When Digital Audio Sucks

Where I work, we spent a fortune on a Klotz digital audio distrubution system, German designed, and used in a lot of radio stations the world over.  Its been fantastic.  We have 10 studios, and hundreds of audio sources and destinations - the system routes it all wherever we want without degradation or noticable delay.  In fact, upon a single core of fibre cable can travel 256 channels of 24bit 48kHz stereo audio.  It's really cool.

However, there are some disadvantages to digital audio, and no, I'm not going to get into the argument over analog v digital audio quality!  After all, in a radio broadcast environment, digital audio quality (at the right spec) is more than enough and its ease of distribution is the major benefit.  However, digital audio does bring with it a host of painful quirks, and most of them are to do with word clocking. Unless you need to distribute audio here, there and everywhere, do youself a favour and STAY in the analog world.

We have an automation or "playout" system.  It plays all our audio automatically, and its connected via several AESEBU digital audio channels to our distribution system.  A couple of years in, we noticed that occasionally, the very start of some audio events were clipped.  A lot of the time, it wasn't noticable, but if the event was a song which started with a high transient such as a drum hit or guitar strum, or if the event was a spot starting with a voice-over, it sounded terrible.  So we went hunting... and hunting... and hunting.  We found nothing, after talking to every software and hardware vendor involved in that part of the audio chain.  We updated drivers, downgraded drivers, experimented with performance enhancing drugs, but still nothing.

I don't get many epiphanies, but I did get one about this problem, in the middle of the night.  I was lying awake, wondering what had changed to introduce the problem.  Everything was the same, hardware, software, wiring - situation normal.  Aha!  We had a new Receptionist!  You can never underestimate the impact of someone new in a key role like Reception...  But no, that wasn't it. 

Aha again!  We had introduced some different sample rates, because we had made the decision to change our house sample rate from 44.1 to 48kHz so that we could avoid resampling all the way through our satellite network.  That meant that our automation system contained old audio that was at 44.1kHz and new audio at 48kHz.  No problem.  Klotz can convert sample rates on the fly - in fact, we paid extra to have SRC on AES inputs in anticipation of this very problem.  Turns out though, there's a catch.

After querying a Klotz engineer about the way Klotz SRC works, it confirmed what I suspected might be the problem.  If the sample rate feeding a Klotz AES input changes more that 0.4%, the input is MUTED for the amount of time it takes to adjust the word clock.  MUTED!!!  Naturally, it has to be muted or else you might here something even yuckier, like a pitch change or noise.

So, on the infrequent occasion that an audio event was fed to a Klotz input at a great than +/-0.4% different sample rate, a couple of hundred milliseconds of its front would be chopped off.

The fix was easy, but expensive.  A new set of beautiful Audio Science audio cards for the automation system.  These cards can be instructed to output a consistant sample rate, regardless of the rate of the file they are playing.  Hey presto - it works beautifully.

A waste of Klotz SRC?  In this case - yes.  Could we have done the same thing in the analog realm with a lot less hassle?  In this case - yes.  Would our listeners in the deep jungle of Indonesia, China and India picked the difference?   Hmmmm... I don't think so, but at least they're hearing the start of their favourite songs.