Martin Colloms IOA Conference Blog 'Audio Matters'

'Reproduced Sound -22', held at Oxford Nov 3-4 2006.

I went to this event feeling a mite guilty that I last attended back in 1991. Part wearing the HIFICRITIC technical editor hat, I picked my way amongst the often highly erudite presentations. Topics ranged from wave field synthesis and acoustic holography abetted by advanced Fourier transform methods and software, to subjects of more direct relevance to the Hi Fi community. The following summary is unashamedly audio oriented and in consequence does not fairly represent the broad spectrum of professional work presented at the conference. (see IOA.org.uk)

Cherry picking with a vengeance, and summarising the titles, we have 'Crossover network capacitors and sound quality' by primary author Paul Dodds, University of Salford, who reported on the progress into capacitor sound quality, construction and loudspeaker performance at 18 months of a two year investigation, DTI supported. In particular, simple axial measurement of frequency response for a reference speaker showed significant measurable and audible differences between capacitor types even though their electrical performance was almost beyond reproach. Further work has confirmed the suspicion that mechanical self-resonances, which are themselves dependent on the type and parameters of construction, were also affecting sound quality, adding coloration in the treble band. Industrial Capacitors (ClarityCap, www.claritycap.co.uk), are associated with this research. Here at last was some learned confirmation of effects on which reviewers have long speculated.

Keith Howard (of HFN and Stereophile) authored the invited lecture 'High Definition Audio, How is it delivered' and he succinctly and unambiguously exposed the cynical attitude of some record companies and their supposed HD audio releases by analysing their spectral content, their bandwidth and peak levels via a computer, rendered much easier than for previous work by the digital nature of the recordings themselves. Those music data numbers may be crunched to find out what the tracks are made of. He found that some HD releases, both DVDA and SACD had plain CD content, i.e. the CD layer and HD layer were the same lower resolution, while no labelling was present to help the purchaser decide. Keith indicated the he also suspected that in a very few cases hybrid discs were rendered with a CD layer of lower playability (e.g. reduced pit contrasts) so that this layer would play with reduced fidelity relative to the primary SACD layer. His analysis showed other unwanted manipulations including re-sampling, and a failure to go back the analogue master for an HD release when a wide bandwidth analogue master was available as a source. Fortunately a few specialist audiophile HD discs were found to be of genuinely wider bandwidth and quality.

The BBC's Andy Baker described some details of the key component of the newly introduced digital audio broadcast chain, founded on a massive hard disk unit. He recounted that consideration of data reduction, to try and reduce the data transmission load and the storage of audio, led them to choose uncompressed musically robust, lossless formats on hard drives. The contents of the drives are dynamically copied, in triplicate, at exclusively linked networked locations, drives geographically separated for reliability. The data formats including track info and other metadata and automatic logging for rights management. Tracks are pre-cued and gain levelled to uniform peak output (but not limited). Whatever format a program was broadcast in, the only data losses present are now those determined for that channel format at the point of release. He explained that 90% of BBC Radio output is now via this hard disk system. For example, 'BBC7,' a new digital station, operates completely automatically for more than 20 hours a day in conjunction with the new disk system and its inbuilt control capability. It is good to know that for the storage and distribution chain, the BBC is maintaining archive quality, even if the fidelity of digital radio often leaves much to be desired.

Mark Dodd of GPAcoustics (KEF division), recounted his experiences in attempting to master the directional response and smoothness of a tweeter dome mounted down a cone. You could view the cone as a form of horn loading for the tweeter. We already know such a technology as the patented UniQ invented by former Technical Director Laurie Fincham, and which is now free issue in some Countries after its 20years of patent life. It is appropriate that Mark's success at improving the technology has resulted in a new patent filing, just in time for KEF, assuming of course that the application ultimately goes to grant. Mark is known for his work at Tannoy, in particular on the famous Dual Concentric series, notably the still current Tulip waveguides, and also for the mechanical resonance damping of film type crossover capacitors. Mark offered a refreshing change of view, eschewing the tradition and nomenclature of horns and keeping his eye on the wave properties of sound. Adopting a synthesis stimulus, which has both time and frequency content, he first explained that a planar baffle was in fact a 180degree conical waveguide, and then showed that while its behaviour was innocuous enough for a 1" flat piston radiator, it causes all kinds of trouble for a rigid dome profile. Significant response errors were present right into the low treble; so much for our ideal of the infinite baffle when that practical radiator shapes are involved. He then showed that there was an optimum shape for a practically sized, modified flare, 'cone' shaped waveguide, where the sound radiation for a given practical dome radiator could enjoy a pleasantly uneventful route to the air load beyond. In consequence, exceptionally smooth and consistent on and off axis responses were predicted from the modeling, newly strengthening the coincident driver technology. Improved KEF Uni-Q loudspeakers are to be manufactured which will benefit from this advance in waveguide theory and design. Of the many illustrations, I can reproduce a couple of revealing overhead views of the sound radiation pattern. Uniform colour density is considered optimal. The first is for a rigid hemispherical cap radiator on a flat baffle, i.e. 180 degree included angle. The dome output has a surprising rough acoustic ride in trying to get clear of the dome.


Figure 1 (from ' Improved Diaphragm and Waveguide Geometry for Coincident Source Units' Mark Dodd. 2006 IOA, RS-22)

The second graph, though here plotted for a narrower 75 degree radiation angle, is for an optimal 80 degree angle dome, correctly placed down an 80 degree, shaped cone 'waveguide,' and shows great promise for impressive control of the uniformity and quality of sound radiation.


