Should concert halls use “assisted resonance”?

I read recently a very interesting article on the sonic deficiencies of some major classical concert halls and the possibility of using “assisted resonance” a.k.a. electronics and DSP to improve their reverberation characteristics.

It seems that there have been some expensive acoustic disasters over the years, where new concert halls have failed to live up to expectations. London’s Royal Festival Hall, which opened in 1951, is one of them, and it does seem a shame that a very expensive building designed and built purposely for music, has acoustics where performers apparently “lose the will to live”.

The science of concert hall acoustics has become better understood recently, but even if the hall works as predicted, there is no such thing as a one size fits all characteristic that is optimal for all types of speech and music. Even if there were, the number of people in the seats for a particular performance has a significant effect on the nature and length of reverberation. Starting from scratch, a good strategy might be to go all out for reverberation, with optimal dimensions and hard surfaces that could be covered with retractable curtains as required, but for many existing halls it is too late; they were built with the wrong materials and have the wrong dimensions, and it would be too costly to modify them.

And this is where electronics can supposedly come to the rescue. Microphones can be placed near to the stage and around the auditorium, and their output processed with DSP and fed to loudspeakers. Acoustically, the system can subjectively give the impression of changing the materials the hall is made of, or its dimensions. There are various commercial and experimental systems, and their use seems to be quite widespread.

Acoustic feedback from the speakers to the microphones is a factor that has to be managed, and is a limitation on the designers’ ability to create any response they desire (although modern DSP techniques reduce the feedback problem, but possibly with audible side effects). It was also the actual basis of one of the earliest attempts at electronic reverberation, known as “assisted resonance”, which was used in the Royal Festival Hall in the 1960s.

So reverberation enhancement ‘works’, but should it be used? Well, as a person who is pro the use of DSP in audio systems, I find myself unable to embrace it enthusiastically for classical concert halls, and I would much prefer to remain in ignorance if it is being used in any hall I might go to! I wonder how the majority of audiophiles would feel about it? Personally, I think I know too much about the reality of electronics and the people who inhabit that world! I don’t attribute the characteristics of art, craftsmanship, music and musicians to audio equipment. The reality is that audio equipment is created by technicians who are not steeped in art and have not served a musical craftsman’s apprenticeship. Do they have any business in a classical concert hall?

Electronic reverberation enhancement would no doubt be a mixture of approaches: custom design by computer programmers and acousticians in offices, and then physical construction of the system using standard microphones, amplifiers, DSP units, speaker drive units and custom MDF enclosures crammed into whatever corners and spaces of the hall that were convenient. Gaffa tape and the wearing of heavy metal T-shirts would be involved in the installation.

No one can say for sure what the ideal hall response should be, and even if they could it wouldn’t be achievable in every seat of the house. By definition we would be retro-fitting the system into an existing hall so would not have free rein to place speakers and microphones in all the optimal (if we even knew how to define optimal) locations. I have no doubt that, given a full 3D model of the auditorium, the acoustics with and without the electronic system could be simulated and plotted quite accurately, but this wouldn’t tell us the optimum settings for the system in order to maximise performance throughout the auditorium. If we felt able to specify criteria for “performance” then we could set a computer running with the task of finding the best compromise using simulated annealing or similar. We could go for best possible performance in the most expensive seats and not worry about everywhere else, or go for the best average performance throughout the auditorium, say. But notice what would have happened there. The future sound of classical music performances would have been set by:

  1. Arbitrary placement of transducers
  2. Sparse coverage of transducers
  3. Imperfect transducers
  4. An incomplete model of the auditorium and all possible configurations of stage, audience and placement of performers
  5. An incomplete simulation of the acoustics
  6. Arbitrary criteria for what makes ‘good’ acoustics
  7. Arbitrary criteria for distributing ‘good’ performance throughout the auditorium

