Saturday In The Park

This time, the task is to design a sound system to cover the 19-acre Gas Works Park in Seattle WA for an open-to-the public fireworks event. The park is on the National Register of Historic Places, the last remaining coal gasification plant in the United States, converted into a public park in 1975. The plant’s long-retired, massive steel tanks and pipe structures, relics of early 20th century industrial engineering, form the architectural centerpiece of the park. The sound system will radiate outward from the architectural fence line to cover the audience areas surrounding the structure.

Our event will require equipment spread over a large area, and due to the terrain, most of the gear will need to be pushed into place by hand, so we want to be efficient with our equipment choices to minimize weight and crew exertion.

The majority of the stated 50,000 person attendance capacity will occupy the lawn to the west of the plant structure, where the terrain rises to a hilltop about 250 feet away. On the east side of the plant, a paved area hosts food vendors, and beyond, more seating area on the lawn sloping down to the waterline about 270 feet out.

To the south, we have a relatively flat VIP seating area overlooking the lake, and additional VIP seating along the lakeshore to the southwest. The TV broadcast stage in between the two VIP seating zones needs to be out of coverage so as not to interfere with the broadcast. In addition, the entire north lawn will remain uncovered by our system, to accommodate a small mobile stage with band performances throughout the day. (As an aside, we were asked to accept a board feed from that stage, as well as send a feed from our system to their mixer. Care must be taken at both ends of the signal chain to ensure that neither operator can accidentally create a giant feedback loop. Fellow systems engineers are invited to consider how they might accomplish such routing.)

We will apply the design strategy outlined in my book – use the largest / most powerful system to cover the largest audience area, then use fill systems to fill in the gaps. Rinse and repeat. Here, our mains are 32 elements of a small-format line array cabinet. The West and East coverage areas and throw distances are about equal, so we will use half the boxes for the west side and other half for the east side, allocating 8 elements for each of the four main PA tower locations – NE, SE, NW, and SW. (If one side required a significantly further throw than the other, we would allocate the inventory accordingly.)

To make the most of our set up and load in time, and help ensure the vendor doesn’t pack too much or too little equipment on the truck, we want to have our design as solidified as possible before we arrive on site. Google Earth Pro satellite info can give us a bird’s eye view of the area, and even preliminary elevation data that we can use to translate into rudimentary audience planes. We can start by marking out on-axis “landing strip” elevation data for each of the four PA tower locations (FIGURE 1). The two West arrays will throw uphill, while the two East arrays are throwing slightly downhill. How high should we trim? The 25’ maximum height of the lifts makes the decision for us. A pair of dual-18” subwoofers at the foot of each tower will supply the LF extension.

FIGURE 1 - Satellite view, with red lines indicating the “landing strip” coverage of each array tower, viewed from the SW. Elevation data is visible in the bottom pane. Google Earth Pro

Here we must pause to examine a shortcoming of the simple on-axis “landing strip” elevation method for determining splays. Our array has a 100° horizontal dispersion, and the summit of the hill, which is about 15 vertical feet higher than the on-axis point, lies off axis horizontally. If we are only considering the ground elevation of the on-axis landing strip, we will miss covering the summit. We need to overshoot the on-axis geometry to make sure the listeners on the higher ground off axis are properly covered.

FIGURE 2 - Imagery from the POV at the base of the NW PA tower, looking due West. Red line shows the on-axis landing strip, and green indicates the additional vertical coverage needed to ensure covering the listeners atop Kite Hill. Google Earth Pro

FIGURE 2 shows the satellite imagery from the perspective of standing at the base of the NW PA tower, and the red line indicates the on-axis landing strip elevation that we might traditionally shoot with our rangefinder and input into prediction software. The green line shows the additional vertical height that we must achieve with our coverage in order to properly cover the people at the top of the hill (who arrive a full twelve hours early to claim this “perfect viewing spot” for the fireworks show). When only considering the on-axis landing strip splays, the coverage seems to be excessively overshot, but considering the entire 100° horizontal dispersion of the cabinet, we see it’s there for a good reason.

A similar issue can arise with traditional side hang / outfill geometry in arenas and stadiums, when the array is projecting onto the audience plane at a non-orthogonal angle. FIGURE 3 is a generic stadium side seating plane. Normally we would overshoot by a box or two to prevent the HF rolloff occurring in the top rows, but I’ve aimed the top element precisely into the top seating row for clarity here.  The prediction reveals the severe under coverage situation at the edge of the horn nearest the array. To rectify this, we need more overshoot than would otherwise seem necessary. Using 3D prediction software is the best way to protect against this type of under-coverage error, as it is not readily apparent in 2D landing strip predictions whenever the source is not impacting the listening plane head-on (what physics-inclined individuals refer to as a “normal” angle).

FIGURE 3 - aiming the top box into the last row on-axis results in severe undercoverage for listeners at the top left corner of the audience plane due to the non-orthogonal angle of incidence. MAPP 3D.

