January 15, 2013, by Ken DeLoria
In the beginning there were volume knobs, and that was good. Then came bass and treble controls, and that was better still.
But more was wanted, so more was created and thus the equalizer was born. Over time, equalizers have been shaped and honed into the various formats that are ubiquitous today.
Graphic, parametric, semi-parametric, paragraphic, and more – such as Lake’s rather brilliant Mesa filters – entered the market years ago and we can’t seem to get enough of them. At last count, retailer Sweetwater has 70 different models in it’s catalog, with GC Pro offering an even larger portfolio.
How do we sort them out? It starts by matching EQ types to EQ tasks. The use of equalization falls pretty squarely into four primary processes – let’s call them layers – that they’re expected to perform. It’s important to approach these layers in a logical order.
Room. The first layer should always be correcting low-frequency room resonance and any other room-related issues. Usually, this is best accomplished by means of a loudspeaker management tool that includes a generous bank of user-adjustable EQ filters, as most now do. This layer is the foundation that everything else is built upon.
Loudspeakers. The second layer is likely to be far less linear than a console, power amplifier, cables, most microphones, and so on. Even top-shelf loudspeaker systems will rarely be the flattest ruler in the tool box, so this is the layer in which we solve frequency and phase response problems, fix misalignment in the time domain, and integrate the mains with the subwoofers.
Fortunately, modern DSP-based loudspeaker controllers offer a wide selection of corrective tools, enabling precision alignment. Such tools include an extensive range of filter types, incremental delay, protective limiters, and more.
It’s important to note that if a given loudspeaker system is seriously deficient and nowhere near flat, it needs to be addressed first, then the room, then the loudspeaker system one more time. Don’t mistake room resonance for loudspeaker performance problems.
Instruments, Microphones & Vocals. The third layer is tailoring the sound of the vocals and instruments to support the style of music in consideration of what the artist wants to achieve. We add filters to suppress feedback, make that piano, sax, or Sousaphone sound like it’s meant to sound (and/or whatever sound is appropriate for the mix), and shape the overall aural “feel” of the show.
Vocal tonality should flatter the vocalists, instrument tonality should flatter the instrumentalists, and everyone needs enough gain-before-feedback to be clearly heard and understood. If the event includes track playback, then tracks should be given the attention they might need to satisfy the desired production values.
This layer can usually be handled by channel EQ on the console, but outboard devices – graphic or parametric, depending on what you’re comfortable using – can be very useful for taming difficult sources such as podium mics, lavaliers, strings, grand pianos, vibes, xylophones, poorly recorded tracks, and any other source that requires more sonic shaping than the channel strip is able to provide.
Preferences. When the other layers are firmly in place, you may wish to make tonal changes that are purely preferential at the start of sound check. This may be a one-time thing, or it may vary from moment to moment.
Perhaps the guitarist demands an unusual tonality for a certain solo part that defines his signature sound and needs help to obtain it. Maybe the vocalist performs in a wide range of styles and wants to come across differently from one tune to the next (think David Bowie). Or it might be as simple as occasionally compensating for changes made on stage, such as swapping out a guitar or horn.
In any case, preferential EQ changes should ideally be programmable from console presets, or if that’s not available, then programmable from outboard equalizer(s) so that the baseline EQ can always be found again and reverted to. Preferential EQ should never compromise the system’s overall linearity.
Someone very clever, a long time ago, figured out that room resonant characteristics tend to fall somewhere around one-third of an octave in width. Not always, but fairly often. This makes it seem like those 31 bands on your graphic equalizer can take us anywhere we want to go in terms of room correction.
But room volumes and reflective characteristics vary considerably from one architectural marvel to another (and from one trashy club to another).
And this is the important part: room resonant modes have not yet formed a committee to agree that they will always resonate at the ISO center frequencies that are the backbone of the graphic equalizer.
While the foregoing is intended as a joke, it’s absolutely true that ISO centers are arbitrary and the room you’re working in is completely ignorant of them.
So what happens when you identify a room resonant mode with a band center of 140 Hz, very narrow in width but high in amplitude, and the only way you can cut it is to adjust the 125 Hz or 160 Hz filter on your graphic equalizer? Well, it might perfunctorily help the room problem…but it probably will not. Instead, it is more likely to exacerbate the issue (Figure 1).
