12. Sonic Solution

1990-1992

As early as August 1988, Brian Hodgson had commented on how digital signal processing (DSP) could usefully enhance existing audio material, especially in the area of noise removal. By the summer of 1990, such processing was contemplated as an extension to the main Radiophonic operation. The two main systems were NoNoise, from Sonic Solutions, and CEDAR. Unfortunately, both products were expensive and presented in a nebulous way, giving the impression that they were a bit of a ‘con’.

Another system, known as DAWN (Digital Audio Workstation Nucleus) was tested early in 1991, but didn’t impress and lacked digital inputs or outputs. In comparison, the NoNoise demonstration was faultless. As well as disposing of interference or other unwanted sounds from existing recordings, it could be used to transfer the processed material onto CD-R. Hence, this system was chosen for the new Studio X, which was to be on the site of the original Studio B, formerly known as Room 38.

The Sonic Hardware

Initially, the Sonic Solutions system consisted of two NuBus cards installed in a Mac IIfx computer, each of which contained special 27 MHz processors. At first, it didn’t work in ‘real time’, which meant processing took at least twice the duration of the material. Later versions of the software and hardware did run in real time, but required three separate ‘passes’ to remove crackles, noise and hiss. Fortunately, processing could be done in the ‘background’, freeing the computer for other tasks.

The original Mac IIfx was eventually replaced by a Quadra 900, and the system expanded to consist of three updated cards. The first two of these cards provided a ‘virtual’ four-channel mixing desk, complete with faders, pan pots and equalisation. As the system developed, these cards were linked via a Small Computer System Interface (SCSI) to 2.8 GB of hard disk storage and a Sony CD Encoder. From this encoder, audio data passed via optical fibre connections to five Sony CD Writer units. This allowed processed material to be recorded on up to five audio CD-R discs at once.

The third card of the updated system was devoted entirely to noise removal and had a pair of stereo digital audio inputs and outputs. These fibre-optic connections were plugged to an interface box that provided digital access via AES/EBU and SDIF2. Although the AES/EBU interface was the ‘industry standard’, the older Sony Digital Interface 2 (SDIF2) format was still useful for older devices.

Operation

Source material was loaded into the Sonic system, usually from DAT, to create a ‘soundfile’ on the hard disk. The effect of different kinds of processing was initially checked using the on-screen mixer. Having chosen the best setting, the system could then process the material in the background.

The final soundfile could be edited on the Mac screen as a ‘foreground’ task, with sound waveforms shown graphically on the screen. The user could ‘zoom in’ for more detail or move out for a general view. Sections of sound could be replaced with ‘black’ silence or the original material could be joined up around the ‘gap’. Also, portions of sound could be repeated or swapped between the stereo tracks.

A Table of Contents (TOC) for recording a CD-R could be created, either manually, using ‘flags’ on the graphical display, or automatically by setting a ‘silence’ threshold and duration. The final soundfile could then be ‘dumped’ onto tape, DAT, CD-R or any combination of these formats.

The Basic System

By the end of 1990, the design of an elongated version of the ‘circular’ console had been finalised. As shown in the diagram below, this included circular elements similar to those in the original console design, but with additional straight sections that contained extra cupboard space.

Early in 1991, the engineers had moved out the old equipment, relocated the RF feed from the original Studio A and installed a new technical earth. Meanwhile, Jeremy Quinn measured up the area for the console whilst the BBC’s Equipment Department began work on nine mini-racks.

As well as the Sonic system, the installation contained three Yamaha DMP7D mixers in a digital ‘cascade’ and assigned respectively to general sound sources, synthesisers and effects devices. Each DMP7D was fed via an AD8X 20-bit 8-channel analogue to digital converter, whilst extra effects were supplied by a Yamaha SPX1000 effects processor digitally connected to each mixer. Connections were controlled by four Akai DP3200 matrixes, accommodating up to sixteen effects devices. The digital output of the last DMP7D was connected to a Yamaha DA202 digital to analogue converter.

