The increasing role of DSP in loudspeakers

In this week's tech review Professor Rumsey explores the role of DSP technology in loudspeakers
Publish date:
Social count:
In this week's tech review Professor Rumsey explores the role of DSP technology in loudspeakers

What are loudspeakers for? Reproducing what goes into them as accurately as possible? Reproducing the sound that the engineer heard in the control room? Reproducing something that sounds nice? The answer could be any or all of these, depending on the context. The loudspeaker in a mobile phone is never going to reproduce monitor-quality audio, but you might be able to make it sound nice. Unfortunately any consideration of loudspeakers can’t be divorced from one of the rooms in which they are housed. Move loudspeakers a foot and they sound different, too! Design engineers have been tempted in recent years by the potential for digital signal processing (DSP) to ‘correct’ the characteristics of loudspeakers and their interaction with the surrounding acoustics. Wouldn’t it be great if you didn’t have to worry about where you put the loudspeaker, or even whether the room had decent acoustics? For many installations, even at the high end, there are advantages to loudspeakers that are capable of being adjusted to suit their surroundings, and even the best loudspeakers still have anomalies that a little DSP-based tweaking could improve. At the very low end of the market, there are the crappy little transducers you get in mobile devices, which may not have much of a response below 1 kHz, or there are ‘difficult’ designs like flat panel televisions that don’t have any space for enclosures. These can certainly use some help. So what can you do with DSP in loudspeakers? Among other things you can adjust frequency response, adjust time response, design crossover filters, enhance the sound quality, phase or time-align multiple units, protect the drive units and adjust the directional response. The question, however, is should you do it and how much? Whatever the right answer, there is really no substitute for having good acoustics in the first place.
The traditional way
Traditional methods of loudspeaker equalisation use one-third octave or parametric EQ along with a pink noise signal to measure the response. The averaging effect of the filter hides the detailed peaks and dips in the spectrum, though, and the pink noise does not act much like a music signal in the time domain. The best you can achieve is an averaged adjustment to the magnitude response. In digital systems, low order minimum-phase filters can be employed, but they don’t do much about time-domain problems. Systems usually aim for a target response curve, such as a flat one-third octave response ±3dB, or might aim for something different like the X-Curve used in the movie industry (I’ll deal with this in an upcoming issue). However a flat curve is often not the most desirable thing to aim for, as it doesn’t sound very natural. (It’s not what we tend to hear in real spaces.) A microphone is needed to calibrate the response at the listening position, using either single or multi-point measurement. A spatial average of multiple points can give a better result over a range of positions when the listener’s head is moved. It’s a question of whether you want something highly corrected for one position, or something that works reasonably well over a range of positions. The latter is generally considered preferable, as there is no point in the engineer hearing great sound when the producer thinks it sounds terrible. Correcting for very large peaks or notches is another thing to avoid, so a more gentle approach is usually adopted. It’s all a trade-off between the amount of correction and side effects. Push it too far when attempting to boost a missing frequency, for example, and you can overload the loudspeaker or run out of amplifier steam.
Adaptive bass
The bass response of loudspeakers is strongly affected by where you put them in the room. This is because boundaries act as reflectors that modify the total power radiated. Bang & Olufsen’s revolutionary Beolab 5 has Adaptive Bass Control (ABC), which uses its own little mic that pops out at the bottom of the loudspeaker to measure the acoustical radiation resistance when test sounds are emitted by the loudspeaker. By comparing those measurements against stored references, the bass response of the loudspeaker can be adapted for the position in the room.
 Dereverberation filters attempt to deal with anomalies in the time domain response, reducing the decay time of problematic modes for example. Fielder, writing in 2003, found that small variations in measuring position (1.3 cm) gave rise to very large changes in audible resonances, and at that time he found such filters to be impractical in most cases, particularly in larger rooms. If the correction is restricted only to relatively low frequencies, say below 250 Hz, some control of room modes can be beneficial without generating undue side effects. It’s not generally a good idea to attempt to remove the room decay completely, as we don’t like listening to completely dead rooms. They sound weird.
Other clever stuff
 DSP in loudspeakers also allows the designer to create very precisely controlled crossover filters, which have things like linear phase, time delay compensation and unusual slopes. It also allows some directional control of the high frequencies, as a result of the control over driver interaction. By having a model of the loudspeaker parameters its possible to ensure that the DSP doesn’t drive the loudspeakers into regions that are likely to damage it. A model can effectively predict the temperature of the voice coil, for example.
To market, to market…
One well-known manufacturer that has brought loudspeaker DSP to the mainstream professional loudspeaker market is Genelec. Its AutoCal package is said to be useful when working in rooms with little acoustic treatment or when travelling, as it gets you a similar response wherever you are. The GLM version uses four parametric notching and/or shelving filters or bass roll-off, depending on the loudspeaker. Genelec’s network interface can connect up to 30 loudspeakers on a network, enabling automatic alignment and compensation for distances of loudspeakers from the listening position, for example. Subs can be automatically phase-aligned with a main channel. Oxford Digital’s TinyCore platform, on the other hand, has been used in systems like portable devices, mobile phones and flat-screen televisions to make frankly awful loudspeakers sound quite impressive. Using highly efficient code on dedicated silicon it’s possible to implement bass enhancement, loudness enhancement and frequency response correction, amongst many other things, at the cost of only a small amount of processing power and battery juice. That’s crucial in mobile platforms, especially when the loudness enhancement is achieved with no increase in the peak battery voltage requirement.



Dolby Mobile explained

The first of our in-house professor’s ten-minute technology reviews, designed to keep you up to date with the latest in audio engineering, writes Francis Rumsey.