Figure 1: The top graph (A) depicts room resonance centered at 140 Hz. In the middle graph (B), we see the response of a graphic equalizer attenuating 125 Hz and 160 Hz. In the bottom graph (C), it’s clear that the two filters only served to exacerbate the room resonance problem instead of solving it. (click to enlarge)
Think about how a 1/3-octave RTA and a companion 1/3-octave graphic EQ are like audio venetian blinds. Look through a partially closed blind and you see an image, but you don’t see it all. The brain fills in the gaps. It’s easy to tell that the girl outside the tour bus is a girl. But can you precisely see her form, her clothing, all of her features? You’re viewing a limited amount of visual information and your brain does it’s best to fill in what’s missing.
A similar effect occurs when measuring a sound system with a 1/3-octave RTA and then tune the system with a 1/3-octave graphic EQ. Unfortunately, the end result is not going to be the same as looking at the girl through the venetian blind. I can’t emphasize enough the importance of viewing a high resolution response trace (at least one-twelfth octave), and then using an equalizer that has at least the same (or better) resolution, if the results are to be precise.
Anechoic chambers are built to have almost zero resonant modes because they absorb the sonic energy rather than reflect it, but virtually all other rooms (and for that matter any vessel that contains a volume of air without equivalent absorption) will resonate in response to the air mass that’s being excited by sonic energy. It’s why flutes work, it’s why saxophones work, it’s why pipe organs work…and it’s also why large theatres (and sports arenas) tend to have a lot of acoustic “mud” in the low frequencies.
Resonance is much more pronounced in the low-frequency region than the high-frequency region, except in special cases such as tent structures in which the LF energy is absorbed by the flexibility of the tent walls. In very large rooms (Radio City Music Hall, for example), it’s not uncommon to see LF peaks that are so high in magnitude that it takes ganging two analog filters on top of each to provide the required level of attenuation.
These days, that’s largely become a non-issue because most DSP-based equalizers have a very wide range of cut capability (often -40 dB), but it was a genuine problem in the “good ol’ days” when most analog filters maxed out at -15 dB.
While a well-designed graphic equalizer can be a great tool for shaping the sonic qualities of an individual instrument (I once spent two hours with an 11-band graphic to get a problem snare drum to sound “just right” on an album project), is it also the right tool for tuning a sound system in a resonant room? Numerous professionals make it their first choice. Some believe it’s not even possible to tune a system with a parametric equalizer. Let’s look at this more closely, because some basic engineering concepts can help you make the best choices.
Peaks & Valleys
Resonant problems can actually be improved rather easily, but only if you can accurately identify the amplitude and Q (the ratio of the center frequency divided by the bandwidth) of the resonance.
A high-resolution analyzer is essential for this task. You’re not going to see the precise characteristics of the various resonant modes on a 1/3-octave RTA (real-time analyzer); it’s simply too coarse.
Conversely, an FFT (Fast Fourier Transform) is ideal for measuring the response of a loudspeaker system in a room, as long as it has very fine resolution in the lowest five octaves of the audible spectrum ~20 Hz to ~640 Hz, because that’s where you’ll detect and defeat room resonance.
Of course, a room can’t actually be stopped from resonating with an equalizer alone, but the energy from the loudspeakers can be attenuated at the resonant frequencies. Doing so will radically improve the clarity, intelligibility, and musicality of the event.
Typically, there will be one dominant LF peak, followed by several harmonically related peaks (Figure 2). The frequencies of these peaks are a direct function of how much air volume is in the room. Other peaks, at other frequencies, may also be present in one or more compartmentalized acoustic areas, such as under or over a balcony. These areas need to be examined individually, but only after the main system has been flattened.
Figure 2: The top graph (A) depicts the three primary room resonant modes; note that they are harmonically related. The middle graph (B) shows the response of the parametric filters, accurately adjusted to attenuate the resonant peaks. The bottom graph (C) shows the final result of the correction.(click to enlarge)
Every resonant peak that is identified needs an attenuating filter that precisely cancels-out that peak. The filter should be set dead-on, on the center frequency of the room peak and as narrow in Q as possible. The goal is to surgically cancel the room peaks. Larger rooms usually require four, five, or more filters to remove the primary room resonant frequencies.
