Riding the Digital Wave: How I Moved From Analog to Software
From analog loudspeakers to DSP, FIR filters, and measuring sound for 80,000 people in a stadium: how I stepped onto the bridge from hardware into software.
When I started at Meyer Sound, everything I built was analog. Loudspeaker systems made from real components on real boards, no microprocessors, no operating systems, no software. Just analog circuitry doing its job to create sound. And I loved it. But I could feel the ground shifting under the whole industry, and I wanted to be standing where it was moving to.
The audio world was going through a huge transformation. Everything consumer was heading toward digital, CDs, CD players, all of it. But in professional audio, none of the early digital stuff was reliable enough, and the quality wasn't good enough. So the real story of that era was digital signal processing slowly getting good enough to trust for professional applications. I happened to be in the industry at exactly the moment that tipping point arrived, and I threw myself at it.
I'd actually started before I ever joined the company. My undergraduate thesis took a loudspeaker system and ripped out all the analog circuitry that made it work, replacing it with digital signal processors, two Motorola DSPs, one for the high-frequency channel and one for the low. Then I built a measurement system around it and designed very long finite impulse response filters, FIR filters. Those long filters let me compensate for every little peculiarity of the speaker, so that when you sat in one spot you got an absolutely flat, faithful reproduction of the sound. That project convinced me that this was where I wanted to spend my energy: building the next generation of speakers using this kind of technology.
Here's why the problem is genuinely hard, and why I found it so rewarding. If you look at the frequency response of most loudspeakers, the ones in your home, whatever you listen to, it's not perfect. Headphones get much closer to perfect, but a headphone can't move real air; there isn't enough low-frequency energy to fill a room. The moment you step into the physical world of real sound reproduction at real volume, the systems get complex fast.
Now scale that up. You load a stadium full of speakers and you have 80,000 people listening. How do you actually know the people over there, and over there, and everywhere in between, are getting sound they can hear clearly, the intelligibility, the quality, the balance of highs, mids, and lows? You know it through measurement. We used FFT-based frequency-domain analysis, fast Fourier transforms, to look at the frequency response of the whole system in the room. Doing that in a stadium, with 100,000 people present and a hundred speakers flown around the stage, is where the real difficulty lives, and that difficulty was exactly what I loved.
So I built systems to solve it. I implemented a lot of signal processing on Texas Instruments DSPs, embedded processors built specifically for floating-point operations. These things could do on the order of hundreds of millions of floating-point operations per second, which at the time was remarkable. My systems ran the analysis, and then went out on the road so that engineers could measure real installations and confirm the speakers were performing.
That was the bridge in my career from hardware into software. I never stopped caring about the physical device, the speaker, the board, the air being moved. But digital signal processing was where the physical and the computational met, and stepping onto that bridge shaped everything I've done since. If you can see a wave like that forming in your own field, my advice is to go stand where it's about to break.