The Scots’ Church Rieger

by Stephen Phillips

The fourth in a series of articles dedicated to what many regard as Australia’s finest church organ – and the story of its hard-won virtualization by Inspired Acoustics.

Part Four: Putting It Together with SPIRAL

The long awaited consummation of all this effort would take place, ironically, in the vanishingly low noise floor of 24/32-bit digital architecture. We will see what perils and pleasures might await this final phase of bringing the Scots’ Church Rieger, in virtual guise, to an expanded audience of players and admirers. We might also expect to learn something of the meaning behind the rather curious choice of title for this final instalment. What, 'pray tell' – you may well ask – is this 'SPIRAL'? (I can almost see the quizzically raised eyebrow.) Let's find out...

In the first installment of this blog series, we traveled back in time to the very distant past, to the ancient rocks of the Australian wilderness, before coming forward again to the time of white settlement, the gold rushes, and the early history of the Presbyterian Church in Australia.

As we finish out this extended look at the Scots’ Church Rieger and the battle for its successful virtualization by Inspired Acoustics, we once again go back in time – not nearly so far as previously, this time just upwards of ten years – in order to properly tell the story of the period lasting, as of this writing, almost exactly 24 months, occupied by the development, release, and subsequent revisions of the virtual organ product.

Prelude to SPIRAL: NDB

Our journey back – as it did in mid-2012 for the sampling effort Leader, Dr. Csaba Huszty, at the conclusion of the long weeks of work in Melbourne – leads to Budapest, Hungary where, ten years ago, a new and seminal product had been released to global acclaim – the Notre Dame de Budapest Pipe Organ Samples (NDB). This new ‘sample set’ (originally prepared for the Tascam Gigastudio platform – one of a number of diverse platforms that would see this and subsequent IA creations) achieved what no product up to that time had been able to – the convincing portrayal of a large symphonic organ (in this case, two of them) including very plausible swell chamber, tremulant, and crescendo behavior, the sounds of the organ engine, registration action, valve speech, etc. all included – even the sampled squeak of trouser-cloth on varnished wood as the organist 'assumes the position'! An exhaustive and pioneering undertaking, it fully delivered what it was designed to do. Avoiding any possible side-effects of looping the samples, and taking advantage of the hard-disk streaming capability of Gigastudio, these beautiful samples ran up to one full minute (!) in length. The promise of well-engineered convolution reverberation was also most effectively demonstrated – the technique similarly applied to the swell enclosure progressive filter characteristics – providing a seductively real sophistication of acoustic presentation not since seen in a single product. Importantly, the laborious and time-consuming reiterative proprietary noise reduction performed on these early high-resolution samples no doubt was a major contributive element to the undeniable success of the project, and provided an important pillar upon which Inspired Acoustics could base further research and development. Original research in the field of Impulse Response and convolution processing, first exhibited commercially in this offering, would be another mainstay of the emerging company’s activities.

Methodology refinements: PAB

As an early adopter of NDB – almost certainly, the very first on the Australian continent – and (at the time) being not overly-familiar with computers and sampling software, I inevitably found occasion to take advantage of what proved to be the excellent technical support provided by this specialist company. I could not know that, in these generous and carefully considerate communications, the field was being prepared for a friendship, of great personal and ultimately professional value, between Dr. Huszty and myself. The first of three trips to Hungary (to date) was made in October of 2005; a second in November 2007, with a third, only one year ago at the time of writing which, while for happily auspicious personal reasons, understandably trespassed on a good deal of organ-related ground, including the second of two visits to the Basilica of Esztergom in the far north of the country. A considerable volume of email traffic across this time, apart from confirming this clear early rapport, delved into many fascinating (and from my point of view, somewhat mysterious) technological explorations, including significant leaps forward in multi-loop sampling, something that would, by early 2008, be very concretely demonstrated. In typical Inspired Acoustics style, this was to be no half-hearted effort.

The instrument of choice for this initial foray perhaps naturally suggested itself, and for a number of reasons. In the several years since NDB had stunned the sampling world, on the east bank of the Danube, just north of the Rákóczi Bridge, had sprung up the state-of-the-art Palace of Arts, a superb world-class performing arts centre. In a world-first, as the centre’s showpiece, the Béla Bartók National Concert Hall, took shape, simultaneously with it, in a coordinated effort, did the 92-stop Pécsi-Mühleisen symphonic grand organ, designed by distinguished local talent led by the elder statesman of Hungarian organists, István Baróti, of Esztergom Basilica fame. The second of the organ’s two consoles, freely manoeuvrable upon the stage, allows for more intimate interaction between the organist and the audience; it is this console we see modeled in the Palace of Arts Budapest Pipe Organ Samples (PAB).