Typical amplitudes of room resonant peaks range from just +3 or +4 dB in relatively small rooms to as much as +20 dB (or more) in large rooms. Don’t be afraid to use the full extent of an equalizer’s capability to cancel the peaks. Cutting filters (attenuation) are safe, generally speaking, and will not harm the phase response of the system; in fact, if the amplitude, Q, and band-center are accurately dialed-in, the measurable result will be an improvement in the phase response of the system in those spectral regions where the cutting filters have been applied.
On rare occasions it’s acceptable to use an accentuation filter (boost) to “fill-in” a hole in the system’s response, but this should be done carefully, and with enough time to critically listen to the results with familiar tracks.
In most cases, some aspect of the loudspeaker system’s response, not a room-related issue, is being corrected. Rooms don’t often exhibit acoustic cancellations that need to be boosted, but if the architectural elements are complex enough, or the building materials unusual enough, then it’s possible.
The Loneliest Number
When there is only one measurement microphone available on a project, move it around a lot. If there’s more than one, time can be saved and accuracy improved by instantly comparing one part of the room to another.
In either case, take note of the differences in each room location in respect to others. Don’t adjust EQ for one “sweet spot.” A typical approach is to measure where the console is located, as that’s the quickest and easiest thing to do. But it’s not the optimal approach. Moving the measurement mic around the room, as time permits, will lead to a far better outcome.
It’s vitally important to not use the full system when taking measurements and making EQ corrections. If the room is symmetrical (and most are) start by measuring and EQ’ing only one side of the PA then transfer those filter parameters to the other side (assuming left/right sources). Otherwise you’re measuring the acoustic addition and subtraction of the various L/R PA elements interacting with each other, rather than identifying the true room resonance.
Of course, after all elements of the system are initially adjusted, listen (and measure) the entire system as a whole. Inevitably the LF region will be more pronounced and dominant than when only measuring one side of the system. This is an excellent time to use an LF shelving filter to gently reduce the LF build-up, rather than to add more attenuation to the surgical attenuation filters.
An LF shelving filter can easily be trimmed on the fly by ear, with no dire consequences to worry about. A narrow Q surgical LF attenuation filter cannot.
Sooner or later, audience members will enter the room, thereby displacing a measure of air volume with their relatively solid human bodies, and thus raising the resonant frequency centers. This is a simple effect of physics: the room now has less air volume so it resonates at higher frequencies, just as a shorter pipe on a pipe organ produces a higher frequency than a longer pipe.
Acoustically, the presence of the patrons will be very noticeable in most rooms, both by ear and by measurement. If you have the means to re-measure the system’s response with the audience in place, then you can make some real magic! Products such as Rational Acoustics Smaart, Meyer Sound SIM, SATlive, and others, foster accurate measurements using walk-in music or even the opening strains of the performance itself.
What you’ll see is that the resonant modes, so carefully identified when the room was empty, have now shifted upward in frequency. Sometimes it’s a lot, others only a few Hz. But keep in mind that a few hertz in the lower frequencies is actually a considerable part of the LF spectrum; e.g., 20 Hz to 40 Hz is a full octave, but that octave contains only 20 actual Hz of LF information (in terms of integers).
Due to the surgical cuts instead of broadband swipes, the audible difference that you (and your audience) will hear when you touch up those precise cuts, causing them to once again fall on the new band-centers, can be astounding. Shifting a filter upwards by a fraction of an octave can have a profound effect? Hearing the results, particularly when the program material comprises natural instruments, makes it obvious that the resultant accuracy is well worth the time and effort it took to get there.
At this point it should be pretty clear that this type of precision tuning can’t be accomplished with anything other than a parametric equalizer, and one that has 1 Hz or 2 Hz of resolution in the LF section. Not all do – some jump around quite a bit in the LF region – so choose your weapon wisely, Mr. Bond.
Ken DeLoria is senior technical editor for Live Sound International and ProSoundWeb has had a diverse career in pro audio over more than 30 years, including being the founder and owner of Apogee Sound.