Initially, a single computer was considered for both NoNoise and MIDI effects work. However, the spectre of a ‘MIDI crash’ bringing down the NoNoise system led to the use of two computers. Unfortunately, a separate keyboard and mouse had to be used for each machine, since the Apple Desktop Bus (ADB) would fail if switched or disrupted, and the keyboard lead of the NoNoise machine had to be extended. The ‘MIDI’ computer was provided with a Mark of the Unicorn MIDI Time Piece, connected to a Digisound MIDI to CV Interface for the Roland 100M synthesiser.

Although the installation also included a Roland S50 sampler, its buttons weren’t accessible, since they were beneath the upper level of the console. This problem was solved by locating the ‘MIDI’ computer and its keyboard above to the S50 and installing software that could directly control the device.

The initial NoNoise system arrived in the Spring of 1991, as did the pre-wired studio cables. The console also arrived, complete with a spray-on melamine finish that looked very tough. A set of mini-racks was provided by Equipment Department (ED) but, due to inaccurate construction, none of the constituent parts actually fitted together. ED later supplied new tops for the racks but these were then too big! Pete Scott, then attached to the engineering team, was responsible for creating special ‘combiner boxes’ for the mixer auxiliary outputs, as well as a special power supply unit (PSU).

A new matrix control application called QuickPatch was developed by Tony Morson. This replaced one vital role of Cue Card in the move to a new generation of computers and MIDI interfaces.

Other Elements

This studio had an ‘off the shelf’ monitoring system, the AMU-8S from Media Products. As Murphy’s Law dictates, this failed several times and had to be returned each time to the manufacturer. Eventually, it worked perfectly well, although when, in October 1992, Ray Riley suggested that such a unit be used in Studio B, Peter Howell claimed it was a‘Chad Valley’ device. To be on the safe side, Pete Scott made a simple control box that could replace the unit in case of failure. In Studio X, the AMU-8S was wired via a multiway connector to a large plastic box. This had jack sockets, as well as cables wired to Amphenol connectors that plugged into a DP3200 matrix. Unlike earlier installations, all the other cables connected to the matrixes were fitted with an Amphenol connector at one end and jack plugs at the other. This obviated the need for an adaptor panel at the back of each DP3200.

For sound effects work, several of the older generation of Roland effects units, including the SDD320 Dimension D, were retained, as well as the Roland 100M analogue synthesiser and other devices. New equipment included a Roland SN550 Digital Noise Eliminator, a Precision Power Phase Chaser, an Audioscope spectrum analyser and a Koch CD Tester.

The Roland SN550 processed material in ‘real time’, releasing the Sonic system for more arduous tasks. The Phase Chaser eliminated phase errors that shifted over a period of time, as often created whilst copying material from a tape machine that had an azimuth error. The Philips Compact Cassette format, with its slow tape and relatively crude mechanism, was particularly prone to this problem.

The Audioscope could be used to find any frequencies in a recording that were attenuated or emphasised, either before or after processing with the Sonic system. This device provided an RGB output suitable for a Philips CM8833 colour monitor. Unfortunately, the 9-way D connector on the Audioscope lacked the necessary +5 volt feed to switch the monitor into RGB mode and therefore had to be modified. The author also had to extend the monitor connections to the Audioscope and to the Roland S50 sampler by using 8-way inline couplers. As with the DMP7 ‘cascade’ cables, these had to be created from the shells of a pair of sockets that were joined and encased in heat-shrinkable sleeving.

The Koch CD Tester allowed a recorded CD to be checked for errors and other information. This device was connected to a standard Centronics printer, allowing the results to be kept on file.

Stereo recording was accommodated by Sony PCM2500 and PCM7030 DAT machines. Sadly, the PCM2500 allocated to this area had been modified to accept an external clock signal. The presence of this clock, usually running at 44.1 kHz, unfortunately prevented recording at 48 kHz. So this machine had to be replaced with a ‘normal’ machine that extracted its clock from the incoming AES/EBU audio data. The PCM7030 was useful in that it worked with SMPTE timecode and could be directly controlled by the Sonic system, thus accommodating automated loading and dumping of sounds.