Figure 2 (from ' Improved Diaphragm and Waveguide Geometry for Coincident Source Units' Mark Dodd. 2006 IOA, RS-22)

In my own paper I discussed the need, or not, of an ultrasonic bandwidth for higher fidelity sound reproduction and hope to develop a version of this content for HIFICRITIC in the future. I concluded that a wide band was worthwhile for recording and production, the case for replay remained unproven.

Juergen Peissig of Sennheiser explained and demonstrated a novel 'parametric' loudspeaker. This was a compact panel carrying an array of small transducers working in the ultrasonic band. It is designed to create a very narrow, projected sound beam which could position a compact sound image, e.g. for sound effects, on remote, reflecting walls. The operating principle relies on the non-linearity of the air at very high sound intensities in order to demodulate the narrow beam carrier.

With an operation essentially limited to the mid and treble range due to effective limits in respect to allowable modulation depth, it could invisibly reproduce the sound of running water and birdsong at positions all around a large room with intriguing effectiveness. With a rated output of 120 dB, 4 metres, for its 100kHz modulated carrier, it needs to be used with care, though when directed at them at more than 4metres spacing no one could hear the ultrasonic beam itself except in one case. Here for one participant it was powerfully and painfully demodulated by his hearing aid driven to overload.

On the subject of bass response, extensive listening tests for product review have indicated that the more extended and well damped (but not excessively damped ) the bass response is, the better is the overall sound, this including the bass. Lara Harris University of Southampton, was using MTF (sorry, Modulation Transfer Function) as the stimulus for controlled listening tests conducted in an echo-less test acoustic chamber, and was able to report good correlation between this measure for the documented low frequency behaviour of an augmented loudspeaker, and the perceived sound quality itself. There is more to do here.

One of the more informal projects was a workshop run by Paul Malpas of Arup Acoustics. He organised a capable PA with active two-way speakers lent by JBL, type LSR28p, aiming to demonstrate to a panel of 40 plus members the range of music quality expressed by various lossy compression methods, MP3, and AAC improved, such as used on Apple Ipods and similar devices. Pre-recorded tracks were made at nearly all the available bit rates, 64,80,96,112,128,160,192,320kbps, and the 1.411kbps CD resolution WAV material. These tracks were compared with the lossless originals, generally rendered from CD originals stored as WAV files.

Once the audience of more than 45 had settled down it was clear that all the data reduced options degraded sound quality, even under such less than ideal lecture room conditions. On hearing these losses delegate Dave Bruml of Funktion One, who is involved in the design and specification of high quality club and discothèque sound systems, commented that this was nothing new to him, as he had found that some 'celebrity' DJs who were arrogant enough to bring their play list on an Ipod, produced a sound quality which in dynamics, punch clarity, rhythm and not least absence of fatigue, was well below the standard familiar from direct feed at CD resolution, and that most clubbers knew that too. In some cases there was correlation between the number of active dancing ravers and the technical sound quality and origination of the tracks, and you probably don't need to guess how this one goes. This rather contradicts the popularly disseminated presumption that we are supposed to happily dance with our data reduced 'pod music feed.

One result appeared to contradict received opinion concerning the supposed low fidelity of modern levelled-up and highly compressed pop material. One might have thought that lossy MP3 or equivalent Ipod coding would not significantly damage such material but this was not the case. For the Artic Monkeys excerpt, for me 'heavy thrash' music, we found that several aspects were corrupted. The lead guitar lost some twang and bite, the top end was slightly dull, the drums were soft and plodding, the vocal lost some articulation and expression, it was less involving and it lacked rhythm.

Further, when listing to the data reduced version it seemed harder to follow the plot, even fatiguing. Reversion to the lossless original gave a sense of relief and aural comfort, even on the Artics. For those subtler and lower density tracks some audience members found discrimination more difficult, and an explanation was put forward. Because the lossy coders are trying to conceal their mechanics and pack the more fundamental sounds into smaller places very dense, loud-all-the-time music is actually the most difficult to code well, while episodic quieter and simpler stuff is handled with somewhat greater accuracy. Nevertheless what went missing on this second type of material was transparency, image depth, low level reverberation, focus, and again, rhythm, but here of a of more subtle degree.

My notes indicate that the maximum 320kbps compressed coding was adequate for background music, but not for high fidelity. At the lowest rates, 64kbps was clearly lacking in impact with poor dynamics and an audibly distorted kick drum. Improved AAC at this rate was better, still more so at 128kbps and obviously beat straight MP3 at this data density. I noted (without knowledge of the format), that was about 1/3 of true CD quality. The odd thing is that the more you try to listen for the errors the more difficult it becomes. Its like a moving target, it really tries to conceal its errors. What surprised a number of participating 'acousticians' was the sense of relief when the replay was switched from data reduced to the real thing. They had not anticipated this. It is as if the coding artefacts have been design to confuse, to avoid direct recognition, but the result is disturbing and fatiguing, but understood when the blurring and confusion is removed.

At this point we have to point the finger at the designers of data reduced music formats, who in order to save on then costly memory, sold the reduced data coding idea to the industry on the premise that they knew that we would not hear what they were taking out.

The industry is going to have to wake up to the fact that lossy coding and High Fidelity are simply incompatible; lossy coding sounds broken, and now we forever have a duty to explain and demonstrate the difference.