Some of you out there may have tuned into a recent webinar I gave for Fedele De Marco, who runs Il Lato Oscuro della Fase: Designing, Verifying, Tuning, Adapting, which focuses on acoustical education for audio engineers. In the webinar, I discussed the challenges live sound engineers face in delivering a consistent, high-quality listening experience to each audience member, particularly in the subwoofer band (<100 Hz).
While a ground-based centrally distributed subwoofer array is a common and straightforward solution, it can be impractical and unsafe in certain situations (which is a topic for another day). Often a left/right subwoofer system (ground-based or flown) is a better choice. The problem with a left/right configuration is that there will be severe comb-filtering, causing inconsistent horizontal coverage. To avoid this issue, the left/right signals must be decorrelated. Existing approaches involve unique EQ applied to each side of the PA (which isn’t great from an efficiency viewpoint) or the use of allpass filters (which generally result in a reduction in audio quality).
My former PhD student, Jon Moore (now working for Midas Consoles), and I took a decorrelation approach developed by Prof. Malcolm Hawksford for DMLs and adapted it to operate on conventional loudspeakers. This approach is called Diffuse Signal Processing (DiSP). The DiSP algorithm has been devised to decorrelate audio signals without any perceptual artifacts. While the primary intended application is for left/right subwoofers in live sound, it could also be applied to:
- Horizontal subwoofer arrays (live sound)
- Stage monitor systems (live sound)
- Cinema B-Chains
- Home cinema (even with a single subwoofer)
- Loudspeaker crossover networks
- Headphones (for externalization)
- Speech intelligibility (possibly… more research is needed here)
A good overview of how DiSP works is given in the recording of the webinar, which can be viewed here:
During the webinar, I tried to run a live demo of DiSP, but didn’t quite get all the bugs worked out within the online presentation system. I’ve therefore redone the demo, which you can listen to in the video below. The first half explored static DiSP, which is best for non small-room applications. To deal with the problem of correlated early reflections, dynamic DiSP was developed, where the decorrelation is varied over time. This is the focus of the second half of the video. Apologies for the clicks and pops in the demo when I switch between the unprocessed and processed audio streams – the matlab program running the demo wasn’t designed for this sort of thing! (audio sample credit: Newton Faulkner – Teardrop)
One remaining bit of work regarding DiSP to to optimize the algorithm to deliver the decorrelation with as low latency as possible. At the moment the latency is too high for use in live sound. The reduction of latency should be relatively straightforward, I just need to find the time to do the work!
In the meantime, here is some selected reading on DiSP if you’re interested to learn more about its development over the past 20 year or so:
- Hawksford (1999): MATLAB Program for Loudspeaker Equalization and Crossover Design
- Hawksford + Harris (2002): Diffuse Signal Processing and Acoustic Source Characterization for Applications in Synthetic Loudspeaker Arrays
- Hill, Hawsford + Newell (2016): Enhanced Wide-Area Low-Frequency Sound Reproduction in Cinemas: Effective and Practical Alternatives to Current Calibration Strategies
- Moore + Hill (2017): Optimization of temporally diffuse impulses for decorrelation of multiple discrete loudspeakers
- Moore + Hill (2017): Dynamic diffuse signal processing for low-frequency spatial variance minimization across wide audience areas
- Moore + Hill (2017): Applications of dynamic diffuse signal processing in sound reinforcement and reproduction
- Moore + Hill (2018): Dynamic diffuse signal processing for sound reinforcement and reproduction
- Moore (2019): Dynamic Diffuse Signal Processing in Sound Reinforcement and Reproduction (PhD Thesis)