Secrets Q & A
- Written by Brian Florian and Colin Miller
- Published on 30 November 2007
Many surround sound receivers and processors these days offer an "Auto" set-up routine that attempts to configure the basic set-up for the consumer, including whether a speaker should be high-passed or not, which frequency to high-pass it at if applicable, the distance/delay setting for each speaker, and of course the calibrated level (loudness) for each one.
The process often includes EQ for some or all of the channels, usually with the tag that doing so "corrects the room". Seem like a good idea? Yes, a great idea, but when it comes to the practical application in a real room, the results can leave much to be desired.
In such systems, the manufacturer supplies a microphone with known anomalies that can be accounted for in the calibration software. You plug in the microphone, place it at a listening position, the system cycles test signals from different speakers, comparing what the microphone picks up to what it sent to the respective speakers, in terms of time arrivals, frequency response, phase response, polarity, etc., and then applies 'correction' in the form of compensatory delays, level-matching, polarity swaps (if necessary), and frequency response equalization as best it can to make up the difference.
So what's the problem? In terms of setting delay and matching channel levels, if the software/microphone combination works correctly, nothing. It should still be verified with manual calibration (because sometimes these systems produce inaccurate adjustments, or you might want to fine tune it after the fact), but in handling the basics, there's no reason why this portion of the auto set'up can't work just fine. The problems really arise in "Auto EQ."
"Auto EQ" (a.k.a. "Room Correction"), in many cases, ranges from a marginally helpful 'band-aid' to useless or severely detrimental alteration.
All rooms start as inherently flawed in terms of acoustic transparency. If they've got walls, floors, and ceilings, and most do, they have surfaces that offer reflections in additional to the direct sound that, assuming a reasonably good loudspeaker, we want to hear primarily, if not almost exclusively. Getting a room that controls these reflections through deliberate absorption and diffusion (dispersed reflection) treatments (use of wall and ceiling panels, i.e., Room Treatment) is a mixture of science and art. Note that, for the sake of discussion, we differentiate here between "Room Treatment", which we define as the use of physical objects in the room (absorption and diffusion panels), and "Room Correction", which we define as the DSP applied in SSPs and receivers to change the sound that you eventually hear. We have a previous article on the types of materials used in room treatments.
There are many consumers who have actually spent time and effort addressing acoustic issues by using room treatments in a sound system who will testify that it's a far more fruitful upgrade than the time and money spent swapping out cables, components, or even loudspeakers. Of course, actually doing this, compared to playing audiophile nervosa, is a real effort, an effort most consumers, and even hobbyists, will not tolerate, and are eager to rationalize into the lowest possible priority. In extreme cases, some would rather debate about refining their sound with digital audio cables than talk about slapping up some rigid fiberglass panels on reflective surfaces.
Enter "Auto EQ", or "Room Correction".
If you could electronically correct for room acoustics, that would be great. Unfortunately, the fact is, you simply can't.
The acoustic character of a room is imparted when the loudspeaker (or instrument, or person) creates a sound. That sound then travels in multiple directions. Some of the sound travels directly at the listener(s). Most of it travels somewhere else, bounces off of surfaces, and much of it eventually reaches the listeners after the 'original' sound that took the direct path. No matter what you do to the sound before it leaves a loudspeaker, you can't undo what happens to the sound after it leaves the loudspeaker.
This reverberant sound field, the sound that reaches the listener through reflection, is fundamentally different from the direct sound field that travels without the alteration of reflected surfaces. It's not only delayed from the direct sound field, but also delayed among itself, in that there is a gradual decay, and instead of instantly stopping, gradually fades away. The rate of decay is described as its T-60 time, or the time that it takes for a sound to drop 60 dB, or a factor of 1 million (for each 10 dB, the sound decreases by a factor of 10, and 106 = 1,000,000).
When this reverberant sound field recombines with the direct sound field at the listening position, the multiple delays as well as the differences of reflection characteristics of the various surfaces, add relative time arrivals, stretch out the duration of sound, add to the directionality of the sound, and alter the frequency response of the system.
If it's done in a controlled manner so that nothing extreme happens, the effect can be pleasant, and even more dimensional or believable. After all, this process happens in the real world, and our cognitive systems process this information to provide information about the location of the sound, and the environment in which the sound occurred.
That doesn't seem so bad, does it? Maybe not, from a subjective standpoint, if you like the effect, but it masks any acoustics actually recorded, and in any case, diminishes the ability to resolve the original content. We're not advocating that people listen to music in an anechoic chamber. After all, the recording engineers anticipated some reverberant character in playback systems. But, for optimal results in serious listening applications, the 'live' character that's so nice to enhance your singing in the shower is the worst case scenario for an audio system, and if you want something resembling accurate playback, the typical living room is a nightmare.