Two Studer A812 quarter-inch tape machines, complete with ‘centre-track’ timecode facility, were included in the installation. For operational convenience, the decks of these recorders could be tilted at an angle. One composer said he was ‘fuming inside’ about the allocation of these machines to this area, finally referring to the author and the assembled company as ‘po-faced gits’.

In June of 1991, the studio received a special ‘78’ player, as developed by Technical Services. This machine could be used to play old records without causing them any damage. It consisted of a modified EMT turntable, complete with a Revox ‘linear tracking’ arm and a Shure V15 pickup cartridge. The total development cost of this machine had come to around £100,000. Unfortunately, it couldn’t play modern ‘33’ or ‘45’ disks, so a ‘standard’ gramophone had to be provided as well, complete with a switch box to select between the two turntables. The extra styli and cartridges for the ‘78’ player cost almost £1,000 and were eventually installed in a special rack for safekeeping.

Fine Tuning

As in previous ‘circular’ studios, a considerable amount of experimentation was inevitable. For example, in May of 1991, Ray Riley investigated the control of an Akai DP3200 matrix via a MOTU MIDI Time Piece. This worked, although ‘dummy’ data bytes were required to convince the MTP that the instructions were valid MIDI data. In addition, only three matrixes could be controlled over a MIDI circuit. Hence two circuits had to be used, each connected to a pair of matrixes.

In June, the digital links to the NoNoise system were tested. Meanwhile, both the MTP and Video Time Piece were connected to the ‘MIDI’ Macintosh. However, Apple’s MIDI Manager software didn’t like AppleTalk, since the latter required the Mac’s printer port. To get round this, the printer was connected directly to the printer port whilst the MTP was only plugged into the modem port. But this didn’t work either, so the MTP was reconnected to both ports. The printer port was then used purely for controlling the matrixes whilst the modem port was used for other MIDI data. To prevent AppleTalk from monopolising the printer port, an older version of MIDI Manager was installed. Everything then worked, apart from the printer itself. Finally, Tony Morson tested his application for controlling the matrixes, the links between the four matrixes giving him a considerable headache.

Complications with timecode revolved around the recording machines. For example, timecode fed into a Studer A812 would appear at its output if its ‘input’ button was pressed, whilst the PCM7030 DAT always copied incoming timecode to its output. Both peculiarities could send the studio’s timecode system into a ‘loop’. Worse still, if the timecode selector at the rear of the PCM7030 was set to ‘Int’ it would ignore any incoming timecode and generate its own from the ‘absolute time’ on the tape!

By the summer of 1991, the five CD Writers and the Encoder had arrived. Unfortunately, the first version of the software only allowed you to ‘burn’ five CDs or none at all! By now, Dick was working with NoNoise in earnest, although inevitable problems were encountered. For example, the Sonic system would sometimes fail late in the day, apparently due to overheating of the cards.

It was also discovered that the NoNoise equipment didn’t like long SCSI leads, although these were well within the limits set by the SCSI standard. This meant that all the disk drives then had to be relocated. Dick also had to remember to reformat disks every eight weeks to prevent any fragmentation of data. And the rather inconsistent behaviour of the ‘flags’ sent over different types of digital interfaces made it necessary to switch off pre-emphasis throughout the studio.

And in the autumn, the studio was still being updated. For example, the MIDI Time Piece, a very silly device that still wouldn’t remember its settings, had to be relocated. And an extra 1.2 GB hard disk drive had to be added to the Sonic system. Dick, having created yet another dud CD, decided to attach to it a battery-powered clock mechanism. In cahoots with Malcolm Clarke, he developed the idea into a ‘back room’ industry, although his products weren’t to everyone’s taste!

