Audio Calibration

Anthem Room Correction (ARC) System - Part 1

ARTICLE INDEX

Measured Results

Measured results position 1

The curve set below show the frequency response curves before and after the ARC room correction is applied.

The upper curve is my acoustical measurements of the speaker from the listening seat with ARC off (ARC Out of the system). As discussed above, I made nine spatial separated measurements with my test microphone positioned coincidentally with the ARC microphone before and after the ARC room equalizer was activated, and then averaged them. The uncorrected response is +/- 5dB. Most of this response error is room related and is not an accurate representation of the speakers anechoic response characteristics. To see the speaker's excellent intrinsic performance, you need a copy of my Infinity C336 review, published in the November 2010 on the digital AudioXpress magazine. It is currently viewable only with a digital subscription.

The bottom curve is with ARC enabled (ARC In the system). The corrected response from the speaker low frequency cutoff to 5 kHz is ± 2dB. The overall result is very impressive. Most room EQs performance degrades with decreasing frequency, but not with ARC.

I repeat that the graphs are smoothed to 0.1 octave and the curves have a compact span of 21dB. Both of these settings magnify frequency response errors in comparison to curves more frequently seen in magazines with a 40dB span and 0.33 octave smoothing.

Generation of a single pre-equalization response curves form individual microphone measurements

Below is a pair of graphs showing the pre- and post-equalized results for the nine individual measurements (nine points around an 18 inch square and its center).

Here, the nine measurements with my test microphone positioned coincidentaly with the ARC microphone before and after the ARC room equalizer were activated. The average of these curves was shown on the previous pair of graphs. As expected, the individual curves are noisy and it is impossible to trace a single curve in this composite display. At this juncture, I ask whether the post-corrected response exhibits unexpected behavior in any of the individual corrected curves.

The nine measurements receive the same correction because only a single electrical correction is applied at each microphone position. The spread of amplitude in the set of independent curves at a given frequency varies significantly, as can be seen in the pre-equalized curve set. The peak at 80Hz has a very low spread. The spread is larger in the 150 – 200Hz range, then narrows as the frequency rises above 500Hz. The same thing happens in the post-corrected curve. The average shape of the curve is flatter across the frequency range, but at any one frequency, the spread is static. This is to be expected, given a single electrical correction curve generated form the single DSP filter.

The ARC uses a simple average of the nine point spatial average and the pre-equalized response is flatter at some positions. Post-equalization, the center of the nine point average will yield the flattest response. Since I took the measurement set around the center of a single seat, the listener in this position will hear the optimal improvement.

I should note that the first placement of the ARC microphone is used to calculate the level offsets of each speaker, so it should be at the prime listening seat. The first microphone position is also an in-room frequency response measurement position just like subsequent positions.

ARC does not calculate distances between the speaker and microphone. These need to be manually entered using the on screen display of an Anthem AVR or Pre/Pro. This is the one downside of the USB microphone. It is not possible to accurately estimate the delay from the ADC and USB encoder in the microphone, and the USB decoder in the computer. If an analog microphone and preamp were connected to one side of a stereo USB sound card locally at the computer, then the other channel could be used as loopback for a timing reference.

I see a major upside in placing the entire analog signal processing inside the microphone, and little downside in having to use a tape measure, especially since some room EQs appear to make significant errors in the automatic distance measurement. Trinnov (at much higher prices now) is a special case in which accurate distance measurement in three dimensions is essential to its operation. The four microphone array needed to do this requires the use of analog transfer to four local ADCs. As a result of its special requirements, the Trinnov system can make only one measurement, and thus cannot do spatial averaging discussed above.

It is not necessary to simply average the curves in the pre-equalized plots. Some may be given more weight than others. This will produce a different electrical correction curve. With a weighted average, no post-equalized curve will have the flattest response. This is how the Audyssey system works. By shaping the post-equalized average response curve, no single position provides the flattest response. Instead, an attempt is made for the response at each microphone position to have a similar frequency response deviation. All are now suboptimal, but less variation will be seen at each microphone position (see my review of the Audyssey Professional Sound Equalizer in the November 2007 issue of Sensible Sound).