This new product instantly broke the record for the largest virtual organ to that point (a record, by my reckoning, which it still holds, though others have nibbled at its chunky heels). This landmark instrument, further extended in the monumental 149-stop Gravissimo incarnation – the magnitude of this high quality virtual extension, again, on an unprecedented scale – has been embraced by professionals, teaching academies, and organ enthusiasts, even inspiring treatises drawing upon its extensive tonal palette in the study of historic organ registration. Many of this instrument’s advanced and innovative console control features have subsequently appeared as native elements in the specification of alternative related products of various vendors.

The approach to the virtual Scots'

A wide range of diverse documentations/virtualizations have followed upon this heavyweight offering, from highly-prized historic baroque organs to the ambitious work-in-progress that is the Grand Organ of the Esztergom Basilica, where the stupendous 9-12 second (frequency-dependent) reverberation presents a unique challenge in acoustic sample retrieval. As a fortunate repeat visitor to this extraordinary space, it is both impressive and truly moving to experience, from the other side of the world, this unforgettable acoustical effect (not to mention the glorious and distinctive historic pipework). Further to this, not much can, or need be, said. The breathtaking authenticity of the Esztergom Pipe Organ Samples (EGOM) – if I may put it thus – literally speaks for itself.

As work on these various projects has proceeded, background developments at Inspired Acoustics in signal processing and indeed all aspects of virtual organ creation have not only kept pace with rising expectations, but also have suggested choices of prospective instrument that would formerly have been likely regarded, for a variety of reasons, as unfeasible. In other cases, emerging technical difficulties have prompted ingeniously novel engineering solutions, building on an increasingly significant body of original research.

The most recent of the Inspired Acoustics stable of instruments, and one specially relevant in this context, is of course the subject of this series of articles, The Scots’ Church Rieger. In Parts One-Three I have told the story of the church and its history, the arrival of the current organ in 1999, and the unique challenges encountered in gathering the audio sample data upon which the essential viability of the Scots’ Rieger virtual instrument would rest.

The virtual Scots’ Rieger takes shape

Back in Hungary, work on constructing the virtual instrument commenced without delay. Attention would now turn to laborious manual data processing – an enormous quantity of data resident on a small militia of hard disk drives transported with the greatest care from the far side of the world. From these many hundreds of gigabytes of multi-channel audio data, the desired sample material would have to be separated out, sorted into a hierarchy against applicable criteria, and sent on its way to undergo the many stages of necessary refinement for inclusion in a playable instrument. Not to be forgotten were the great many graphic records, obtained in that first night’s work on site, from which references the photo-realistic 3D console would be constructed.

Having encountered the serious environmental noise issues related in Part Three of this blog series, it was important for us to establish to what degree these might potentially compromise the ultimate accuracy of the virtual instrument; once best-existing processing techniques were invoked in a series of test filtering procedures performed on the first-acquired sample data, we judged that a successful outcome could reasonably be expected given the achievement of certain ‘limit’ or minimum conditions. All subsequent sampling was therefore conducted in a manner informed by these critical tests. In terms of Pros and Cons, this meant we knew at least knew what was needed, while resigning to the inevitability of a greatly extended recording phase, making every effort to ensure sufficiently ‘high-quality’ audio would be obtained.

The knowledge that sequential processing tools could be trusted to perform this sonic cleansing, to a high degree of utility, was no small encouragement through those long nights of the sampling effort. Of further encouragement was the promising array of advanced new tools undergoing continuous refinement by Inspired Acoustics’ Technology Development division towards an anticipated breakthrough in pursuit of an all-inclusive interactive signal processing matrix.

In the subsequent months, Inspired Acoustics continued to pursue a broad front of complementary activities: the ongoing building of new relationships with various communities, location recordings for future virtual organs and acoustic documentations, signal processing of works-in-progress and pending releases (including Scots’), the incremental development of the very tools used in this processing, and – as noted above – of an ambitious multi-faceted signal processing and management project still in prototype (offline research) form, awaiting completion and full testing before being pressed into active service.

Once the currently available sophisticated noise reduction and signal processing had been applied judiciously to all sampled ranks of the Scots’ organ, a cumulative effect upon the subjective impression of ‘reality’ when many stops were required to sound in combination, owing to the required treatment of the excessive ambient noise contamination, was clearly observable. This difficulty of gradual ‘pile-up’ or ‘believability drift’ when dealing with large organs, even those residing in comparatively quiet environs – neatly solved a decade back, with NDB’s large-scale combination registration samples – would need, this time, a different – and yet more powerful – manner of assistance. Fortuitously, such assistance was to be forthcoming earlier than expected.

A technical breakthrough: SPIRAL comes online

Almost exactly at the same time the virtual Scots’ was first released the research team achieved an important breakthrough in one of their key project objectives, permitting the prompt assembly of a suite of command line software tools that would ultimately prove entirely central in the fortunes of this ever-challenging virtual organ. The extreme and persistent challenges in resolving this organ’s development to the high standards required and hoped for would now see the software prototype brought online for the first time, in the fashioning of what has come to be styled Scots’ II.

