The Next Dimension
Roland RSS System
Three-D sound placement from two speakers - that's the secret of Roland's revolutionary Sound Space system. Vic Lennard looks at the technology and the theory.
You've probably already heard the buzz about Roland's "revolutionary" RSS 3D sound system, but what does it profess to do and how?
One of the major developments in the history of sound reproduction was that of stereo. With it came a fresh set of possibilities - not just in the novelty of hearing someone apparently "walk" from one speaker enclosure to the other, nor from the abundance of stereo effects that take advantage of this two-enclosure system, but from such things as the placement of instruments to create the impression of the physical presence of their players. Essentially, stereo was a huge step forward in recreating the sounds of the real world.
The sounds of the real world, however, are more complicated still than those available from a conventional stereo system. The logical line of development was to use more speakers. In the 70s the Hafler system, two amplifier/speaker pairs were used - to the left and right were the left and right stereo images as usual, while in front was the sum of the stereo images and at the rear, the difference.
Consequently, the front speaker emphasised sounds that the two stereo speakers had most in common while the rear speaker provided ambience, or audio information at the edges of the stereo image. While it offered a more natural sound image than stereo from a conventional stereo source, this system had various drawbacks. The main one of these was that it required two amplifiers, two sets of speakers and a large listening area.
Another alternative system uses the IMX Stereo Expander (which can be found on many cheap Tandy graphic equalisers). This uses a mixture of in-phase and out-of-phase signals to give a spacious quality on playback, with the balance between the two phases dictating the depth of the effect. Again any sound source can be used, but the result is quite disappointing. Spacey, yes, but in a directionless manner.
Neither of the above methods are capable of "placing" a specific instrument or part of a sound; instead they affect the entire stereo image. To be able to locate, say, a piano in the stereo image requires the encoding of the original recording in some way.
An alternative approach to that of multispeaker systems is to "encode" sounds to be replayed on a stereo system such that they sound to human ears as if they exist outside the stereo field. While it can be argued that it's scientifically impossible to create a true 3D image from just two loudspeakers, the success of systems which attempt to do this vary from person to person.
Some time ago, Roland embarked upon a line of research and development intended to provide a more realistic image from an electronic piano using a built-in pair of loudspeakers. The culmination of the years of research which followed is the Roland Sound Space system - RSS.
The original stereo recording technique involved a pair of identical microphones crossed at 90 degrees. While this process records the ambience of the recording environment, it's often too imprecise to locate specific instruments. Better localisation is obtained by individually recording each instrument - as in present multitrack practice. The recording then has to be treated with reverberation to recreate the ambience of a "real" acoustic environment. This is necessary because the panning of each instrument only moves it around within the stereo image; the "depth" of that image has to be artificially created.
Over 30 years ago, the idea of accurate binaural recording was investigated. Analysis was made of the way in which the human ear responds to sound by the use of a dummy head with microphones inserted where the ears would be. This facilitates faithful sound recording as long as reproduction is achieved using headphones. One exponent of this technique was Hugo Zucherelli, who used a dummy head fitted not only with microphones for the ears, but also with transducers to pick up the vibrations through the skull's bone formation. "Holophonics", as he called his technology, was used on various early 1980s recordings from Pink Floyd and the Cure. However, there were problems in getting binaural recordings to sound correct on a loudspeaker system, which is why the idea never really caught on.
In part this is due to "transaural effects". These occur only when you listen to sound via loudspeakers; due to the dissipation of sound in the air, your right ear not only gets the signal from the right speaker but also part of the signal from the left speaker. With a binaural recording, this crosstalk destroys the integrity of the image, which is dependent on the correct component of the signal being heard by each ear.
While the use of a dummy head allows reasonable accuracy in the recording of data for the creation of binaural recording, using humans gives more accurate results (due to our composition and the manner in which we absorb sound). By locating microphones in the ear canals it's possible to analyse the results using a Fast Fourier Transform (FFT) and to create a database of characteristics of sound depending on the position of the sound source.
"Some will clearly hear a sound moving behind them, while others will just hear it moving over their shoulder. Either way, the sound certainly moves outside of the confines of the stereo image."
Roland's solution to the transaural problem is to transmit an out-of-phase version of the right-hand signal from the left-hand speaker with a compensating time delay, and a delayed, out-of-phase, right-hand signal from the left. Without delving too far into the technical aspects, the attributes of the binaural database are programmed into a Digital Signal Processor (DSP) chip which, along with the transaural compensation processor, then alters incoming sound to produce the RSS psychoacoustic effect.