Only the correction curve shape has been changed by the non-linear weighting. The spread at any given frequency pre and post-equalized is the same. It is impossible to have a different spread because the sound captured by the microphone is sent through a single electrical equalizer. The electronics have no knowledge of the microphone's position, and therefore supply the same correction regardless of position. The equalizer is a single-input / single-output system.

To reduce variation seat to seat in a home theater application, properly placed, multiple subwoofers would provide a more optimal result, in the range of the subwoofer, than any single-input / single-output electrical correction system can provide. Above the subwoofer crossover, passive room equalization can also improve the situation, although it is expensive for lower frequency room response improvement.

I have a strong preference for a linear weighting, as used in the ARC system, that provides approximately optimal results at one seat even when the multiple seats are in the room. If the seats are symmetrical around the primary seat, measurements can simple be made at each seat. If the speakers are not symmetrical around the listening seat, and you desire to have optimal correction to be focused on the primary seat, extra measurements could be made around that seat. This in effect allows you to determine the weighting preference for a preferred seat.

ARC allows for two different microphone measurements to be stored. The memory settings are called Cinema and Music, but these are just generic legacy names. You could, for example, use one memory to hold measurements taken just around the prime listening seat (the method I used in this review), while the other set holds measurements over a wider set of seats in the room but still with the primary goal of listening to music.

The Cinema and Music memory settings have many applications that are beyond the scope of this section. Many do not require any additional set of microphone measurements. Each memory setting allows for almost all basic and advanced control parameters to be adjusted independently. The ARC PC display allows you to view all the speakers' frequency response characteristics as measured and after correction, with all advanced modes selected for each memory setting independently in a separate window.

The electrical inverse response curve

In the curve set below I added a third curve (the center of the set), which is the electrical response of the room equalizer measured from the six-channel analog input left channel to the six-channel analog output left channel. The electrical correction curve should be the inverse of the pre-equalized acoustic response. If it has this shape, then the post-equalized average response should be flat. The nine point averaged acoustic measurements (top and bottom curves) made with my equipment are identical to what was presented above.

The electrical inverse correction curve (center curve above) cannot respond to every wiggle in the pre-equalized response. Its response is limited by how many computations can be made in real time by the DSP which generates the electrical transfer response. Attempts to fill large dips in the uncorrected room response will not be attempted since trying to put energy into a room null will not be effective.

Trying to correct each wiggle would also be a futile exercise. As we saw in the set of individual curves, the variation in amplitude between microphone positions at any single frequency is larger than the wiggles in the post-corrected average curve.

The area on the electrical inverse correction curve that has a red circle over it shows the area that the ARC is trying to extend the lower frequency response of the speaker. Note the 6dB limit at 30Hz. Unfortunately a boost at 30Hz will produce no positive result, given that the pre-equalized response shows the speaker has no useful response, in this large room, below 40Hz. Any negative effect would have to be identified by doing pre and post EQ IM distortion measurements.

The three curves are text book examples of how an electrical room correction should work. This is the best result I have seen. It all comes down the how well the electrical inverse correction curve shape can remove the response variations in the pre-corrected curve. Producing an optimal electrical inverse correction curve is dependent on the algorithms used to create the filter from the data provided from the acoustical measurement and limitations of the filter size which is restricted by room available in the DSP chip. It is impossible for you to understand the depth of the intellectual property Anthem has amassed over the years of this system's development to produce this result, not just for this example but with other speakers in different sized rooms.

In the curve set below, a copy of Anthem's ARC PC display (bottom) joins the pre - and post-acoustic averaged curves I measured independently at the top and center respectively. The red curve is the pre-equalized curve, and the green is the system's estimate of the post-equalized response. The ARC graph appears to be smoothed to approximately a sixth octave instead of the 0.1 octave smoothing used in my measurements. I cropped and scaled the ARC plot to match my graphs.