Despite all the changes, the new Studio X was opened on Friday, 18 October 1991, accompanied by pink champagne and a yellow cake made to look like a CD. A good time was had by all.

After the Event

Even after its official opening, Studio X continued to evolve. In the spring of 1992, extra memory was fitted in the computer for the updated NoNoise software. Unfortunately, 32-bit addressing hadn’t been enabled, causing NoNoise to run out of memory and to wreck five more CDs. And a later version of the Sonic software was so bad that Dick reverted to an earlier version, which wasn’t easy, since the original registration code hadn’t been written down. Dick also once recorded a set of CDs at a sampling rate of 48 kHz, instead of the standard 44.1 kHz. They played, but sounded rather slow!

The ‘Sonic’ Mac IIfx computer was soon replaced by a Quadra 900. Unfortunately, this had five NuBus slots, unlike the six in the IIfx. This meant that the three Sonic cards would have been closer, increasing the risk of overheating. To avoid this, the NuBus video card for the Calcomp monitor was removed and the display was persuaded to work from the built-in video output of the Quadra 900.

The digital interconnections to the Sonic system were also changed. For example, the Audio & Design ProBox, a device that converted digital audio between AES and SDIF2 standards, had to be relocated to avoid the usual timing problems associated with digital audio.

By the autumn of 1992, a new version of the Sonic Solutions hardware had arrived. Sadly, fitting the cards caused the computer to fail, giving the dreaded ‘Death’ sounds at startup. Another machine was rented and was eventually purchased as a spare. The new system also included a Universal Clock Module (UCM) that allowed NoNoise to accept a ‘site’ reference clock, such as composite video, SMPTE ‘time of day’ timecode or incoming AES/EBU digital audio. Sadly, it couldn’t directly read or write timecode, although SMPTE information could be accepted via the UCM’s serial port. Worse still, the UCM only accepted its master clock via digital audio inputs 3 and 4. This wasn’t very convenient since, at this time, Studio X only had optical interface boxes for channels 1 and 2!

Simultaneously, a small LocalTalk network was created, allowing the resident laser printer to be used with the ‘MIDI’ Mac IIx or the ‘Sonic’ Quadra 900. By now, the Mac IIx was fitted with an Opcode Studio 5 MIDI interface. Since this had a dislike for AppleTalk, it could only be connected to the computer’s modem port whilst the LocalTalk network could be wired to the printer port. Unfortunately, AppleTalk still had to be switched off to use Apple’s MIDI Manager.

Level Headed

Early in the summer of 1992, a customer complained, with justification, that the Sonic system was producing low-level CDs from normal analogue source material. Unlike BBC recordings, where a ‘headroom’ of around 20 dB was provided, the sounds on CD had to be at the highest possible levels. This meant that the bargraph on the studio’s main AD8X A to D converter had to be illuminated to just below ‘clipping’ at maximum levels, with the NoNoise trimmed back by ‘1’ on its on-screen display.

The levels fed into the Sonic system were increased by fitting a 12 dB attenuator in the monitoring circuit connected to the studio’s main output. This meant that when the studio’s meter showed PPM 6 (+8 dB) the actual level received by NoNoise was +20 dB, just on the verge of clipping. As a result, a lineup level of PPM 4 (0 dB) corresponded to a feed of +12 dB into the Sonic system.

With this arrangement, a limiter was needed or levels had to be monitored very carefully. A CD test recording of a section of Prokofiev’s music sounded loud enough on a domestic CD player, generating an output from the player that peaked to +5 dB, where +7 dB corresponded to clipping.

An Audio & Design Limiter, modified to have a threshold 5 dB above the normal +5 to +15 dB limit, was subsequently installed, as well as a sample rate converter. A digital fader, also from Audio & Design, was later rejected since it removed the DAT ID codes used by the system. Hence, the studio’s analogue main fader was retained and was fitted with a switch to provide a mono studio output.

Next

©Ray White 2001.