Digital Audio Done Right

Sunday, April 19, 2009 by Peter Clarke

In the 1990′s and early 2000′s the debate about analog versus digital audio raged freely. There were many arguments on both sides that over time seem to have simmered down into a grudging truce between the most vehement opposition. In recent years I’ve found many of my contemporaries feel that whether you go with a purely analog system, or a purely digital system is more a matter of function and cost than anything else. The sound quality of digital components has greatly improved over early offerings and our understanding of proper application has also improved.

I recently set out to see how I could configure a typical live sound system for maximum sound quality in the digital domain. I chose to focus on two major components, the digital audio console and the DSP (digital signal processor). For this test I used a Yamaha M7CL-32 and a BSS London BLU-80. The BLU-80 was loaded with both digital in and out cards as well as analog in and out. The analog output of the BLU-80 was connected to a pair of JBL LSR 2328 studio monitors with an LSR 2310SP subwoofer in line. Commonly, sound systems are sent into the field with digital components interconnected with analog means. The drive lines from the console to the DSP are analog as well as the outputs from the DSP to the power amplifiers.  I believe this is done for several reasons. First, analog interconnects are easily understood by even the greenest of audio professionals. Second, most companies have analog snake and drive line systems already in place. And third, when your livelihood as a live sound engineer depends on making the show happen without interruption or chance of failure, you tend to go with what you know and trust will work flawlessly every time.

I had the console connected to the DSP down a 300′ snake which would be typical of any number of shows. I then listened to tracks off a CD player with several of my co workers and noted the results. All of us felt the system sounded perfectly decent. There was nothing glaring or obviously wrong. This represented the status quo. Next, we switched the BLU-80 to the AES-3 input that was coming from the console’s 2-track AES-3 output. This was being fed down 300′ of AES-3 cable to simulate the length of the analog drive lines. We all noticed an immediate improvement in the quality of the vocal range. The track used for testing was “Sand In My Shoes” from Dido’s 2003 album “Life For Rent”. The first two minutes of this song are excellent for making a subjective analysis of audio quality. The vocal recording can sound harsh and a touch brittle with a bit of sibilance if not played back on a quality sound system. When the system was connected entirely analog, The voice sat forward in the mix and had a slightly harsh edge to it. When we listened via AES-3 interconnect, the voice sat a little farther back in the mix and greatly smoothed out. Other details in the mix began to emerge such as the sound of wind effects and short bits of looped material and also the bass tightened up considerably.

To take the experiment farther, I wanted to see what effect would be had by clocking the M7CL-32 from a better clock source. There was a Yamaha PM5D-RH nearby, so I connected a 25′ 75Ω cable from the word clock output to the word clock input of the M7CL-32. We listened to the track again and were stunned by the results. The mix had changed again and in a very pleasing way. The loops and bits that had been in the background now came farther forward and the vocal that had seemed too forward in the purely analog interconnect test now sat neatly back in the instrumentation. We started hearing things in the mix that we simply hadn’t been able to make out in the purely analog test. Of note was the improvement in the bass. The leading edge of the bass guitar notes was discernibly sharper as were the kick drum hits.

Based on this success, we thought that if one good console could clock another and make it sound better, what kind of an improvement would we get from a dedicated outboard clock? I grabbed an Apogee Big Ben off the shelf and connected it to the M7CL-32 with a short 6′ 75Ω cable. We listened again and were pleasantly surprised by the results. The vocal smoothed out even more, more elements were audible in the mix, and the bass tightened up further. It felt like the moment in a movie theatre when the projector has been out of focus during the previews and the projectionist finally remembers to focus the lens right before the feature.

I had honestly not expected these results to be so dramatic. I figured the digital connections would sound better simply because there were fewer conversions involved. In the purely analog setup, the CD player has to go A to D to get into the console, D to A to get out of the console. The same process happens through the DSP. With the digital interconnect, two A/D stages are eliminated between the console and the DSP. That was part of the improvement in sound quality; the other was clocking. An AES-3 signal has embedded word clock inside it so when connecting from one device to another clock is carried along. By connecting the Apogee Big Ben to the console, not only were all of its converter processes improved, but subsequently those of the DSP which was being clocked from the console.

In the real world of setting up and operating live sound systems, I don’t believe there is any more or less risk in making digital interconnects than in making analog interconnects as long as it is done properly. There is no more likelihood of the failure of one output connector on a console than another be it digital or analog. The same is true of the DSP. In the BSS London product for instance, all I/O is accomplished on 4 channel cards. There would be no more risk of one type of card failing than another. The cable architecture for AES-3 is easily as reliable as analog and due to the tolerance the 110Ω cable is built to, perhaps a little better. No special skills are needed for the cable. If you can solder an XLR connector on an analog line than you can solder an XLR connector on a digital line (it doesn’t even care if you get the polarity wrong). As a side benefit, a single twisted pair of AES-3 carries two channels of audio, so the amount of copper being carried is cut in half. Certainly the broadcast video world understands this as AES-3 has been their audio standard for many years.

Based on the clearly audible results of this test, I believe that to do digital audio correctly it must stay in the digital domain from end to end and have the highest quality clock source available. For the engineer driving this kind of system, it will be incumbent upon their mixing skills to present the best quality experience to the audience.