Since we have almost 200 linear feet between the NW and SW towers, we can expect extensive center gap along the fence line between the arrays, although they converge and overlap at distance, and prediction confirms this (purple shaded area in FIGURE 4). Since we don’t need to go very far, we can fill this gap with two point source loudspeakers on sticks located at the two dogbone bends in the fence line (blue icons in FIGURE 4). Since a secondary fence around the architecture in the center provides a buffer zone area open only to production crew, we don’t need to worry about people in the immediate nearfield getting blasted, thus these auxiliary fills are only tasked with covering from approximately 12’ to about 50’, at which point their coverage feathers into the energy from the main arrays.

FIGURE 4 - EASE Focus 3 prediction showing the coverage gap between the west side arrays (purple) and the proposed delay locations (blue icons)

The two PA towers on the east half of the area are essentially mirror images of their western counterparts, albeit with different splay since the audience area slopes downwards instead of up, and without the need for auxiliary fills, since the first fifty feet or so is occupied by food vendors. Likewise, the south and southwest VIP areas are relatively small and easily covered by additional point source elements on tall stands operating at low SPL, which minimizes energy spill into the TV broadcast area at the SW corner of the fence line surrounding the old plant.

Once we are on site, we use our rangefinder to confirm the accuracy of the elevations we pulled from the satellite data and make any needed adjustments to the splays for each array tower.

The decentralized nature of the system requires us to give the alignment process some thought. The four principles of alignment outlined in my book can guide our order of operations here when coupled with the practical considerations: we have a front-end DSP located at the SW corner control position, which is distributing signal to each of the four PA towers, plus additional feeds for west fence line fills, south VIP and shore line VIP, plus a variety of utility feeds to the broadcast truck and the band stage on the north lawn. We also have per-box processing for each of the arrays and subwoofers, but not the networking infrastructure to control them remotely (although that’s high on the list of improvements for next year’s iteration). So, we will visit each tower with our laptop and measurement microphone, verify the array and subs, push our HF air compensation filters from prediction as a starting point, and then tune the array and subs to achieve the desired tonal target curve for this event. Since all four towers are the same box count with a similar amount of curvature, the EQ for each array ends up being very similar. Once we have visited all four towers, we will turn our attention to timing, and then setting level and timing for the fence line delays to fill the gaps. All of this deals with the interaction between sources, so we can work using the front-end DSP without having to reconnect our computer to the individual tower systems.

Since the east-side towers are basically twins, both firing forward into similar audience geometry, and we have no fills to deal with, we don’t have any real alignment decisions to make. The west side is more complicated – the NW and SW towers are aimed on different vectors, and covering terrain of different heights with different splays – they are asymmetrical.

From up on the hill, when we get to the south edge of coverage for the NW tower, we are already hearing quite a bit of energy from the more distant SW tower. We have a rather large equal level region (what I refer to as the “seam”) from very displaced sources, and so much abandon any thoughts of finding some perfect delay time to the tenth of a millisecond, as comfortable as that notion may be. No, the seam itself is large enough that there is a decent amount of “time slip” across it. We will solve this the old-fashioned way – grab a radio and fire up our favorite tuning song, calling delay values back to a colleague operating the DSP at mix position. Standing in the area where I found the time offset most disruptive, we increased delay on the nearer (NW) tower until I found it least objectionable as possible, then walked the entire transition a bit more to confirm. With source displacements this large, single-millisecond changes don’t clear things up the way they might in smaller-scale situations, so I recommend bumping up or down in 5 ms or 10 ms increments until you hear it clear up, keeping in mind the goal is to do the best we can for the most listeners.

With that sorted, we go down and listen to the fence line fills from about half their coverage depth (approximately 25 feet), call for level adjustment until it matches what we were hearing in the coverage of the mains, then back up until we creep into the edge of the array coverage, and call delay times for the point sources until they snap in. Smaller seam, so timing can be more precise here. Rinse and repeat for the south shore fill, which picks up where the SW array drops out. The south VIP area fills have no seams with any other sub-system – they’re on their own out there, with no other arrivals to hear – so we can just set them to an appropriate level and we’re off to the races. Using this method, the entire alignment process took less than half an hour despite the large amount of ground to cover.

Although the approach here might seem a bit atypical to regular readers, the conceptual workflow is the same as usual – start with the large main system, set the tonality and shade for uniformity, then work your way down, bringing on the supporting system to fill the gaps and setting the timing relationships at the seams as we go.  

 The author wishes to thank Audio Engineers Northwest and Own The Night Productions for their collaboration on this project.

Southwest PA hang and the plant’s Semet-Solvay gas generator towers and primary and secondary scrubber towers

As the sun sets, the audience settles in on Kite Hill in anticipation of the fireworks show.

With the sound system and the gas plant silhouetted, the show begins.