You might say that something very similar would have occurred during the design of any modern acoustic-only concert hall: computer simulations and the setting of arbitrary criteria. But I would point out one crucial difference: a physical space and its acoustics form an entirely consistent system where the sound at any point is the sum of the direct sound and multiple delayed reflections. Even if the acoustics are not ideal they are consistent within themselves, and by moving throughout the space and sampling the response to impulses generated from known positions, multiple viable models could be constructed of the auditorium, which would gradually refine down to a single viable, consistent model. Electronically-generated acoustics cannot do this throughout the whole space. That is, they cannot be guaranteed to simulate a building that actually exists – certainly not at every position in the hall. Maybe the stationary human listener cannot hear the inconsistency, or maybe they can, but I don’t think it would be possible to guarantee a totally convincing effect at every point in the auditorium – unlike the case of an acoustic-only space no matter how flawed.

Other inconsistencies would include:

  • A ‘cognitive dissonance’ between the dimensions and materials of the hall and its sound (maybe it is obviously constructed from soft materials yet sounds like a stone church with different dimensions to the actual space)
  • A disconnect between the auditorium’s acoustic effect on sounds made by the audience itself (yes, coughing probably!) and its different apparent effect on the sounds made by the performers.

I realise that none of this may be the huge problem I am making it out to be. It’s just that I am wary of hype, and sceptical of the abilities of technicans! If a person who is adept at audio installation, mathematics or computer software tells me that they possess special powers enabling them to create the world’s finest concert hall acoustics with a few microphone capsules and polymer cones then I am not wholly convinced. Even if they are experts in their field (and this field could be very relevant like synthesising acoustics from first principles within 3D computer games) it does not automatically mean they can really do it.

The way I envisage the installation, technicians with laptops would pore over colourful charts on their screens, talking about “waterfall plots” and setting the system up to their own best guesses based on the methods they often use in sports stadiums, pop venues and shopping malls. Driving home at night they would be playing the latest Rihanna album on their car stereo, not Harrison Birtwistle; I would expect meaningful communication with the concert hall people to be limited simply because of the gulf of understanding between them.

In use it would become apparent that there were rough edges to the sound, but the concert hall people would be incapable of describing it in a way that could be understood by the technicians. Despite repeated attempts the sound would never be great. In short, the hall would become the offspring of two cultures that do not understand each other.

Over time the system would degrade. The microphones in very awkward-to-get-to places would gather thick layers of dust, changing their response. Occasionally, mysterious sounds caused by a spider living in one of the mics would be heard but never solved. Cables would be damaged by roofing contractors and repaired using Blu-tack and sellotape. Sonic anomalies like a metallic ringing particularly audible from rows C to E in the balcony would never quite be fixed. At some point the suppliers of the system would lose the original configuration files, making modifications impossible. Occasionally the system would pick up mobile phone interference. Yes, this is how I imagine such a system would be.

Would the system have fixed settings, or require a man at a mixing desk out in the auditorium to make proper adjustments for each performance just like a pop venue?

And then there are the ‘philosophical’ implications. I think that when we go to a concert we engage in some ‘suspension of disbelief’. Of course deep down we know that the hall is purposely-designed to sound good, and built for profit, and that the performers would rather be at home watching Game of Thrones that night. But we like to imagine that we have stumbled serendipitously upon a cultural happening with like-minded people in a magnificent hall built primarily as a gathering place, witnessing a group of performers doing what they do for the love of it. What happens there is spontaneous and not entirely predictable. Maybe the crowd will love the performance and the performers will feed off that reaction, or maybe they won’t. Maybe the organ will resonate in tune with the hall, or maybe because of atmospheric conditions and a particularly full house tonight it will be different and lend a new twist to the piece – without anyone analysing it of course.

