overdubbing, you can use the same processor on different tracks by patching in the processor during the overdub process. However, because virtual tracks are played back by MIDI instruments in real time, all processing needs to be added in real time. You may not have enough processors to handle all of the tracks, especially if you want to use a high-end processor for some exotic effects.

MIDI effects can also cause problems. It's possible to create some great MIDI echo effects by copying tracks and altering velocities to change levels, but this can eat up polyphony.

The Fix: Bouncing to Audio

To get around these problems when they first surfaced in the '80s, many pro studios with multitrack recorders would bounce completed MIDI tracks to analog tape. This overcame the limitations mentioned above: You could patch in processors during the bounce, computer stability was improved because the computer had to send out only a limited amount of MIDI data (whatever was currently being bounced to the track), non-repeatability didn't matter because the tracks were no longer virtual collections of data but actual audio, and there were several ways to adjust for synth latencies, such as offsetting a sequencer's start time compared to timecode data recorded on tape.

In the early '90s, digital audio remained track-limited and costly. As a result, many programs combined MIDI sequencing with digital audio, on the theory that you could use your precious digital audio tracks for vocals, piano, guitar, and other "non-MIDI" instruments. But times have changed. With fast hard drives and processors, it's commonplace to play back 24 or more tracks of digital audio, and throw in some plug-ins too. As a result, MIDI seems increasingly valuable as a temporary sketchpad medium -- a way to edit a performance before recording the sound produced by a MIDI-driven tone generator into a hard disk recorder.

Typical Candidates for Bouncing

Whether you need to record MIDI tracks as audio into your digital recording media depends on a number of factors: how you work, the gear you use, and your production goals. Maybe . . .

* You've created some drum or bass loops using MIDI-sequenced tone generators, and want to record the looped sound into a digital audio program.

* A composition was started within a hard disk recording program, but now requires some additional MIDI parts or MIDI controller information.

* A composition was started within a keyboard workstation sequencer, but now you want to switch over to a digital audio-based system to add non-sequenceable parts (vocals and the like).

Bouncing Techniques

MIDI-plus-Audio Sequencers.

It's easy to bounce MIDI sounds into digital audio with this type of program, because there are no sync issues. Just work out your MIDI part, send it out to your tone generator, feed the tone generator's audio output into your audio interface's input, and press the record button. If there are any timing latency problems, you can fix them by sliding the newly recorded audio track forward or backward in time to compensate. To improve the timing stability, you'll want to mute all of the other MIDI tracks in the arrangement while recording.

One of the beauties of MIDI is that if you later decide that the track needs more editing than is possible by manipulating the audio, no problem: Go back to your MIDI track, do the tweaks, then re-transfer.

A corollary is that if you do run out of digital audio tracks, you can always premix several MIDI instruments to a smaller number of audio tracks. Unlike the bad old days of analog, if you blow the premix, so what? You can always re-edit at the MIDI sequencer and re-bounce the audio until it's right.

External Sequencers.

If the sequencer is external (a workstation keyboard, a different computer, or a stand-alone unit), synchronization becomes an issue. Generally, I make the computer that's recording the audio the master, and slave the other sequencer to it via MIDI clock. (Why force the sequencer to do complex SMPTE calculations if it's just following a tempo?) The one crucial caveat is that you have to nail any tempo changes before doing the transfer. Once MIDI is converted into digital audio, it's tougher to add slight tempo changes.

Another possible problem is sync stability. In other words, the keyboard sequencer might have good timing when playing back using its internal clock, but jitter somewhat when following another clock. There are some workarounds for this problem:

* Don't worry about sync: Transfer the audio into the computer on the fly, and drag it into place afterwards.

* Transfer each sequenced track individually rather than trying to record, say, eight tracks at once by feeding eight multiple outs from your tone generators into eight input channels on a digital audio recording interface. This also lets you disable all channels not in use, which may improve the timing.

* If all else fails, include a countdown at the beginning of the sequence -- four sharp clicks are ideal. If you're transferring individual tracks or premixed combinations of tracks, make sure you record the countdown at the beginning of each pass. Then, after the tracks are recorded, you can line up the clicks so the tracks at least start at the same time. Unless you have a very unstable system or a very long tune, this kind of "free syncing" will usually work well.

You may also encounter some timing problems due to soundcard or synth latencies. Thankfully, hard disk editing makes it easy to slide tracks forward or backward to compensate.

In some situations using external sequencers, I transfer the MIDI data recorded in a workstation over to the MIDI sequencer portion of the program for editing before turning it into audio. The only real caution here concerns optimizing the timing stability. To do this, transfer one MIDI track at a time, at a very slow tempo (such as 60 bpm) for the best timing accuracy.

If you already have a rough mix at your MIDI tone generator, transferring that mix will force you to make a few decisions. For example, suppose a string pad is mixed in the background. If you record it into your hard disk recorder at a low level, you won't be taking maximum advantage of the analog-to-digital converter's dynamic range, which will reduce the signal-to-noise ratio of that track.

One option is to send over the part at full volume, then re-create the mix in the hard disk recording system. If the mix depends on using controller 7 volume data, you may be in luck: You can strip off the controller information to a separate sequence track, bounce over the audio at full volume, copy the controller information in a MIDI-to-MIDI transfer, then (hopefully) be able to apply that controller info to the hard disk recorder's mix. Yet another option is just not to worry about it, and transfer the part over regardless of level, as it won't be very prominent in the mix anyway.

One final advantage: If you ever need to do a new mix, it's a lot easier to go back and work with audio tracks than it is to set up your MIDI gear, make sure channels are set correctly, load any sys-ex program files that may be needed, and so forth. And as far as I'm concerned, easier is definitely better when it comes to using gear in the studio.

This article presented courtesy of Keyboard Magazine.