One shortcoming of RSS, and indeed any binaural system, is that to receive the full effect of the recording, there is a specific listening position at which the recording will be most graphic in its aural illusion. This position is on the bisection of the lines of the speakers, and such that there is an angle of 60 degrees between the speakers. At this point the time delay for the transaural compensation is optimised. Acceptable results can be heard at most other positions on the bisector, but results are less convincing if the listener is off this line.
The complete system comprises three elements; the RSS-8048 Sound Space Processor, two ADA-8024 analogue-digital-analogue converters and SCC-8004 Sound Space Controller. The RSS-8048 is capable of processing up to four independent signals, each of which then require stereo outputs; consequently, the ADA-8024 has two analogue inputs, one output and two input digital (optical) connections and two pairs of stereo analogue outputs with rear-panel switches for emphasis on/off and sampling frequency selection (44.1kHz/48kHz). They also offer the best digital quality currently available, with 18-bit linear analogue-digital and 20-bit linear digital-analogue conversions. The RSS-8048 is also on the leading edge of current technology with 24-bit signal processing. The simplicity of the front panels of both the processor and converters bely the power within - only input level meters and status LEDs for the sampling rate and emphasis are visible on the ADA-8024, while the RSS-8048 simply has the remote channel number.
Control over the power within the system is exercised through the SCC-8004 remote controller, a most futuristic-looking device which can actually control up to 16 different RSS-8048s by the use of different remote ID numbers.
The way in which RSS is used to facilitate full control over the positioning of the processed signal can be considered as follows: if you imagine that you are seated at the centre of the earth, you need control over two aspects of the sound position. The height of the source above or below you puts the signal onto a circle (effectively latitude) around which it can travel, so you also need to know its position on that circle. The height is termed Elevation while the position within the circle is the Azimuth; there are four sets of rotary controls for these on the controller, one set for each channel.
As rotaries go, these are the business. Perfectly smooth in rotation, they are unlikely to be mechanical, especially bearing in mind their accuracy; I'd be prepared to bet that they are operated by optical vanes. They resemble large alpha wheels - a Roland speciality - around which there are 36 LEDs split between red and green; the colour shows when you have changed the sense of movement. In the case of Elevation this will be from top to bottom, while for Azimuth this shows change from front to rear. The actual position of the wheel is far more accurate; a change of position is transmitted to the processor every three degrees, giving 120 possible positions.
Above each channel is an accurate, 12-segment bar meter showing the operating level in dBs below clipping. While the four channels can each be run independently, it's also possible to work them in stereo pairs with either of the controls acting for both. For instance, you could take a stereo drum image and effect both sides of the signal by rotating one or other of the pair of controls. This is true for either the Azimuth or Elevation rotaries. This stereo facility is brought into operation by pressing the Link button. There is also the facility to have the rotaries acting in different directions; one clockwise, the other anti-clockwise. This is activated by pressing the Mode button while using the Link function.
The only other button of real interest on the remote is the Send button. This transmits the current positions of all eight rotaries via MIDI. While RSS operates in the digital domain, all movements of the rotaries - and hence the position of sounds in the three-dimensional sound field - can be recorded on a MIDI sequencer, which then recreates their movements on playback. Roland have decided to use a feature which is rarely used on MIDI devices; Polyphonic Aftertouch. Few keyboards transmit this, and even fewer sound modules recognise it. Polyphonic Aftertouch allows different values of aftertouch for individual notes, as opposed to Channel Aftertouch, which usually averages the total pressure currently being exerted on notes being played.
The MIDI Note number corresponds to the Azimuth position, with the 120 notes between C-2 and B7 being used, while the pressure value is used for Elevation, with values between 0-119 being used here. The use of 120 values of MIDI information means that the resolution of the rotary controls is kept on playback of movements via MIDI - an important point. The system requires a two-way (handshake) MIDI connection; the return of the MIDI data from the sequencer via playback or a "soft" Thru is used to show the current rotary positions via the LEDs; without the MIDI return connection to the Controller, the LEDs don't light up.
There is a fundamental problem regarding the use of poly aftertouch in this manner; you can't individually record Azimuth and Elevation on separate takes of the sequencer. Both sets of information for one channel have to be recorded via MIDI in one take. And with the high resolution of the rotaries, this is awkward to say the least. It would have been more sensible to have either used a couple of undefined MIDI Controllers or, even better, Non-Registered Parameter Numbers which exist specifically for this kind of situation. Perhaps Roland will change this in the future.
"Movements of the rotaries - and hence the position of sounds in the three-dimensional sound field - can be recorded on a MIDI sequencer, which then recreates their movements on playback."
As all monitoring systems are different, there is a setup procedure to follow when first installing RSS into a system. Using a hidden mode, a continuous sound playing back from the speakers is used to set the Level of Effect and Crosstalk Cancellation Angle. The former adjusts the level of direct signal, while the latter is used to fix the outer limits of the Azimuth. Once completed, the settings are sent to the Processor and remain in memory until changed again. The RSS system is then ready for use.