Or at least that’s how I fondly imagine it. For me, electronic reverberation adds a new layer of ‘contrivance’. An analogy would be the use of electronics in a sports car to enhance the sound of the engine as heard by the driver (oh yes, they do that these days). There’s something not quite right about the direct, calculated, artificial ‘enhancement’ of something that is meant to be a fortunate by-product of something else. Even worse if it is created using technology in another ‘domain’ so that it is impossible to rationalise it as a “power valve” or whatever. Besides, it can never be perfect enhancement, for practical reasons such as that it is impossible for the car’s stereo speakers to create the low frequency vibrations that should accompany the harmonics we’re hearing. At some level, consciously or not, we may detect that the sound and physical sensations are not consistent with each other and decide that the whole thing is ‘fake’.

And then, if the performance is recorded there’s the idea that the recording I am listening to is a combination of real acoustics and some technician’s idea of good acoustics reproduced from imperfect speakers and then re-recorded by the mic! For better quality should the reverb instead be injected directly into the recording as a separate track? And maybe, just as we now recoil in horror at dated effects that were once thought to be timeless classics (e.g. gated reverb in the 1980s), will we only understand the true reality of having allowed young technicians to play with the hall’s acoustics when we listen to it again, decades later? Not a nice thought. But at least when the recording is re-mastered in years to come we can replace the reverb track, and its now-discredited algorithm, with the latest Steinberg Carnegie Hall impulse response – even though the performance was recorded in the Albert Hall.

As you can tell, I’m not keen.

UPDATE 20/06/15

Saw a link to a New Yorker article about someone’s experience with the Meyer Sound Constellation system. Reading it, I began to feel embarrassed: maybe my doubts about such systems are ill-founded, and in fact they are pretty much perfect. Maybe there really is a technical wizard who understands precisely how to solve this problem.

He clapped his hands; the sound resonated handsomely. Then he signalled for the power to be turned off. Suddenly, the clap was clipped and lifeless. The crowd gasped and applauded…

…it demonstrated the Meyers’ ability to conjure a plausible performance space. I was particularly struck by the sound of the tenor Nicholas Phan, in the Britten; the singer’s tensile strength and tonal colors came through intact. “It feels like a completely natural and real acoustic,” Phan told me afterward. “It even changes and feels different depending upon how full the audience is.”

But at the end, the author confirms what I might have expected:

All the same, I was never entirely convinced by the string timbre, especially the cellos and the double-basses. At full force, they had a slightly puffy, plastic quality—a familiar handicap of amplification that Meyer technicians haven’t yet overcome.

There is something philosophically disquieting about the Meyers’ work, as there is in any digital makeover of reality. Both at Oliveto and at SoundBox, the Constellation process never seemed obviously fake or too good to be true, and yet I had a sense of being ensconced in an audio cocoon. In the concert setting, I missed the thrum of floorboards under my feet—the full physical tingle of reverberation. Traditionalists will insist that there is no substitute for a first-class hall, and they will be right.


You can hear it in the street, see it in the dragging feet

Gosh, but I’m cynical. I was just reading a review of a very audiophile-ish amplifier where the remote control can only “trim” the volume – any major volume change involves getting up and turning the knob on the amp itself. This apparently results in a simpler signal path. What..? I was trying to think how and why this might be. Apparently the amp uses electromechanical relays to control the volume.

I have become so cynical about this whole business that I actually came up with this thought: maybe the manufacturers were having trouble with the remote control receiver and were worried that it might go haywire and go to full volume so they limited it to “trimming”. Then I came up with the rather more charitable view that it was simply a question of ‘bits’. Maybe for some reason their remote receiver only has a few bits, so they have to choose between coarse control over a wide range or fine control over a narrow range; they chose the latter. In the end, my final guess was that maybe the remote control applies to a separate attenuator in series with the one controlled by the knob rather than being ‘OR’-ed into the control of a single relay-based attenuator. This means that a quiet volume setting from a wide-ranging remote control would restrict the range achievable by the knob, and vice versa, hence the necessity to limit one of the controllers to trim-only duties.

