Sound savings: Portable audio recorder takes on tape, part 1

Since Sony introduced DAT (digital-audio tape) in 1987, it has been the preferred format option of many journalists, musicians, and live-concert and lecture-audience members, who want to capture a sonic snapshot of the events they attend. Record-label-piracy concerns and consequent copy-protection and royalty complications, though, have curtailed DAT's acceptance in the consumer-electronics world. With recording studios moving to hard-drive-based setups, DAT's popularity is even further on the wane, and recent reports indicating a near-future end to tape-drive-mechanism shipments will likely put the final nail in the DAT coffin (Reference 1).

In retrospect, DAT's demise isn't devastating, because it was not an ideal audio format for a number of reasons. For example, if you accidentally damage a DAT, you disrupt the audio bit stream and have big and, likely, unrecoverable problems on your hands. Compare this situation with mutilating an analog-audio tape: A piece of adhesive tape and a pair of scissors gets you up and running again with only a brief sonic disruption to mark the impairment. Along similar lines, a slight head maladjustment, which would result only in attenuation of high-frequency details in the analog-audio-tape generation, would leave you listening to the "sounds of silence" with DAT.

Because DAT supported sample rates only as high as 48 kHz and per-channel samples as large as 16 bits, it was incompatible with emerging high-resolution-audio trends (references 2 through 4). And, to move the DAT-captured data to a computer, you had three choices: You could connect the digital output of the recorder to the S/PDIF (Sony/Philips-digital-interface) input of your PC's sound card and transfer the audio at a slow-as-molasses 1× rate. Alternatively, you could connect the analog output of the recorder to the PC-sound-card-line input, and suffer the additional misfortunes of quality-degrading interim digital-to-analog and analog-to-digital conversions. Or, you could install a DDS drive in your PC, along with software, such as DAT2WAV or Vdat (Reference 5).

Len Moskowitz, the owner of Core Sound, which has long catered to the audio-recording community, was well-aware of these DAT limitations when he set out to design a high-resolution-audio-capable DAT replacement (see sidebar "Other options"). He noticed that many former tapers were migrating to notebook PCs as their next-generation recording platforms. He also observed the exponentially increasing processing and storage capabilities of the latest handheld computers. Like many of us, he was also acutely aware of the talent and enthusiasm of open-source-operating-system and application developers (Reference 6).

Moskowitz, an electrical engineering alumnus with 20 years' experience at AlliedSignal (now Honeywell) and ITT Avionics, decided to tap into that worldwide talent to help develop his system, which at first glance had a number of seemingly contradictory requirements: It had to be low-cost at low shipping volumes. It also needed to have low power consumption for long recording times, even when diminutive PDA batteries powered it. Its interface had to be compatible with both PCs and handheld computers. It also had to support large samples and high sample rates if the target system hardware and software could process them, and it had to inject little noise to preserve the fidelity of captured audio. The resulting PDAudio system, which is now entering production, employs novel approaches to address these requirements.

The heart of the PDAudio architecture, the $199 PDAudio-CF card, is the result of many hours of brainstorming Moskowitz spent on the proper form factor, the amount of hardware integration, and the appropriate hardware-versus-software partitioning for his target customers (Figure 1). Digigram's VxPocket PCMCIA card was one system whose approach he evaluated before coming up with his own. To maximize the SNR of the analog subsystem, he decided he'd need to move it outside the noisy PC environment. An all-in-one unit that combined microphone phantom power and preamp circuits, analog-to-digital conversion, and digital-processing functions would need to connect to the computer over USB or IEEE 1394 (FireWire).

However, USB Version 1.1 had unreliable latency times and inadequate bandwidth, and USB Version 2 and IEEE 1394 had insufficient penetration within the installed base of computers at which Moskowitz was aiming PDAudio. Further, none of these bus options were appropriate for PDAs. Examining the VxPocket, he also noted that the device's analog-audio-output circuits, although appropriate when Digigram designed the unit several years ago, were redundant in today's handheld and notebook computers, which have those functions built in. The VxPocket's digital output and time-code input weren't critical needs of the audio-recording community, either.

By developing a modular system that split the analog and digital portions of the recording chain, Moskowitz also enabled potential customers who owned microphone preamps and phantom power modules to upgrade their recording setups piecemeal. In its final definition, PDAudio-CF is a two-channel, digital-only device, supporting both coaxial and optical S/PDIF inputs at sampling rates as high as 192 kHz and per-channel sampling as large as 24 bits. He incurred the additional cost necessary to support both S/PDIF input types after his analysis of earlier Core Sounds customers' purchases suggested a roughly 50/50 usage split between the two formats in the recording community.

Inside PDAudio-CF you find few components, enabling Core Sound to squeeze the design into the CompactFlash form factor that many PDAs directly support and that, via a low-cost CF-to-PCMCIA adapter, notebook computers can also handle (see sidebar "A software smorgasbord"). Moskowitz included coaxial-to-TTL and optical-to-TTL S/PDIF transceivers, along with a small PROM to store the unit's CIS (card-information structure). One LED indicates a valid S/PDIF input signal; the other is available to application control and might find use in communicating the incoming sample rate or alerting the user to media-nearly-full or low-battery conditions.

An Actel antifuse FPGA houses most of the remaining logic circuits and translates the incoming serial bit streams into formatted data words for subsequent transfer over the CompactFlash bus. Why an FPGA? ASIC NRE (nonrecurring-engineering) charges would have been too extravagant, and minimum order volumes would have been too high for the customer demand that Moskowitz expects PDAudio-CF to achieve. He also valued the antifuse device's single-chip nature and its low power attributes over those of SRAM-based FGPAs. (PDAudio-CF draws an estimated 45 mA of current while operating and roughly 500 µA in standby.) His application also didn't require the hardware configuration flexibility that flash- and SRAM-based FPGAs offer.

So far, we've talked about the digital-domain conversions that the PDAudio system tackles. But how do the analog sound waves hitting the microphones' diaphragms transform into a digital bit stream, in the first place? In the next issue of EDN, the second part of this article series will discuss the other half of the PDAudio hardware chain, Mic2496, along with the system software that completes the symphony.

Contact technical editor Brian Dipert at 1-916-454-5242, fax 1-617-558-4470, bdipert@edn.com, and www.bdipert.com.