RSS is certainly mixer-channel hungry, requiring four sends and eight returns. The sends can be via auxiliary outputs or individual channel direct outputs, the input and output levels of the ADA-8024 being set at +4dBm. Having set the returns on the desk, it's then a matter of fishing through the audio signals to find those on which RSS has the required effect. This may be a dramatic result, such as moving a sound around a complete circle with the Azimuth control, moving a solo instrument through a figure of eight using Link and Mode or simply a widening of the stereo image by lifting the Elevation a little and setting the Azimuth to create additional width beyond the speakers.
Different types of sound respond to RSS treatment in different ways. To my ear, those with high-frequency content tend to work least well - hi-hat, cabasa, cymbals and the like can be disappointing. Kerry Hopwood - engineer for Tim Simenon (whose single 'Winter in July' was recently used to premiere RSS on BBC's Tomorrow's World programme) - suggested that it was the more harmonically complex sounds which gave him problems.
Whatever your impression of RSS-treated material, the effects are due to the fact that the treatment changes the tonal colouration of a sound as it moves it around; the greater the high-frequency content, the more I become aware of that tonal change. This type of change does occur in real life, however. If you stretch an arm above your head and click your fingers and then slowly bring your arm down in an arc continuing to make the same sound as you go, you'll find that the tone of the click changes. Roland have attempted to duplicate this effect, but the tonal changes are currently too severe for my tastes. This is an area which is being investigated; if Roland can get this right, then RSS may well be able to manipulate all types of sound with equal success. However, it's worth bearing in mind that a severe change in tone only occurs when RSS is being used to its extreme; the same can be said of aural exciters and other enhancers, although I'm not implying that such devices could synthesise the RSS effect.
RSS seems to be at its most effective when operating on sounds free of reverb or other effects. This could be because the reverb effect is restricted to the normal stereo image and so attempts to pull the sound back into the stereo image. One interesting effect is to put the reverb from a sound through two channels of RSS, one side moved to the furthest left while the other is just pulled slightly to the right. This asymmetrical placing is quite off-putting and certainly enhances the stereo image effectively. A similar effect on a piano or other stereo-panned instrument is equally dramatic.
Slowly-moving sounds benefit most from RSS; you can clearly visualise a position associated with the movement. However, the precise-positions of movement are very much down to individual interpretation. Some will clearly hear a sound moving behind them, while others will just hear it moving over their shoulder. Either way, the sound certainly moves outside of the confines of the stereo image. When I first heard a demonstration of RSS, I heard a buzzing sound coming some metres from behind my head - and then realised that my back was against a brick wall...
The MIDI side of RSS is in many ways a bonus. The ability to be able to concentrate on a particular part of a mix, record it to a sequencer and then continue, safe in the knowledge that the movements just made will not need to be recreated at mixdown, is reassuring to say the least.
Finally, RSS appears to playback well via a monophonic system, and it is this aspect which augurs well for its future TV and radio usage.
The cost of the complete RSS system hasn't been mentioned. It's £25,300 (if you need to ask whether this includes VAT, you can't afford the system; actually the price is exclusive of VAT). I don't think many studios will be adding RSS to their armoury, but a day's hire for RSS to be used on a final mix isn't out of the question. The current system is very easy to use; you can be connected and up and running within an hour.
You should soon be able to decide how well the system works for yourself - if you haven't already. There are now several recordings available which have used RSS and these include Bomb the Bass' 'Winter In July'. The b-side of the follow-up, 'The Air You Breathe', is called 'Liquid Metal' and RSS is heavily used on this too. 'Spiritual High' by Moodswings featuring Chrissie Hynde, uses RSS, while on the album side there is Cooltide by John Martyn, Flashpoint by the Rolling Stones (their producer, Chris Kimsey, was one of the first people to use the 1990 prototype) and the forthcoming Swing Out Sister album. Perhaps the best opportunity to check out the musical uses of RSS for yourself is on the sample CD Pascal Gabriel's Dance Samples (reviewed MT, December '91), which has a section of treated sounds which you can sample and try in the context of your own songs.
With a company like Roland behind RSS, it's likely to have a healthy future. With stereo TV receivers quickly becoming a domestic standard, film soundtracks may be one of the beneficiaries of RSS. Three-D radio adverts are another realistic possibility for the near future - the next time you think that the rear speakers in your car are performing particularly well, just remember that you haven't actually got any.
Gear in this article:
Feature by Vic Lennard
mu:zines is the result of thousands of hours of effort, and will require many thousands more going forward to reach our goals of getting all this content online.
If you value this resource, you can support this project - it really helps!