You see, that’s what I have become. I read about a perfectly innocent and highly-desirable audiophile design feature such as a “trim-only” remote control, and my mind wanders off, coming up with these sceptical thoughts. This is not an isolated example of the way I regard the ‘high end’ audio industry…

I, too, am an electronics designer. The way my mind works, I would have identified the control of volume as absolutely the central function of my pre-amplifier and noted that the interaction between remote control and front panel knob was important. If one of my marketing ploys (careful of that cynicism…) was to have no nasty, dirty software in the box then it would certainly be a design headache; less so if I was ‘permitted’ to use a microcontroller (we are talking audiophile prejudices here and some people would hate to have a microcontroller in the same box as the audio signal). But I would have completely failed to realise that a trim-only remote is a perfectly viable product in high end audioland, and can even be marketed as a feature!

Other aspects to consider would have been volume resolution (how many steps do I have available) and logarithmic vs. linear response. If my gimmick (damn, it’s that cynicism again) was going to be relay attenuation then I would have to consider finite switching time and contact bounce. A clever arrangement of relays and resistors giving me high resolution with few relays, but where more than one relay changed state at the same moment, might result in some pretty ugly volume changes. That would need thinking about carefully.

The relay attenuator is one of those ideas that really appeals to the audio ‘high end’. It has everything: the ability to choose expensive resistors, the steampunk-ness of relays and the fact they make a mechanical noise when operated! They are necessarily large, physically, so in a long chain of them the signal is forced over quite a long, convoluted path with many solder joints. I imagine that no one worries too much about shielding – do they put the attenuator in a tin can? How would the punter see the lovely expensive resistors if they did?

And another thing that seems less than desirable: people used to go to great lengths to avoid putting switch contacts in series with the signal. If we are talking about micro-diodes being a noteworthy factor in cables, then here is a case where a slightly oxidised, damaged or contaminated contact really could have an effect (and of course even if the relay is just shunting to ground it is still in the signal path). Relays for small voltages and currents need to have gold contacts otherwise oxidation builds up and is not burned off by arcing (yes, controlled arcing is desirable in power relays) but even gold-plated contacts are not immune from atmospheric damage working its way under the gold. An alternative that might have worked quite well, the mercury-wetted relay, is banned in the EU. It is probably desirable that the contacts are in airtight chambers to prevent ingress of dust, moisture and, that great enemy of switch contacts, silicone which can “creep” over long distances and reduce the effectiveness of electrical contacts. Such arrangements may be found in reed relays, or hermetically sealed conventional relays.

So which has the more signal-degrading effect: an expensive relay attenuator or a £2 interconnect cable? If I had to bet…

In the rational world we now have electronic attenuators such as the ones that live in the multi-channel amplifier that I use for my system. These are a natural result of the need for digitally-controllable volume, and have many advantages (high resolution, high precision, high reliability, low distortion, low noise, small size, low cost). Their disadvantage is their high end non-marketability: they are hidden away in integrated circuits and therefore immune to being infused with musicality by skilled designers and artisans. Nor are they are well-suited for low depth-of-field photographs in brochures.

The fact that the ‘high end’ unquestioningly prefers the steampunk version with all its obvious faults to the ‘perfect’ modern alternative is today’s hi-fi industry in a nutshell.


An example of the difficulties of designing a relay attenuator:

I just looked at a beautifully-presented DIY project on the web. If we delve into the data sheet for the miniature relays specified by the designer, we find the following stipulation:

“Min. permissible load: 10uA at 10mV”

with an added note:

“This value was measured at a switching frequency of 120 operations/min and the criterion of contact resistance is 50 Ω. This value may vary depending on the switching frequency and operating environment. Always double-check relay suitability under actual operating conditions.”

To me, selling this as a commercial product would look like a bit of a minefield. The contacts in these relays are not in an airtight chamber and so are exposed to the atmosphere. While my prototype might work OK (as far as I could hear, but measurements might reveal a different story), what if I left it for several months without operating the contacts?

It would be fascinating to know which relays are used in commercial products. I wouldn’t bet much on them being radically different. The ones in a photograph of the interior of a certain high end amplifier certainly look very similar…