And it cannot have escaped anybody’s attention that A-to-D and D-to-A converters are better left to dedicated, external sub-systems. Again, for reasons of economy, converters fitted to most off-the-shelf PCs have a parts cost in single digits of dollars; committing your master elements to these types of device makes no sense at all. Better, so current wisdom dictates, to spend a few hundred (or thousand) bucks on conversion systems from manufacturers that really know how to get analog signals into and out of the digital domain, and let the host PC manage the resultant digitized bit streams.

Which segues nicely into how we can perform such computational tasks on multiple data tracks. Within a simple editor, we will need to generate timecode-based pointers to data files stored on fixed and removable storage media, and then move elements relative to one another against a common timeline. Crossfades might be performed in real time, while some older systems would re-record them as separate sound files. (Think about it: performing a real-time crossfade requires that the PC access both streams simultaneously and then perform the math to produce a blended track.) Given the data bandwidth of early SCSI-capable hard drives, we might be pushing the envelope to secure adequate data throughput to provide eight or (at most) 16 audio replay tracks.

Until recently, we had limited number of options for designing a DAW capable of offering mixing, equalization and dynamics processing on multiple replay channels. Individual NuBus, PCI and similar cards can only go so far in offering individual processing of signals while they are being played into the system. But if we are looking for expandable track capacity with no inherent limitation on what we can call into action at any point during the tracking and mixing session, an alterative topology is necessary.

Several DAW manufacturers use a variant of Time Division Multiplexing (TDM), utilizing a DSP-intensive plug-in board within the PC host. Now, a tightly regulated high-speed bus is used to transfer signals from the main processor to the computational-aggressive DSP engine, and then extract the results for output to the relevant audio ports – either individual direct outs for off-board mixing, or as blended stereo/multichannel mix balances. Now, assuming that we can run the TDM bus and DSP engine fast enough and with sufficient bandwidth, we can develop a pretty powerful digital audio workstation.

Limitations occur, however, when we need to add more I/O channels to a TDM-based system, or look to run the DAW at enhanced sampling rates. At least one well-known manufacturer is facing a major re-design challenge in an attempt to address just such a dilemma, in addition to other operating anomalies such as timing errors between non-adjacent tracks traveling along the bus. (Despite highly accurate clocks and scrupulous software implementation, it is easy to end up with offsets of several milliseconds; not at all useful if track #1 and track #28, for example, are carrying stereo drum overheads, or we are looking for critical timings between close mikes elements of a rhythm section.)

Intel Pentium IV processor

And many of these previous-generation TDM-based workstations were based on DSP chip sets that are no longer state-of-the art; replacing them with current versions is far from trivial, however, and would require major costs and software/firmware upgrades.

The future, as we are now beginning to recognize, lies in native or host-based processing, in which we can run a Pentium IV or multiple PowerPC G3/4s at sufficiently high clock rates with wide, expandable data highways that can accommodate several hundred on-line tracks of 24/96 processing without the need for add-on boards (aside from the familiar multichannel I/O subassemblies). Coupled to current generation, high-rpm Ultra-100 SCSI and EIDE Ultra ATA/100 drives, we can dramatically reduce the complexity – and cost – of a powerful DAW capable of offering all surround-sound mixing, editing and processing functions from a single Windows- or Mac-compatible platform, complete with MIDI sync and on-line video playback. We are even seeing portable PCs capable of offering a lot of DAW power in a smaller package.

Looking to the future, multiple-processor parallel and vector computers will soon become available from a number of vendors; code written in transportable languages will then be able to take full advantage of the Gigaflops of computational potential offered by these emergent designs. Keep watching the skies!

*To the uninitiated, “cubes” refers to cubic inches of engine displacement.

Next Time: “Geography/Community: Who are you working with?” Then: “File Formats & Media.”

©2001 Mel Lambert/Media&Marketing. All rights reserved