Most readers will be well aware that His Holiness the Pope addressed a huge audience of over one million young people during the World Youth Day event staged the weekend of Aug 20/21 on Marienfeld aerodrome near Köln.
They may also know that audio systems were supplied by leading German rental companies Neumann & Muller, Crystal Sound and Sirius, working s a joint venture group; and that 104 towers of PA were used to relay the messages from His Holiness.
But ask yourself this: how do you control such a huge distributed system? Some of these speaker towers where almost 1,345m from the source point; with a mean air temperature of 22° C, and 50% relative humidity sound travels at 345m/s, that’s 3.9 seconds to the most distant tower. Henning Kaltheuner a Senior Production Planning Specialist from Yamaha volunteered to solve some of the technical issues these calculations identified.
“Neumann & Muller preferred to use Yamaha DME’s because of the level of reliability.” Explained Kaltheuner. “We rarely have a chance to join a real world project of this size during its design and performance and we in fact found some software issues to be improved – and these will already be implemented in next version 1.2. The problem was essentially time. Using a single DME delay component you only get 1.3 seconds of delay maximum; for almost every imaginable purpose this is more than enough, 1.3 seconds is approximately 440 meters, more distance than most distributed systems will ever be. But of course with the most distant speakers 1,345 metres away this was problematic.”
There were other significant considerations to make. “The sound system had to fulfil the German regulations for emergency sound systems, the national standard for this is VDE 0828 part 1. The standard insists that the system must be capable of self interrogation, so that a fault on any device or network connection could immediately be monitored from the control room.”
This was realised by adding a 20kHz pilot tone to any of the 104 outputs of the main matrix. “Outputs from main matrix were send through the Optocore networks (6 net’s in total) into DME24 units and further on in analogue domain from DME24’s to the speaker towers. The analogue signal arriving at each speaker tower was then returned to the specific DME, and then onward back through the Optocore network to the DME main matrix. This return signal was filtered by some sharp 20kHz band-pass filters and indicated on a user screen with 104 meters. During normal operation all these meters showed a constant level, a meter with no signal would have indicated a signal path not working.”
Naturally additional fail-safe systems were required. “Mr. Thorsten Schulze, an independent planner, developed a backup signal distribution system by RF links to the speaker towers. The signal management for automatic fall-over from regular operation to RF backup was also realised in the DME 24 set-up.”
So how did Kaltheuner solve the time delay problem? “The job of the main matrix was to distribute the event program mono mix from a Yamaha PM1D to 104 speaker towers and be simultaneously capable of providing the possibility to address an emergency announcement individually to any of the 104 speaker systems, or specifically to certain groups of these speakers.”
“A main difficulty was to provide enough signal delays for the speaker system. A simple cascade of delays could have covered the requirement, 3 long delays in series would provide 3.9 sec. delay time. This would end in a quite simple hierarchical structure, but the problem is each speaker would require 3 delays – at least for those that far away from stage. DME64 has a limit for delay components, a maximum 32 long delays with 1300msec. Each can be used in one DME64. This limitation is not determined by DSP capacity, DME uses dedicated memory for audio signals per DSP chip, which is totally separate from control memory. This architecture has been demonstrated to be more reliable than a structure in which memory resources are shifted around.”
“If the main matrix had been designed according to the scheme outlined above, up to 7 or 8 DME64 would have been required to build the entire 2 into 104 matrix with all the necessary delays. This would have been quite cumbersome to manage, and also costly.”
“A quite simple reorganisation addresses these capacity problems, and in so doing makes the scheme even more simple. Essentially the concept is this: Delays are not used at the output side of the signal path, but in the input processing instead.”
“Program and speech inputs are processed separately by pre-delays with 1000 msec. of delay each. The fine adjustment for each speaker is done by tapped long delays. The pre-delay time of 1000 m/sec. simplifies the calculation of the total delay time. The tapped delay components have the huge advantage that they just require the audio memory resource of one long delay while providing several outputs. They can only be used before the signals are mixed together, otherwise an individual assignment of signals to speakers wouldn’t be possible anymore.”
“The resultant DME distribution and signal management structure provides the same delay and processing capabilities as the straight one outlined earlier, but it only uses 6 delay components for 10 speakers instead of 30! This is possible because only two signals had to be processed – program and speech. But even with 6 different input signals the same scheme would still save some delays. In the end we did the main matrix with just 3 DME64.”
History now records the event a great success, which is just as it should be.