The Hand You’re Dealt

Theater retrofit. Actually a retrofit of a retrofit of a retrofit. Concert hall designed for acoustic transmission, hideous six-horned center cluster added in the 80s, replaced a decade ago with a modern L/R “line array” system and a center sub, and today’s job is to install a pair of flown fills to handle the coverage gap – the first four rows of the outer seating sections are too far under the mains to get HF coverage. Budget and time were both very limited, so the goal is to create as much of a measurable improvement as we can within the (significant) constraints.

Here are measurements of the Left main array ONAX of each box, averaged together into a gold trace that will be our reference to match for the fill response (magnitude vertically offset for clarity here). This system is used primarily for speed reinforcement, reflected in the very flat tuning curve above 100 Hz. Take a mental note of the phase responses for comparison to the next image.

So now we have a target. We start by taking some measurements throughout the first few rows to see how large of an area we need to fill, and how bad HF rolloff is.

Compared to GOLD, we can see how all these positions have severe HF rolloff. The phase response over this area looks like a well-decorated christmas tree and coherence is very low. All three symptoms of being vertically off axis from a multi-box array. No matter. I took many traces labeled with seat numbers, which enabled me to determine the extent of coverage that we needed from the fill, and I’m showing the worst offenders here. Note that these traces have been level normalized (shift-click in Smaart) to 1 kHz, which removes the level variance from the equation and overlays the slopes. Here we’re interested in spectral variance more than level variance.

In a perfect world this would dictate the choice of loudspeaker based on directivity, SPL requirement and matching with the main system. In the real world (lovely place but I wouldn’t want to live there) other considerations are always at play, and the list of options is much shorter. For this job we have a pair of 8″ active two-way boxes with 105° coaxial HF dispersion and a bit of onboard DSP.

For placement I had two options:

1. High center above the proscenium, on either side of the sub. The available rigging spots are either a lighting batten that sometimes changes in height or off the lighting catwalk above the apron which might make aiming awkward. There’s a tieline there back to the system DSP, but no available output channels on the processor unless I want to sacrifice a press feed or something. This turns a few seats into a mono sum mix that localizes high up and in a different direction than all the rest of the listeners, and physically the arrival will be a good 8 – 10 ms late to the listening area, so there’s no hope at all for time alignment.

2. A few feet directly behind (upstage of) the main arrays on a batten for house lights that never moves. Localization is far more consistent (listeners in these seats won’t be conscious of the image shifting away from the mains) and timing is better since we’re within 5 feet of the mains. I can use the LF from the mains and just feather in the HF as needed to fill the measured gap. (There are a few different opinions on this, but keeping lo-mid energy off the stage is a big priority here, as they often have a podium with a condenser mic right under this box. Excellent Jim Yakabuski article about this approach here.) I can just parallel the signal and AC off the back of the bottom box of the array. I chose this option, with the boxes flipped over so the HF driver was on the bottom to clear the adjacent lighting fixtures.

Still have to share processing with the main hang, but the box has EQ onboard. I set the box on the edge of the stage and fed it a copy of the Main L signal off the DSP. The game here is to match the spectral response to the mains above 500 Hz or so, remove as much excess LF as I can using the onboard EQ, and set the level about 10 dB higher than gold because the final throw distance for the box will be about 3x (10 dB) further away from the mic. Get that right, and I shouldn’t have to tweak it too much in the air.

Gold: target trace.

Black: raw fill response

Red: fill response post-EQ

In keeping with the theme of this post, that’s “about as good as it’s gonna get.” Some range of the HF could be aligned to the house average with delay, but as we’ll see shortly that won’t matter. I would have liked a more aggressive HPF and a true LPF at HF rather than the shelf but still we’re hanging within 3 dB of the target from 400 Hz to out past 12k. I’ll take it. Again, these traces are level normalized but the level for the fill was set about 10 dB above the reference trace.

Interesting note: notice that adding the LF crossover filter in the fill box DSP (here at 125 Hz) doesn’t change the LF phase response. The box is designed to mate with several different companion subs, and the manufacturer is using internal all-pass filtering to keep the net phase response over several options for HPF frequencies. This must be accomplished via IIR filtering, as the box latency was around 3 ms, not nearly enough time to use linear-phase FIR at those frequencies.

Once we flew it and aimed it, the level was pretty darn close to where we wanted it to be. Time alignment isn’t possible, as the box is already further away than the mains sharing the same DSP signal. By this point, however the arrival is already lagging behind the acoustic arrival from the onstage source, so if I have to choose a late arrival to live with, this is a pretty good candidate.

We’ve again reached the “about as good as it’s gonna get” point and I think the before (top) / after (bottom) is illustrative:

Coherence is vastly improved and the first few rows come much closer to matching the spectral response of the rest of the room. These are no longer “bad seats.”