The ARC red curve is close to my independent pre-equalized measurement. The green post-equalized curve is optimistic. This is a limitation of generating the green curve without making another measurement with the microphone with the DSP coefficients loaded in the AVR. As stated above, the green curve is unlike the independent measurements I made with my acoustic measurement system with the complete room correction system operational in the AVR's hardware. Here, the Anthem ARC combines the response of the DSP filter that it has generated from the in-room measurements with the measurements themselves.

The ARC algorithm is trying to create the flattest curve with the EQ in the system by creating an electrical curve that is the inverse of its measurements. Note, while optimistic, the green curve is not a ruler; the electrical inverse correction curve is limited in its ability to fill each peak and dip. That is reflected in the green curve. If you use the ARC's window function to narrow the size of the vertical axis, the correction errors become more apparent. The 1/6 octave smoothing also helps to make the errors more apparent.

Note the small perturbation in the green curve, in the 1 – 2.5kHz area, that can be correlated with similar small errors in the middle curve, which is the independent acoustic measurement of the post corrected room response.

When the red curve is above the green curve, the electrical inverse correction curve attenuates the signal. When green is above red, the electrical inverse correction curve is boosting the signal. You can see the correspondence in the electrical inverse correction curve of the previous curve set (center graph).

When the green curve shows significant deviation it is a sign the speaker and listening seat are not well placed, or the anechoic response of the speaker is poor. The use of the Quick Measurer function discussed above allows you to find a more optimal seat or speaker position.

Verification of performance in position 2

Now, I repositioned the Infinity speaker. I did not attempt to mirror the placement of this speaker as one would do when placing a left and right channel in a room. Instead, I put it in a plausible position that creates a different low-frequency response.

The curve set below show the frequency response curves of the average of nine measurements I made independently before (Out) and after(In) the ARC room correction is applied with the speaker in position 2. The same microphone placement was used for measurements in the curve set above (around an 18 inch square and its center). I changed the toe-in and distance to the back wall of the Infinity C336 speaker. This explains the additional frequency response difference (top curve with the EQ out between position 1 and position 2).

Although the response curves between position 1 and 2 deviate significantly, the ARC system provides similar results. This is the key test of the ARC systems robustness. Except for one notch at 180Hz, the post-corrected response is ± 1.5dB. Part of this result is the flatter response of the pre-corrected curve. It is dominated by clustered peaks between 70Hz – 150Hz, with maximum amplitude above the average value of 9dB.

In the plot set below, I again add a third curve (center), which is the electrical inverse correction curve of the room equalizer, measured this time from the six-channel analog input right channel to the six-channel analog output right channel.

The electrical inverse correction curve (center curve above) reflects the characteristics of the acoustic curve it is trying to correct, dominated by a significant attenuation between 70 -180Hz.

Looking at the electrical inverse correction curve, we see the curve moves up 6dB in the 20Hz – 40Hz range. The speaker has almost no output in this range. As explained earlier, the ARC system tries to extend the low-frequency response of the speaker by shifting the electrical inverse correction curve. Advanced functions available in the target configuration panel can eliminate this boost.

Below, I have repeated my acoustic measurement with ARC out of the (top curve) and in the signal path (center curve). The bottom curve is the ARC PC display. Looking at the ARC red curve of the ARC PC display, we again see a graph that is close to my independent pre-equalized measurement.

The green curve is above the red, below 60Hz, which is an area that I have circled. The difference between the red and green curves is the boost in the electrical inverse correction curve shown in the previous plot set (center curve). Using the advanced configuration option called Response Cutoff, the red curve can always be set to be on or above the green curve, indicating that the boost has been removed. You do not need to measure the electrical inverse correction curve directly to see if a boost or cut is applied. Just look at the position of the red curve relative to the green curve.

I should note using the default flat Response Cutoff setting often provides the best result. Trying to keep the red curve always above the green in the low end will result in some loss of usable low end response of your speaker. It is very speaker-dependent, reflecting the distortion characteristics of the speaker below approximately 80Hz.