Feature

Sonic surprises

Dissecting the high-resolution hype

By Brian Dipert, Senior Technical Editor -- EDN, 2/16/2006

The market for add-in PC audio cards is rapidly evaporating, driven by the notably more robust capabilities of integrated audio in recent years and the in-progress system transition from PCI to PCI Express. When complete, this latter evolution will force audio-card manufacturers to redesign—a costly move in a market plagued by scant profits.

However, moving the noise-sensitive analog audio circuitry out of the EMI-plagued PC chassis still provides some advantage, and both FireWire and USB now provide sufficient bandwidth to stream a robust amount of digital data between a PC and a tethered audio peripheral.

Late August of last year brought a killer deal from Philips' online store to my email inbox: a USB-based PSC805 Aurilium 5.1 Channel External Sound Processor at 60% off its previous $50 asking price (the MSRP on this unit when originally introduced was $100). My immediate question upon viewing the message was, what does $20 plus sales tax, with free shipping, translate to in terms of included hardware? So I bought one with Prying Eyes in mind.

The short answer: Not what, based on the company's DSP heritage and the product's documentation, you might think. I also pursued this project with the hope of answering other burning inquiries, such as: Does the PSC805 deliver on its marketing promises? And can Philips still turn a profit at $20, or for that matter at $50? Read on for the details.

Read more Prying Eyes

The first question that always comes to my mind, when assessing a system's cost, is where it was manufactured. Predictably, the back panel of the PSC805 sports a "Made in China" sticker, although interestingly the system was subsequently packaged in Chile (if another sticker can be believed). My PSC805 certainly was a world traveler, as Philips presumably sought to keep total costs at a minimum. I'm curious how it got from China to Chile and from there to California (and points in between).

Click for larger image

The next issue I always think about: "How complex is the system-board design?" In this case, the answer seems to be "not very." Looking at the overall system-board shot (left), you'll notice that you can see almost all of the interconnection traces.

A qualifier before continuing: In lieu of a detailed discussion with the PSC805's hardware and software developers, I've relied on these traces, along with IC and system documentation, plus my own accumulated engineering expertise, to develop my hypotheses on the system's construction, hardware-versus-software partitioning, and functions. I don't promise that everything I'll say in paragraphs to come is spot-on, but I'm confident that a significant percentage of it is.

You'll also notice a lot of empty space on the board. Philips clearly made the system bigger than it needed to, probably because the company's marketeers calculated that a larger box would enable them to extract a larger sum from consumers' wallets. That is, they reasoned that the incremental cost of a beefier-looking system would be more than offset by the incremental price tag. I'm not showing the back side of the system board because there's little to see—only some more exposed traces, a moderate number of drilled vias, and solder points for through-hole-packaged components.

After perusing the IC documentation (specifically, the pinout diagrams) and the board traces, I strongly suspect that the system board contains buried layers for power and ground planes, but not for additional analog and digital signal routing. To clarify the discussion that follows, please note that the PSC805's back-panel connectors, top to bottom, are:

Toslink digital optical S/PDIF
RCA digital coax S/PDIF
Analog line in
Front left and right speaker analog out
Back left and right speaker analog out
Center and subwoofer speaker analog out
USB (the PSC805 is self-powered via USB, not requiring a separate AC or DC power connection)

Click for
larger image

The unit's packaging touts USB 2.0 support, but neglects to include the "full speed" (12 Mbps) qualifier. The Texas Instruments TUSB3200 streaming audio controller (visible at the center of the picture on the right, with its corresponding crystal above it), which self-identifies as a USB 1.1-to-I²C transceiver, reveals the truth behind the marketing propaganda. The PSC805 also claims to be "24-bit," enabling you to experience "high definition 5.1 audio on your laptop or PC."

Indeed, the Philips UDA1338H audio codec's DACs, which work in tandem with amplifiers to drive up to six speakers, accept 24-bit input sources, and its ADCs output 24-bit, I²C-formatted data. However, eight of those bits are largely wasted, because of the 16-bit sample-size limitation of the USB connection to the PC, coupled with the reliance on PC-based software for the majority of audio processing. More on this attribute in a bit.

Click for
larger image

Philips' product brief trumpets the fact that the PSC805 includes "24-bit/96-kHz audio hardware," and once again the picture reveals that it does: the AKM Semiconductor AK4353 DAC (followed by a headphone amplifier). However the USB bus's 16-bit sample size limitation, along with its 48-kHz sampling rate constraint, attenuates the hype.

Below the AK4353 is the ADC for the microphone input. The PSC805's documentation unfortunately implies that the signal boost from mic to line levels occurs in the postconversion digital domain (thereby magnifying not only the input signal but also its associated noise floor).

One big fiscal surprise upon popping off the side panel and peering inside: This high volume, cost-sensitive consumer electronics product contains a CPLD—specifically, a Lattice Semiconductor ispMACH LC4064V 64-macrocell device. Extrapolating from the circuit-board traces, I suspect that it performs a number of "gatekeeper" functions, such as:

Translating I²C audio to S/PDIF in order to drive the back panel's digital audio outputs
Driving front-panel LEDs and generating outgoing commands that route to the PC over USB, in response to PSC805 front-panel control manipulation, and
Responding to incoming commands coming from the PC over USB and controlling other PSC805 circuitry in response to those commands.