The choice of name for this wide-ranging but deep proprietary technology, SPIRAL, carries far more meaning than first appearances might suggest. The 'word' is in fact an acronym for Synaptic Regeneration Algorithm, a name intended to convey something of the range and depth embodied therein.

So, what exactly is SPIRAL – what does it do, and how does it do it?

The SPIRAL methodology

At this point, I will defer to Dr. Csaba Huszty, whose brainchild this achievement primarily is, as more fitted to describe its technicalities:

“In general, SPIRAL is a methodology with deep analysis and processing techniques involved, specifically engineered to handle pipe organ sounds. It is a series of different processing methods, algorithms and work phases. It is a ‘deep’ process, in that it addresses many issues special to pipe organs that are not found in commercial software, and other tasks that would be prohibitively time-consuming to employ on the many thousands of sample recordings of a typical medium-large organ. Technically speaking, SPIRAL is far more than stationary noise reduction, it includes sound-content-aware adaptive filtering optimized for minimum intrusion, transient and stationary noise reduction, physical measurement-based calibration and signal reconstruction methods, beyond-the-noise adaptive regeneration, polyphonic restoration – everything we basically do in the workflow or processes in handling the organ. It also takes into account perceptual aspects and human hearing. In total, it is list of methods, employed in a given desired, logical order, having been under development now for exactly ten years. It is also highly modular so it can be improved and expanded. At present its modules have very limited graphical interface and are usually run on supercomputers or high-performance systems depending on the nature of data.

“If we ask ourselves why SPIRAL matters, the two main answers would be a major leap in quality and workflow efficiency. Worthy of particular mention are the parameterization and perceptual qualities. Parameterization is an important factor, permitting revisitation or adjustment of any phase without the need of months of additional work. Part of this carefully-optimized process is the capacity for ‘learning’. Importantly, these processes are not self-determinate, but are consciously directed to the optimization of the actual parameter settings themselves. Standard processing of a typical noisy source signal usually introduces artifacts (‘metallic’ sounds) and may significantly reduce the responsiveness of the data. This can be mitigated by applying adaptivity and context-aware intelligence where the algorithm identifies a signal’s ‘content similarly’ much as humans do, but in a much larger range of detail and dimension.

“Some of the noise-related problems are purely of mathematical nature and we can effectively address them with the proper tools in place. The parallel handling and parametrization of linear and nonlinear processing of tens of thousands of recorded audio samples imposes a dimensionality that can be more effectively handled with context-aware methodologies, machine learning and artificial intelligence. This is where SPIRAL has its main merits, in the possibility of providing them.

“Inspired Acoustics’ most recent virtual organ, that of The Scots’ Church Rieger, Melbourne, Australia, provides a demonstration of what can now be expected.”

As a musician, not an engineer, I am obliged to take much of the content of these rather dense sentences largely on faith. I am not obliged, however, to accept an ‘answer’ in the promise of a technology that is at variance with the plain evidence as witnessed by my own ears. There can be no doubt, in direct careful comparison with unassisted, location-recorded audio, that the integrity of the musical speech of the pipe – with SPIRAL processing applied – is fully retained, while noise of all types unrelated to the sounding of the organ is effectively suppressed. For an instrument of the musical pedigree of the Scots’ Rieger, enjoying as it does a complete freedom from compromise – save on account of the cruel and unremitting ambient noise pollution, reported in Part Three of this blog – this impressive achievement is to be hailed as a pivotal breakthrough in signal processing in general, and in virtual organ delivery in particular. Insofar as the precious pipe sounds within the audio signal have been rescued from serious – and otherwise lethal – contamination, the efficacy of the process is clear enough. But it will be in the experience made available for the end user, the player of the completed virtual organ, sounding forth, intact, from silence – an experience not possible in interaction with the real organ in its home, there on that busily noisy inner-city Melbourne intersection – that we will see, fully confirmed, the true magnitude of this achievement.

Conclusion

It is for the many who will come to know, and love, this handsome organ, learning its personality, versatility, and many strengths, that this effort of ours, at times bordering on discouragement, and even despair, has been made, and is considered more than worthwhile. From that first rain-swept Wednesday afternoon in September 2010 to the present, the journey has been indeed long and difficult, ever-beset with new obstacles, each requiring fresh energy and innovative action. It will be the greatest of all compliments to our extended, collaborative labour if, in the rapturous enjoyment of the instrument’s beauty and power, the manifold facts of this tangled tale amount – for the player steeped in the moment of intrepid practice, or fervent performance, or idle improvisation – to not more than of academic interest, paling under the force of the final result. For those of us who tackled directly this most determined foe, in all the vicissitudes of that conquest, the frustrations and anxieties, while not entirely forgotten, have assumed an inconsequentiality akin to the mother’s pains in bringing forth her most dearly beloved child. For that child, now in the hands of a world-wide community of like-minded souls, and for her many new, adoring parents, we wish all success in their future together.