It’s usually not enough to simply design and install a great sound system, using high-quality components, and thoughtful design and installation practices.
Most of the time, once everything is in place and powered up for the first time, the new system will still sound at least somewhat lacking or unpolished.
If you are interested in learning more about achieving better audio within houses of worship, check out the following session, "20 Steps to Better Sound," slated for the WFX Conference & Expo next month in Orlando.
Modern loudspeakers (especially powered models with built-in processing) are getting much better at sounding good “out of the box.” Even then, though, a number of factors, including the room itself, can make for some disappointing listening experiences.
That’s where the art of loudspeaker system optimization, aka “tuning the PA”, comes into play.
Proper optimization can yield stunning results, even on very old and/or lower-quality sound systems. For example, I just retuned a 15-year-old system a few weeks ago, one that was based on more “old school” design techniques: individual drivers arrayed in what’s sometimes (jokingly, even lovingly) called a “junkyard cluster.” Despite the dated look of the system, it had been designed and installed well. The results, after carefully bringing this system back to its glory days, were wonderful.
The tools have never been easier to obtain than they are now, both in terms of selection and price.
However, unless you know what you should really be measuring, how to properly measure it, and how to interpret the data, you likely won’t have much luck getting the desired results.
In this piece, I’m going to walk you through some of the more important points.
The first thing you need to know is that the RTA (real-time analyzer; a bar-graph like display of various frequencies) is not an adequate optimization tool.
Yes, it’s popular, and easy to get as an app on your phone, but it has a very limited perspective.
The biggest problems are that it is “time blind,” meaning it can neither distinguish between direct and reflected energy (such as echoes and reverberation). It also cannot show you phase, and it is unable to give you a real sense of what the PA system and room are doing to your original stimulus signal (e.g., pink noise).
As a result, the RTA display is showing a lot of different sonic contributions all lumped together, and we need more refinement in our data to make educated decisions on what to do next.
That’s where transfer function measurements come into play. With a transfer function, the original stimulus signal you’re using to test with, is compared to the response from the measurement microphone. This allows you to see the exact contribution of the sound system and room to the results you’re measuring; see time-alignment issues so that you can correct them; and even use something called a window function to restrict the measurement to just direct energy from the loudspeakers (under certain measurement conditions).
Transfer functions are available on a various software packages, including the ever-popular Smaart from Rational Acoustics.
Even by using transfer functions instead of RTA displays, we still may compile bad data. This is why it’s so important to choose measurement locations carefully, and to understand how various positions will affect the results.
For example, placing a microphone at ear height to simulate the listener’s perspective often makes sense, but reflections from any nearby surfaces (most notably the floor) will skew the frequency response (via an effect known as “comb filtering”). This can make it appear as though there are undesirable ripples in the loudspeaker system response, but they aren’t really there. Well, they are there, acoustically. It’s just not the loudspeaker system that’s causing them, and therefore we usually shouldn’t try to correct them.
Since it can be difficult to know the difference between the loudspeaker system’s actual behavior, and the associated contributions of acoustical interactions, like the “floor bounce” discussed above, it’s critically important to use multiple measurement locations to find an average response throughout the room. Not only will this prevent making EQ decisions based on what might just be anomalies, it also helps ensure that the system sounds as good as possible everywhere.
You can produce average measurements with many software platforms, by moving your microphone throughout the room and taking successive measurements. You can also do this much more quickly, by using multiple (preferably identical) measurement microphones with a multichannel audio interface, so that your software can show you the averaged data in real time.
Even the inexpensive measurement mics available today are quite good, so it’s a lot less cost-prohibitive than you may think, considering it saves a lot of time and generally produces better results.
Now that we (hopefully) have some good transfer function based, averaged data, we can move on to making actual changes to the PA.
I recommend doing things in the following order:
1. Set the crossover, driver delays, and relative driver levels properly, for any actively-processed (two-way, three-way, etc.) boxes (note that this is not the same as “powered” boxes, which have built-in amplifiers).
2. Adjust the overall loudspeaker box levels of the whole PA (with everything on, except auxiliary systems, such as delay fills), so that you get the most even SPL throughout the room. For line arrays, you’ll get the best results from the array, if every box is the same level, although gradual “gain shading” is sometimes needed for even coverage.
3. EQ one box (for point source systems) or one entire array hang (for line arrays) by itself. Then bring in the rest of the “main” system, and compensate with EQ as needed, for low-frequency build-up. (As a general rule, the same processing should be applied to every box of a line array. Do not be tempted to add different delays throughout the elements of the array, or to EQ them separately, unless you really know what you’re doing).
4. Finally, add in any auxiliary systems (front fills or balcony delays, for example) and get them time-aligned, EQ’d, and gained to match the main PA.
When you’re in the EQ phase, try not to boost things that seem to be “missing.” Oftentimes, those dips in the response are either inefficiencies of the loudspeakers or acoustic issues. Boosting those frequencies will rob you of amplifier headroom, make the loudspeakers work harder than necessary, and often won’t sound particularly good. It can sometimes make sense to boost the high end a bit, to get even top end, but be careful and intentional, when you do.
Finally, don’t forget to use your ears while you EQ. I typically do this process in two phases. First, I get the system dialed in to measure flat, and then I use an additional “layer” of EQ (on flexible architecture DSP units, that allow you to define your signal flow) so that my “tasteful” EQ choices are in a separate virtual processor. A flat line may look good, and it’s worth pursuing initially, but it’s not necessarily the final target.
There’s so much more to say about loudspeaker optimization than can be included in an article, but hopefully these tips I have outlined in this piece, can steer you in the right direction the next time you find yourself wrangling a system.
Just remember that it’s easy to get data, but it can be hard to get meaningful data, so take some time studying the process and experimenting.