At the bottom of the printed circuit board, you can see the CPLD's on-board programming header.

Another big surprise: the PSC805 also contains discrete logic, in the form of two Philips 74HCT541 line-driver ICs, which work in tandem with the CPLD to selectively illuminate the front-panel LEDs.

The final surprise (although, in retrospect, not so surprising given the unit's price tag): there's no dedicated audio DSP. Instead, bass and treble boost, more general equalization, two-channel audio expansion, six-channel audio virtualization, reverberation, and other audio-processing algorithms (many of them provided by QSound Labs) all run on the host PC. One downside of this PC-centric processing approach is sometimes-noticeable latency as incoming audio routes from the PSC805 to the PC, is manipulated there, and then routes back to the PSC805 for playback. Another is the potential for stutter and dropouts as other USB peripherals compete for the scarce bus bandwidth and as the PC's CPU and other resources are distracted by other contending tasks.

The 16-bit/48-kHz limitation of the PSC805 is something that I've long pondered, particularly in light of the aforementioned "24 bit" and "96 kHz" marketing claims. Granted, the PSC805's software touts a "monitor" capability, so it may be possible to direct-route audio coming in over the unit's mic or line inputs to the PSC805 headphone or speaker outputs without disagreeable latency or 16-bit truncation (the >48 kHz sampling rate potential of the AK4353, however, likely goes unrealized).

However, any audio that flows across USB is 16-bit/48-kHz max per channel, per the PSC805 documentation's fine print. And given the cost-slimming PC-based-processing approach of the PSC805, most of the time you'll be listening to audio that has undergone at least one round-trip traverse of USB. Creative Labs has also gotten in legal trouble in the recent past for similar spec sleight-of-hand.

A little math helps clarify the picture. A single 16-bit, 48-kHz audio channel requires 768 kbps of bandwidth; a single 24-bit, 96-kHz channel is 3× larger. Six channels of 16-bit/48-kHz audio PC-to-PSC805 playback gobbles up just over 4.6 Mbps. Now, peer into the PSC805 literature and you'll find the words "full-duplex," meaning that (at least theoretically) the PSC805 is capable of simultaneously recording (transferring from the PSC805 to the PC) while playing six-channel audio. The recording task would consume 1.5 Mbps of audio bandwidth. Worst case, that means the PSC805 alone can consume more than 6 Mbps of USB 1.1 (USB 2.0 "full speed") bandwidth. And in the real world, that's about all the usable USB 1.1 bandwidth that you can count on.

Competitive analysis also clarifies the PSC805's position. The M-Audio Transit is also less than $100, and it really supports 24-bit/96-kHz audio. However, it's only a two-channel audio peripheral and it supports sampling rates above 48 kHz only in half-duplex mode (record or playback, but not both at the same time). Notably too, unlike Philips, M-Audio does its own driver development versus relying on Windows' built-in USB Audio drivers, which helps explain why the PSC805 is only capable of running under Windows 2000 and XP, not on earlier Windows variants—a cost-saving but somewhat market-limiting move on Philips' part. The now-obsolete Creative Labs Sound Blaster Extigy claimed to support full-duplex, 24-bit/48-kHz, six-channel playback and two-channel recording, a feat it achieved by switching to ADPCM lossless compression for high-channel-count USB 1.1 transfers.

In another bit-slimming (and CPU horsepower-saving) move, the Extigy also handled Dolby Digital decoding on its own; you only needed to pass it the compressed Dolby Digital bit stream, not the already decompressed and much larger six-channel audio payload. This greater hardware complexity came at greater cost, however. The MSRP on the Extigy was $150.

The subsequent (and now also obsolete) Creative Labs Audigy 2 NX ($100) switched to USB 2.0 for improved bandwidth, while the still-available USB Live! 24-bit External ($50) is similar to the PSC805 in not only original sales price but also features (and limitations). The even more economical USB Sound Blaster MP3+ ($40) supports only two-channel analog-audio playback. And I can't help but wonder if some or all of these USB peripherals are downsampling (in rate, size, or both) the audio that routes to some of the speakers, in order to further reduce their USB payload.

Bottom line: The PSC805's not a bad deal for $20 (or, for that matter, $50), but don't mate it with an anemic or pre-Windows 2000-powered PC, or rely on it to accurately capture your next high-resolution audio masterpiece.

Author information

You can reach Senior Technical Editor Brian Dipert at 1-916-760-0159, bdipert@edn.com, and www.bdipert.com.

Editor's notes:

For a followup to this article, see this post in Brian's blog, Brian's Brain.
The above is an extended version of an article that appeared in shorter form in the print edition of EDN. This PDF file shows the printed version.

Free Magazine Subscription

Printer-friendly version / PDF