"Room Correction" claims to compensate for the anomalies resulting from the recombination of direct and reverberant sound field information. It does this by comparing the test signal to the recorded signal, and applying DSP to complement the end result, mainly by applying equalization to boost response dips, and attenuate response peaks. In a simple world, this would be a good idea. However, in our world, serious problems arise. These are:
- Lack of precision in EQ, measuring, or application. Without very precise processing and very high resolution, the EQ adjustment is often way too broad. For instance, it may attenuate a band with a peak present to average a flat response. However, that band, say 1/3 octave, may not just have a slightly lower peak, but a couple of new dips above and below that frequency. It may measure flat with a low resolution measurement like a 1/3 octave spectrum analyzer, but in effect you've just shifted your problem into multiple problems. In the cases where EQ can be of substantial benefit (i.e., subwoofer notch filter attenuation), the measurement must be capable of resolving 1/12th an octave or smaller, and the EQ itself should be of the fully adjustable variety, where not only amplitude can be adjusted, but also the specific frequency, and the width of the boost/cut, or 'Q' of the filter.
- You can't undo reverberation. The reverberation of a room is similar to that of a bell. If you hit a bell and ring it, you can't undo the ring by hitting it differently. If you hit it differently, it may sound different, but it's still ringing.
- With the possible and contentious exception of notching frequencies low enough to be relegated to subwoofer reproduction exclusively, our auditory system will pick up the changes in the direct sound. Instead of sounding like the room went away, it will sound like we're in the same room, but something screwy happened to the original event. In attempting to fix one problem, we've done little to fix the problem, and have created another.
- Even if the above problems could be addressed, and the room 'corrected,' there is the fact that the acoustic character of the room changes with listener position. That is, even if we could provide a correct compensation for one listening area, the compensation would be a total crapshoot when it came to the listening experience just two feet (one seat) away from that position.
Bottom line: Simply applying EQ to a speaker does not "correct" the room. Although you can counter frequency response aberrations at the source (such as may be inherent in the speaker), you can do nothing for issues beyond the loudspeaker.
To correct a room, you must physically change what is in the room and where it is in the room. This means a minimum of sound absorption material to attenuate the strongest and most detrimental reflections, as well as spending serious time positioning and aligning the subwoofer, or in some cases using multiple subwoofers to smooth out the response. Then and only then does EQ become the "final tweak".
Even after applying room treatments, many automatic room correction systems fall short in that they only provide correction when taking into account a single listener, the one who sits with their head in a vise at the exact same position the microphone was in during set-up. We can illustrate this by going through the auto-set-up routine with the microphone in different positions and then checking the settings: It will choose very different settings depending on which seat the microphone was placed at when you ran the auto routine. In some cases, the people seated elsewhere end up with WORSE sound!
In fact, you could use a device equipped with two microphones and a custom-built collecting surface, often referred to as our head, complete with outer ears, which is a bit more sophisticated than a single microphone, but you might find the adjustments tailored to a single location artificial or distracting. In other words, the sound could even be worse at the optimal listening position.
Is EQ useless? Certainly not! It simply requires more intelligent and sparing implementation.
An outstanding Auto EQ scheme is the Audyssey system we've tested in a couple of consumer products that have licensed it. As we understand it the Audyssey originated as an attempt to automate the complex and time consuming task of aligning the sound systems of commercial theaters. One of its fundamental features which sets it apart is that it takes several measurements throughout the listening area and applies a "best overall" EQ carefully and sparingly. Coincidentally, when we spoke to the Audyssey creators, they agreed that there is only so much you can do, and when it comes to electronic 'correction,' some things are better left undone. Even then, the most dramatic benefits achieved by the Audyssey 'correction' technology lie in loudspeaker correction. If you start with good loudspeakers, the benefits of electronic processing are far more subtle.
Assuming you have a good subwoofer, start by setting all your speakers to Small, using 80 Hz as the high-pass frequency (we have an essay which makes a pretty good case for this), set your distances according to what they actually are, and set your levels with either the built-in noise generator (if you have a THX unit) or suitable external equivalent using a $35 SPL meter from Radio Shack. After getting used to the sound of a standard calibration, fiddle a little bit here and there (experiment with different crossover frequencies within the operating range of both the main and the subwoofer, and massage your channel levels maybe one dB, or in rare cases two dB). If you are really daring, get your hands on a CD or DVD with low frequency sweeps to get a sense for what frequencies are a problem for your subwoofer and use a fully parametric (meaning that it includes "Q" as an adjustment), extremely specific, manual EQ to cut (never boost) the problem frequencies with surgical precision, and then let your ear be the final judge.