Transitionally Finicky

Sudden pattern or frequency-response transitions can make audio systems do unexpected things.

Please Remember:

The opinions expressed are mine only. These opinions do not necessarily reflect anybody else’s opinions. I do not own, operate, manage, or represent any band, venue, or company that I talk about, unless explicitly noted.

cardioidWant to use this image for something else? Great! Click it for the link to a high-res or resolution-independent version.

Most of us have probably heard the story of “that nice, sweet dog that suddenly bit Timmy.” If you are, or have been an aviation enthusiast, you’ve probably also heard the stories about aircraft that abruptly departed from controlled flight and crashed. In both cases, everything seemed to be perfectly okay, and then, BOOM! The canine or the airplane turned around and “bit” someone.

This disconcerting behavior can also happen with audio rigs. You’ve got a system that seems nice and stable, and the show’s rolling along, and everybody’s happy –

SCREECH! WHOOOOOOOOOM! SQUAAAAAAAAWWWWWWKKK!

The rig goes into feedback and “bites” you.

As with dogs and aircraft, a sound system always has a reason for doing what it does. If the rig got tipped over the edge of stability, there’s a logical, physical explanation for why. When it comes to an audio system seeming to be just fine, and then suddenly behaving in a terrifying way, I’ve come to believe that there’s a primary factor to look for: Where does some part of the system display an abrupt transition in polar or frequency response?

A Polar Expedition

If you’re not familiar with the concept of polar response, it’s actually a fairly simple concept. It’s the varying sensitivity of a microphone or loudspeaker at different angles around the device. Microphones and loudspeakers are the inverse of each other, and so what the measurement concerns is also inverted. For a microphone, the signal source’s location is variable and the observer’s location is “fixed” – that is, we observe the output of the microphone by looking at the voltage from the outputs. For a loudspeaker, the signal source’s location is fixed (the speaker’s input terminals), and the observer’s relative location is what changes.

In the case of microphones, we tend to assume that the polar pattern is the same for both horizontal and vertical angles. A sound source going off to the left of the mic at some angle is presumed to be picked up with the same sensitivity as a source that is under the mic at the same angle. For loudspeakers, life is more complicated. A great many sound-reinforcement loudspeakers are “asymmetric” regarding the horizontal and vertical planes. Standing off to the right of a loudspeaker at 45 degrees may not get you the same apparent output as standing above the loudspeaker at 45 degrees.

But anyway – let’s talk about transitions in polar response. We’ll stick to mics for this article, because the concepts translate pretty easily to loudspeakers and speaker placement.

Changing Patterns

omni

That picture is of a theoretically perfect, omnidirectional pattern. It’s exactly the same everywhere. It’s transitions are infinitely small, because there aren’t any. As such, an omni polar has very predictable characteristics. If someone suddenly grabs an omni microphone and flips it around, its tendency toward feedback isn’t going to change very much. You can’t “point” an omnidirectional microphone at the monitors, because you can’t point one away from the monitors either. An omni mic “points” everywhere all the time, to the extent that its response pattern is perfect. (Also, when I say “point it at the monitors,” I don’t mean glomming onto the mic and shoving it right up into the monitor wedge’s horn. I’m talking only about the orientation of the mic, not a change in its distance from any particular thing.)

Whether or not you can get generally usable gain-before-feedback with an omni mic is a whole other discussion, and a highly application-dependent one at that.

Now, let’s look at some directional patterns, like a cardioid and supercardioid response. In these pictures, zero degrees (directly on axis) is to the right. The numbers are “pressure units” – NOT decibels. The first picture is side-by-side for greater clarity, whereas the picture with responses overlaid is better for comparison.

sidebyside

overlaid

(Please note that, in manufacturer specs, the supercardioid “tail” is flipped around to provide a more intuitive graph.)

