Street Talk: Tweaking DSP for Stage Monitors
Tips and tricks that help maximize the potential

By Bruce Main

In the days when digital signal processing (DSP) first stalked the arena, it was the guys at front-of-house that had all the fun. They would ignore their comm light for long periods of time while staring intently at asymmetrical crossovers on their laptop displays.

But now that DSP is affordable enough to consider using it for a dozen mixes or more, the guys at the other end of the snake are beginning to experience the joys of audio in the digital domain. There has been a lot of discussion in various forums about whether the digital approach to equalization (EQ) is fast enough or user friendly enough for the monitor hot seat.

There are some systems like the Ashly Protea or the BSS Varicurve (as well as a few others) that are designed to make the process of “barking” the monitors a one-person job by providing a portable remote that allows you to roam the stage and EQ mixes from right in front of the wedges. Definitely a time saver. And after the monitors are tweaked you can maintain a heads up stance at the console with your eyes on the band where they belong instead of bending down, grabbing 2.5 KHz on the bottom EQ in the rack with your precious ears a foot from the cue wedge.

STARTING IT FLAT

However, show EQ is not what I want to cover here. The real beauty of DSP in monitorland is in optimizing wedges and side fills so that little or no equalization is needed to minimize feedback. You’ll still have a rack full of graphics, but the show will start out with all of them set flat. These units will only be used for on-the-fly feedback reduction or to “personalize” a mix. And with DSP, a preset can be designed for each model in your inventory, particularly useful if you have if you have multiple wedge types. So let’s move on to the tweaking part.

Measured with TEF, a high-frequency device before equalization.

Very few monitor cabinets are inherently “flat.” There are compromises made due to factors like box size versus low frequency driver selection, box size versus horn selection, low frequency driver dispersion versus crossover frequency, etc. Your favorite microphones also have frequency anomalies, especially in off-axis response. Combine the two and things can get quite complicated. Many of these problems cannot be completely cured, but with some DSP horsepower and some modern test equipment, they can be minimized.

The first step is to achieve flat on-axis response from your monitors. I would highly recommend that if you don’t already own (Gold-Line) TEF, (SIA) SMAART or at least a real-time analyzer (RTA), then beg, borrow or steal some audio analysis equipment for this portion of the procedure. At an AES convention a few years ago, an equipment manufacturer that I have a relationship with set up a loudspeaker, an EQ and a pink noise source in their booth. Passersby were invited to try to equalize the pink noise by ear to achieve a flat response from the speaker. One person was spot on, but most of the results were pretty scary, even though the participants were all audio professionals. We need accuracy of +/-1 or 2 dB to provide a truly flat baseline to work from. Some may have the ears to accomplish this, but most of us don’t. If you’re using an RTA, be sure to get one that displays in increments as small as 1 dB.

Measuring the low-frequency response of a loudspeaker minus the room’s acoustic contribution is a somewhat tricky proposition. A 40 Hz wavelength is approximately 28 feet long. In order to measure those frequencies properly with a time windowed measurement system, the window has to be long enough to contain at least one full wavelength. Unfortunately that means it is long enough to contain room reflections that contaminate the measurement. At higher frequencies this is not a problem because the wavelengths become short enough for the windowing to provide anechoic measurements. Because real-time analyzers are time blind, they include room reflections at all frequencies.

TESTING METHODS

There are a couple of ways to deal with this. One is to do your testing outdoors far enough from any buildings to minimize reflected energy. My favorite approach comes from Don Keele, one of the really smart guys in our industry. He places the measurement microphone about one inch from the center dome of the woofer, takes a measurement, then places the mic one inch from the port, takes a measurement, and then sums the two responses.

The signal-to-noise ratio of the measurement is improved greatly because of inverse square law gains that result from being so close to the source. A word of caution here: if the device under test has a maximum output of 120 dB at one meter, its output at one inch will approach 136 dB. This may be enough to do bad things to your expensive measurement microphone. So start out at a fairly low volume level and work your way up. This test will give you a good idea of what is going on from 200 Hz on down.

The first parameter I set is a high-pass filter to prevent signals that the box is not capable of reproducing from wasting power and potentially damaging components. Then correct any large EQ anomalies with parametric filters.

Again with TEF, the data of a high-frequency device after equalization.

The microphone can be moved to the normal listening distance for this and all subsequent tests. Make sure that the distance is at least three times the longest dimension of the box under test. This puts us in the far field. I do this test with no crossover engaged for the low-frequency device. If you test from 200 Hz to, say, 5 kHz, you get a good idea of the total low-frequency response curve of the box. Next, look at the upper range of the device’s response curve. There will be an obvious point where the amplitude drops off or the speaker gets into breakup modes represented on the test display by narrow notches and/or peaks.

