In my previous article (here), I presented a specific cardioid subwoofer deployment created to deliver an improved result for a concert held in a ballroom-type venue that is wider than it is deep.

This time, we are tasked with the design of a system for one stage of a multi-stage, multi-day electronic music festival. The system must accommodate a wide range of musical styles including a variety of DJs, full bands, world music performances, and acts with acoustic instruments such as sitars and violins.

Our audience area is roughly square, approximately 135 feet per side, with mix position behind, and up a slight incline. Rear rejection will be important, because we need to meet EDM audience expectations for subwoofer SPL in the audience area but not sabotage the acoustic instruments and live microphones on stage (Figure 1, above).

Also, since the stage runs around the clock, the technical crew is working in shifts and camping behind the stage, and we’d like for them to be able to get some rest. So we need efficient rear rejection and high forward output.

We also have some additional restrictions in terms of real estate: the stage deck is only 2 feet high, and there’s a 12-foot-wide thrust in the center for a forward DJ position, so we’re limited both in terms of subwoofer stack height and placement. The stage and set structure have already been designed and built, leaving a 6-foot-high by 4-foot-wide opening in the fascia for PA.

Picking Your Battles

The initial set was designed to accommodate six boxes per side of a dual-10-inch line array. Given the SPL demands of the program material, particularly in the low and low-mid region, and the production goals of the client, a step up in mains power class was advised.

After a series of design iterations, the loudspeaker bay cutouts in the set were enlarged to accommodate six boxes per side of a dual-12-inch cabinet, giving an overall upgrade in SPL capability as well as low-frequency extension and headroom, plus a slight increase to line length as a bonus.

This is an example of knowing which battles to fight – making design changes to the set at such a late stage is unquestionably an imposition on several departments, but we have an understanding of the audience experience the creative team is seeking to provide, and it’s our job to steer them around potholes that would the outcome, or at least make sure their decisions with regard to these matters are informed.

More front-to-back level variance than usual is helpful here to create a “choose your own experience” environment for the attendees; those choosing to stand in the front few rows at an EDM festival have an expectation of deep, tactile subwoofer energy.

However, the overall tonality of the mix shouldn’t be substantially different as we move towards the rear of the audience area (and mix position), so we will use the prediction to examine the amount of low and low-mid drop from front to back (a function of line length) and let the high-frequency energy gently fall off to match with gentle splay angles and minimal shading. This gives the perceptual impression of “smooth” coverage despite the baked-in level variance.

We turn our attention now to the subwoofers; the combination of low deck height and center DJ thrust forces our hand into a left-right configuration: six elements per side, in three stacks of two, with a one-foot gap between stacks; we will need the wraparound paths free for optimal gradient behavior.

Reversing the bottom element in each stack would give a 1:1 ratio between front- and rear-facing cabinets, which is not optimal for rear rejection or for forward-going SPL. With enclosures of this size, a 2:1 ratio is closer to optimal.

And as discussed in my previous article, we don’t want the cancellation to go straight backwards, but rather backwards and sideways, towards the acoustic instruments on stage (more supercardioid than cardioid).

Therefore, we’ll reverse the bottom element of our inner stack and our middle stack, and let both elements of the outer stack face forward. This gives us our 2:1 ratio and sets our vectors up for a cancellation onto the stage. (Simply inverting the innermost stack of elements would oversteer outwards and leave a significant power gap directly in the front of the audience area – a “power valley” if you will.)

Finessing The Process

We could use the manufacturer’s cardioid preset for the rear-facing elements here with some success, but we can wring a few extra dB of rejection by manually finessing the tuning, and also gain some control over the cancellation vector (back at an angle, rather than straight back).

To do this, I placed the measurement microphone at about the quarter mark on the stage (approximately the mid point of that half of the stage, where we want the cancellation to go, labeled as “A” in Figure 2), turned on the four forward-facing elements, and captured a trace.

I then turned on the two rear-facing elements and compared. The level was about 1.5 dB higher than the energy from the forward-facing elements (no doubt, at least in part because of inverse squared differences – the rear-facing elements are, on average, closer than the ones that are front-facing). I decreased the gain of the rear-facing elements to match.

Some practitioners warn against level offset between elements of cardioid arrays, and the resulting potential for “pattern implosion” if some elements hit limiting before others. However, I have found that a little level tweaking can substantially increase the rear rejection, and I did not expect the rig to hit limiting during use. (One could argue that any appreciable limiting is, by nature, an underspec condition, but that’s a discussion I do not wish to dive into presently.)

Consider also that numerous loudspeaker manufacturers’ stock “cardioid” presets include some manner of level offset and / or EQ, and this, as far as the author is aware, hasn’t killed anyone yet.

I also observed a bit of a difference in the high-end rolloff of the rear-facing elements compared to the front-facers. This is expected with physically large enclosures that increasingly occlude higher frequencies at the top of the sub range. Optimal cancellation through the entire sub range relies on matched frequency response (both magnitude and phase) so I bumped the LPF on the rear-facing elements down a bit lower to match the rolloffs.

“But wait!” I hear some readers mentally object. “The phase!” Perhaps unintuitive, but the cabinet dimensions causing the front-facing elements to have decreased HF in the rear also causes phase shift, because it’s a (physical) filter, and therefore causes phase shift as well.

There’s no getting around it. The electronic filtering on the rear-facing elements that creates the match in magnitude response also creates the match in phase response. (Note also that the rigid set fascia creates an acoustic boundary whose effect cannot be predicted properly with typical prediction software, so measurement is required to optimize the cancellation.)

Desired Cancellation

We are now well prepared to increase delay on the rear-facing elements until both the magnitude and phase responses match the front-facing subs, and then invert polarity on the rear-facing ones, creating the desired cancellation.

The “front and center” audience members right in front of the DJ thrust are expecting the most sub-heavy experience, and although the stage is too low for front fills, the mains have wide enough dispersion that we’re not in a HF gap. We are, however, just off the edge of a now-directional subwoofer array, and the impact has lightened a bit – 1.4 milliseconds of delay on the inner-most sub stack bends a bit more energy into the center gap (think of it as a reverse-delay arc, bending coverage inward, not outward) without substantially disrupting the rear-going cancellation.

After a few hours of EDM, I felt confident that the rig would not be pushed to its limits, and took advantage of the extra headroom to apply the filter-based decorrelation described in Part 1 of this series, which sanded the edges off the power alley and the nulls on either side, increasing horizontal uniformity (Figure 3).