Directional responses are great for live-sound mics, because they give us a shot at hotter monitor levels before feedback stops the fun. There’s a tradeoff, though. Both cardioid (blue) and supercardioid (red) responses are more “finicky” than omni, because their feedback rejection is dependent upon the mic’s orientation. Point the mic in the right direction, and everything’s great. If somebody twists that mic around so that it’s pointing at a monitor, and you might have a problem. The problem can even be worsened by you being able to squeeze more gain into your signal flow: Suddenly, all that extra gain – which was counteracted by the mic’s orientation – is now applied without attenuation. Thus, feedback can build far more aggressively.

What about a comparison between cardioid and supercardioid?

The first thing to see is that I’ve scaled the graphs so that “2 pressure units” is an overall reference. We’ll call that 0 dB, and I’ll quietly do the math to transform the other values into decibels.

Cardioid

Supercardioid

0 degrees (On Axis) 0 dB 0 db
30 degrees -0.6 dB -0.8 dB
60 degrees -2.5 dB -3.5 dB
90 degrees -6.0 dB -9.5 dB
120 degrees -12.0 dB -177.1 dB
150 degrees -23.5 dB -12.2 dB
180 degrees -259.4 dB -9.5 dB

The cardioid does have a single, deep null, but overall the response transitions gently towards the front. The supercardioid, on the other hand, has a deep null that occurs in the midst of the front-to-back transition, along with a tighter pattern in general. This is great for getting better gain before feedback, but only for as long as the mic is oriented correctly. If the mic is sideways in comparison to a monitor, and then abruptly turned to face that same wedge, it’s as if you gunned the monitor feed +9.5 dB. That’s 3.5 dB more than the cardioid, and might be enough to push things over the edge.

There’s also the whole issue of when someone “cups” a mic such that it bevaves largely like an omni. The degree to which this is a problem depends on how far away from omnidirectional the mic was to start with. A highly directional mic that changes to an omni has undergone a HUGE and abrupt transition. Supercardioids (and similarly patterned transducers) tend to be less forgiving of being cupped, because they have a tighter pattern than a cardioid. The change they undergo is more pronounced, and again, they may have also been run at higher gain. As such, the problem tends to be compounded.

Feeling Peaky

In much the same way as polar patterns, smooth frequency response is more forgiving than responses with narrow peaks. For instance, here are a couple of graphs of theoretical microphone frequency responses. Which one do you think would be tougher to manage, in terms of feedback?

frequency1

frequency2

I will certainly grant you that a 10 dB transition in mic response is rarely what you want in any case, but look at the difference in the rate of change between the two graphs. One has a relatively gentle 3 dB per octave slope. The other rockets away at 10 dB per octave. The response with a large dy/dx (there’s that calculus thing again) is more likely to suddenly hit you with aggressive, unexpected feedback than the gentler slope – speaking on average, of course. Each system that mic goes through also has its own transfer function, and that transfer function may help you or hurt you when it combines with the mic’s response.

Where things get REALLY hairy is when you have even steeper peaks. They might not even top out at the same magnitude as what I’ve presented here, but that doesn’t stop them from being pernicious little creatures.

See, when fighting feedback, we prefer to use very narrow filters. Their steep transitions allow us to select and cut only a small portion of the audible spectrum, which makes those filters hard to hear. The problem, though, is when a similarly steep peak gets introduced into a device’s frequency response. That “hard to hear-ness” is still in effect, but the peak represents positive apparent gain rather than negative. With very little warning, feedback can take off like Maverick and Goose feeling the need, the need for speed. (Name the movie.)

I had a peakiness experience happen to me recently with an otherwise very nifty carbon-fiber guitar. The instrument sounded nice, seemingly having no strange resonances at all, but it would squeal at 1kHz like nobody’s business…and with no warning. It would be fine, and then go nuts. We eventually killed the problem, but it took all of us by surprise.

If you’re having weird system stability problems that are hard to pin down, start looking for devices (and acoustical phenomena, too!) that display abrupt transitions in either broadband sensitivity or their frequency curves. Their finickiness might just be the source of your issues.