Use parametric filters to flatten the response as much as possible within the useable range of the device. Choose an upper crossover point for the woofer that filters out the nasty modes and only utilizes the relatively flat part of the speaker response.

The high-frequency test is next. I prefer doing this test with no crossover engaged, however the sweep frequency must be started at a high enough frequency to avoid damaging the driver. Check the manufacturer’s recommendation for the lowest suggested crossover point and start your sweep there. If full range pink noise is being used as the test signal, start with the crossover engaged to protect the driver. Using the parametric filters in your DSP of choice, correct for frequency response anomalies.

If the box is using a constant directivity horn you may need to use a shelving filter to increase the high-frequency output above 2 to 3 kHz. Get the response as flat as you can across the full frequency spectrum. Remember, if you leave a 3 dB peak in the response and it happens to coincide with a 3 dB peak in the vocal mic response it will cost you 6 dB of headroom.

You should discover an overlapping frequency range where the low and high-frequency devices are behaving in a fairly linear fashion. The crossover can be set anywhere within that region. As a general rule, if the horn is small, set the crossover towards the high end of the overlap zone. Larger horn mouths provide pattern control down to lower frequencies so if the horn is larger, you can set the crossover point lower while maintaining good directivity from the device.

When you’re done, a configured system should look something like this on the DSP software.

Next, the levels and time alignment between the low and high-frequency sections should be set. With the mic on axis and centered between the horn and woofer, do a full range sweep. Set the crossover outputs so the average volume level is the same across the entire frequency spectrum. Then look at the frequency and phase response at the crossover frequency. Pretty ugly, eh? Using an impulse response or ETC measurement, look at the arrival times for the two devices. Set the alignment delay on the DSP to eliminate the time arrival offset.

Now look at the frequency and phase again. Better? You will need to do a little fine-tuning to get the flattest possible phase line. If you’re using an RTA this part is harder. Try inverting the polarity of the high output on the crossover. You should see a notch at the crossover frequency. Adjust the delay until the notch is at its deepest. Reset the high polarity back to normal. If the time offset is correct the notch should disappear. If your RTA has a 1/12th octave mode, it will be easier to see.

Some real-time analyzers have a speaker timing analysis feature as well. Using asymmetrical crossover slopes can produce better (or worse) off-axis response. Experiment with this if you have time, but this article isn’t long enough to cover all the possible permutations. Program in some brick wall limiters just before the little red lights on the amps start to dance.

Voila! A look at the final measurement result, courtesy of TEF.

LESS FROM TWO THAN ONE? Save these settings as a preset in your DSP of choice and repeat the process for each type of monitor wedge in your inventory as well as side fills and drum monitor rigs. You can also use these settings as a basis for multiple wedge setups. But remember that when you use multiple cabinets of any sort, comb filtering will occur because of the time arrival differences. These peaks and notches are non-minimum phase. That means that they are not “EQ-able”. (Is that a word? It is now!) Because of this, sometimes it’s possible to get less output with two wedges than with one.

But riders being what they are, go ahead and do a preset for dual wedge setups. The crossover and time alignment settings will remain the same, but you may get some summing in the low frequencies. Use the RTA to check the frequency response because a TEF sweep will be too frightening to look at. If you need a preset for absolute maximum output with a particular vocal mic try putting it on a stand exactly as if you were setting up for a show. Plug the mic into the test microphone input on the test rig. The response will be a combination of the speaker under test and the off axis response of the microphone. EQ the response to be as flat as possible. It may not sound pretty, but it will get loud. (At least until the singer cups the microphone sealing off the back of the cartridge, turning it into an omni.)

You may also want to do presets for full range response or one with a higher frequency on the low-cut filter for vocal only. Sometimes it is beneficial to attenuate the lows for an acoustic set to avoid exciting acoustic guitars or pianos. Save a preset and switch back and forth at the appropriate times.

With the current crop of DSP devices, it’s not uncommon to find configurations like four input, eight output that work perfectly for either four two-way mixes or two three-way side fills and a cue wedge. Look for routing flexibility and the ability to store lots of presets.

But most of all listen to the units. Audio quality varies as much with digital equipment as with analog. It’s too easy to buy this type of device based on a laundry list of features and functions when the most important thing is great sound. Just think of how impressed that visiting monitor mixer will be when you tell him that the monitors are all wrung out and the graphics are still flat!

Bruce Main has been a systems engineer and FOH mixer on and off for 27 years. He has also built, owned and operated recording studios and designed and installed sound systems. He can be reached at bmain@vectorcorp.com.

March 2003 Live Sound International