Dr. Peter Schuck, Ph.D., answers in-depth questions about Anthem Room Correction and the engineering, advanced mathematics, and acoustic science that come together to create the world’s most effective room correction system.
ARC algorithms have not ceased evolving since the 2010 launch of the software. A decade of customer feedback and analysis of real-world measurements, not to mention improvements in computer processing power, allow the latest generation of ARC software to deal with a vast array of loudspeakers and rooms. However, the algorithms are only half the story. Anthem engineers have refined ARC Genesis’ recommendation methodology for determining optimum room correction equalization and bass management settings. The design of digital equalization filters is a non-linear process, and no single algorithm can calculate an ideal solution. Therefore, when selecting target response characteristics such as crossover frequencies and slopes, ARC must parse thousands of permutations of potential solutions and analyze the merits of each to determine which is best for your unique audio system.
There are considerable variations in the frequency response of audio throughout a room. Using a room correction system to create equalization filters based on measurement data from a single location may improve the response at that one specific location, but the acoustic performance at other locations can deteriorate. Moreso, if you were to use only one measurement point, you can create an ideal response for one ear but worsen it at the other. The differences in the acoustic response at the five measurement locations are what allows ARC to understand the acoustic signature of your room. By capturing the acoustic response at various points in the listening area of a room, ARC makes corrections that improve the overall audio performance without requiring the listener to keep their head in a vice.
Based on the science of psychoacoustics, Anthem's proprietary algorithms take into account the way humans perceive sound. Our ear/brain mechanism expects to hear specific characteristics when we encounter sound in a room. If a speaker in a room is forced to have a perfectly flat response, it just won’t sound right, and our brains detract from the listening experience by telling us something is wrong instead of letting us disappear into the music and enjoy it. By understanding the relationship between a loudspeaker and a room, ARC can preserve the natural acoustic signature of the room while also correcting undesirable peaks and dips in a speaker's frequency response.
Time-domain is the strength of a signal over time. In the context of room correction, it is a reference to a loudspeaker's acoustic output (impulse response) measured from a specific location in a room. When ARC plays tone sweeps (using short, loud bursts of sound), it measures the impulse response of a loudspeaker. However, when it comes to room correction, time-domain (impulse response) only paints part of the picture. Frequency domain audio based systems are vastly superior to time domain based systems (DTS and Dolby base their codecs on frequency domain).
Once ARC measures the impulse response of a loudspeaker, it has all the information it needs to begin calculating room correction and bass management settings. ARC converts a loudspeaker's impulse response to its frequency response using a Fast Fourier Transform (FFT), which creates a model of the loudspeaker's frequency domain. Once ARC has modeled a loudspeaker’s frequency domain, the software calculates equalization filters in the frequency domain using smoothing and averaging to derive an ideal in-room frequency response.
Sound transmitted from a loudspeaker to a microphone is mixed-phase. Mixed-phase equalization is a method, used in some room correction systems, that analyzes a loudspeaker's response by looking at it as a minimum-phase system intertwined with an all-pass system.
Equalizing a minimum-phase system with precision is effortless. All that is necessary is to apply its invert. However, to create equalization for an all-pass system (the other half of mixed-phase equalization) a time delay is necessary. This time delay can be infinite, so to create equalization for an all-pass system the time delay must be limited. Thus the resulting equalization is not precise.
This imprecision leads to a phenomenon known as pre-ringing, a backward echo that blurs transients, which occurs when mixed-phase equalizers start outputting signals before passing the main bulk of its output. Human hearing is extremely sensitive to this type of deleterious artifact, so time delay and pre-ringing energy must be limited when creating equalization filters. When mixed-phase equalization attempts to correct low-frequencies, the effect of pre-ringing is particularly significant due to the long time delays necessary to create equalization at these frequencies.
Additionally, we must consider that humans have two ears, and for both ears to simultaneously be within an equalized area, it is necessary to examine measurements from multiple locations around a listening position. To pick correction points and perform mixed-phase equalization using data from multiple measurements a mixed-phase room correction system must find frequencies with matching phase that are common to all the measurements sets. While it is easy to pick correction points at high-frequencies above the Schroeder frequency, where phase response becomes random and thus irrelevant, below the Schroeder frequency, where phase response is not random, it is challenging to find finding matching frequencies with matching phase. When only a few frequencies with matching phase exist, the number of available correction points that can be used to create equalization filters is minimal. In other words, if a loudspeaker requires equalization at 50Hz, with a mixed-phase based room correction system, it may not be posible to correct this point and instead may have to attempt to correct the deviation by applying a filter at 60Hz or 40 Hz.
ARC does not use mixed-phase equalization due to the likelihood of harmful artifacts and limitations on establishing correction points when creating equalization curves. Instead, ARC performs frequency domain based room correction which allows the software to select exact crossover points when correcting peaks and dips in a loudspeakers frequency response.
Minimum phase filters have a quick impulse response (with only a small amount of sustained ringing afterward). For example, consider the sound a drumstick makes when striking a drum. The sound appears with a sudden, massive onset that rings afterward, but quickly tapers off, the same way most natural sounds that result from resonating systems occur. Quick impulse responses that result from minimum phase filters mimic this natural phenomenon, and this is why they are ideal for room correction.
All filters that change a loudspeaker's frequency response also change the speaker's phase response. Typical analog and digital filters are minimum phase filters (i.e., filters that have only a small effect on the phase) and change the phase response in such a way that has a particular and unique relationship to the frequency response.
Analog filters rely on passive components (resistors and capacitors) to alter the frequency response. Resistors used in analog filters have tolerances (variances) as high as 1% while capacitors are as high as 5% (more precise components are available, but are prohibitively expensive). 1% and 5% variances may not seem like a lot, but these variances make it very difficult to control the location of a correction point and shape filters with precision, especially if its a filter that deals many correction points (high order filters).
In addition to being imprecise, analog filters are not practical for room correction. To accomplish room correction in the analog domain filters would need to be designed and hand-built on a room-by-room basis to account for the unique acoustic signature of each room. Parametric and graphic equalizers (which allow adjustments using knobs and buttons) offer more flexibility than traditional analog filters, but still lack the precision and flexibility to perform the equalization necessary for room correction.
In the digital domain filters can be created with high precision (using 16-, 24-, 32-bits, or more). This level of precision is ideal for creating suitably shaped filters at exact frequencies.