Digital Audio at Last…

Thinking Digitally by Steve Macatee

In this article, Steve Macatee compares analog audio system to a digital audio system.

There has been very little thinking digitally about audio in my past articles. It’s time to get back to audio.

Analog audio waveforms are continuous in both time and amplitude, thus requiring the study of continuous math functions to implement audio widgets in electronics or systems. Digital audio uses discrete mathematics which encompass, among other things, discontinuous sets of numbers. Lots of numbers.

Since the computer industry is well suited to support myriads of widely varying electronic parts (e.g., chips) and software algorithm tools that handle lots of numbers like humans handle air – very easily – the audio industry is almost wholly dependent on such large industries for the individual pieces needed to make audio gadgets. It’s the mass market demand from other huge industries that often drives the parts and tool costs for manufacturers and audio gadget buyers alike.

Audio integrated circuit (IC) manufacturers offer parts that input continuous analog audio signals and encode them into discrete digital audio number sets. Once the audio is in numeric form, almost anything you’d want to do to your audio can be accomplished using the mathematical processing offered by ICs and/or their related software algorithms: process, store, transform, distribute, transport, duplicate, decode (for playback), data compress…

And compared to implementing any one of the above processes with an analog system, a digital one these days offers significant advantages: less expense in parts, lower power consumption, less performance variation across lots, increased reliability and immunity from noise, better aging & drift performance across temperatures, among others. The drawbacks are stressed less in ads and marketing propaganda: parts have shorter life cycles due to other industries’ obsolescence rates, development complexity is intense, our industry is typically slow to change yet computer industry trend rates are ever increasing, poor product usability and software-based performance dependencies can ruin any good day, product or application. And I haven’t even mentioned the long-standing retail or commoditization concept that most electronics these days are designed to never be serviced. Combine this with fast parts obsolescence rates and life in digital audio land can be scary. But since thinking digitally requires precise rounding to discrete binary values, this means all half full glasses are really completely full. But even better, even half empty glasses round up to totally full!

To encode an analog audio voltage into a digital audio number stream, analog-to-digital converter ICs sample the waveform both in time – horizontal slices or samples – and in vertical voltage slices or quantization – usually expressed as some number of bits. Digitizing is somewhat like placing a very finely spaced piece of graph paper over the audio waveform and plotting the precise cross-points where the waveform intersects the cross sections of the graph paper’s lines. The sample rate dictates how close together the paper’s vertical lines are. The quantization dictates how close together the horizontal lines are.

Professional audio gear typically uses a 48 kHz sample rate, but other common rates are 32 kHz (broadcast), 44.1 kHz (audio CDs), and 96 or 192 kHz for high-end recording or audiophile applications. Quantization, the number of bits used, in pro gear is typically 24 bits, with the 16 bits of audio CDs formerly being commonplace.

In future articles we’ll expand on these basics. sm

Digital Audio - Graph showing bit depth and sample rates