An era came to an end for me recently when, after 14 years of continuous operation, the audio routing matrix I designed and made in Birmingham for Capital Radio in London was finally switched off for good. The picture shows Global Radio engineer Hirjii doing the deed.

A radio station usually consists of a number of studios which take feeds from several outside sources (outside broadcasts, other studios in the building, phone lines etc) mixes them together with local audio (playout systems, microphones) and outputs the result to one or more transmitters. In the early days of commercial radio this was a fairly simple affair as there were usually only two studios which flip flopped between being on-air, few outside sources, and usually only one transmitter feed.

It all got a bit more complicated in about 1988 when the IBA (Independent Broadcasting Authority) ruled that radio stations which up to then had been simulcasting on AM and FM should provide distinct programming services on the two frequencies or risk losing one of them. This meant additional studios and additional transmission outputs, but it was more complicated than that. Studios had to be able to feed AM or FM or both. If feeding both transmitters jingles and commercials needed to be separate - this meant that transmitter unique audio needed to be mixed in downstream of the studio - before the transmitter. At the same time stations started to merge / take each other over so for example a program made in Birmingham might be fed with different ads and jingles to Birmingham and Coventry. If that wasn't enough, this was in the days of the two and a half second profanity delay. It had to be possible to switch a delay unit into the transmission chain so that unwanted profanities (occasionally from phone in callers) could be deleted before transmission. This all complicated monitoring - how could the presenter hear the jingles and ads if they were being mixed in after the studio?

Click on read more for ..erm more

It was at this time, when I was working as an engineer at BRMB in Birmingham and had started writing bits of code in Basic using a Nascom computer that I had built from a kit (from Birmingham based Lucas Logic)  - I still have that machine. I remember taking it into work and demonstrating to my boss at the time, Dave Wood, and the stations MD, David Pinnell, that if I connected some switches to it I could flick switches and get a simulation of circuits being switched on screen. The bosses liked it and so the Kaye Switcher (named in memory of my Mum who saw computers as the coming thing and was pleased when I started messing with them) was conceived. The Kaye Switcher concept used an Omron self latching relay as the switching element (none of your headroom limited digital nonsense here). Not only did this provide a very transparent audio path, but it had the advantage that once you'd pulsed it to either set or reset it, it would stay there consuming no power whatsoever until you told it to switch the other way. Even more, if the whole matrix lost power all the relays would stay put and routes would all be exactly where you left them when the power came back up.

The relays were built into Eurocard pcb's I designed using Cadstar pcb design software. Each pcb carried 2 x 16 way stereo / mon swtich arrays and the cards could be stacked via jumper cables up to about 256 way (limited only by cross-talk). The dual layer pcb's were manufactured by Tates Circuits just down the road from us in Aston, Birmingham company and assembled and tested mostly by me assisted by engineering colleagues at BRMB. Later variants of the boards were assembled by another Birmingham company (the name of which at the moment escapes me) and I built an automatic test rig using a Lindos audio analyser.

The whole lot as you might guess was computer controlled. Remember this was pre-Windows. Cadstar ran on a pc but was DOS based. The CPU for the system was a single board micro computer from Cambridge company Arcom. The development environment for the early systems ran under CP/M and used Arcom's AB88 compiled basic. Later versions of the software were written in C using Borland C and compiled into EPROM. The CPU kept it's configuration and status in an array held in battery backed RAM. Again this made the system resilient to power failure - you could switch the CPU off and back on again and it would know exactly where it had left off.

The final innovation in this system was the use of a touch terminal for the user interface. Touch screens are in nearly everyone's pocket these days but in 1989 the use of a touch screen was quite controversial - would the dj's cope with it? These TCS units were self contained computer terminals with a resistive touch screen designed for industrial use. They looked like a standard 14" monitor. The TCS units communicated with the CPU via RS422 which polled up to 8 TCS ports. Although expensive (each one cost about £2k) these units were very reliable the only problem being burn-in on the CRT screens. Incidentally, I did a cost benefit analysis of the £2k cost of these screens and it was still way cheaper than the cost of wiring up the equivilent number of physical switches - even if you could have got the same number of switches into the same space. 

Configuration for the whole system was via an ini style configuration file (later versions used the same syntax as Windows ini files) uploaded via a serial port initailly via a passive terminal and later from a pc. Among other things this defined the TCS screen layouts, functions assigned to buttons, and the labels attached to inputs and outputs of ths matrix.

In all four Kaye switcher systems were built with improvements at each iteration. The MK1 went into BRMB's Aston studios in Birmingham, MK 2 went into Mercia in Coventry. Here the concept of sustaining services was introduced whereby the last presenter of the day in Coventry could put the transmitters into "Sustain" where they would take a feed up the BT line from Birmingham. The same line was used to relay AA travel news and commercials from commercial production to Coventry during the day. This process was on a timer - so when the AA weren't doing the travel news the line was made available to the commercial traffic department to pipe ads to Coventry.

MK3 was built for Capital when they moved from Euston Tower to Leicester Square. Capital's requirements were beyond the limits of the original basic compiler so the whole lot was completely re-written in C taking up a massive 48k of EPROM space! Originally built for 4 studios this system was quickly expanded to 8 studios and about 8 transmitters and at one time was controlling Capital FM, Capital Gold, XFM, Choice FM, Digital News network and a couple of other feeds the exact nature of which I can't remember.

The onset of digital desks with central racks of DSP spelt the beginning of the end for the Kaye Switcher which, although very configurable was not quite as flexible and, not being pc or ip based, not as easy to interface to. It was however very reliable, which is why it survived so long, and of course used a lot less power!

Average user rating