Extending the ARC System to Include a Subwoofer

In Part 1 of this series, we discussed the Anthem ARC (Anthem Room Correction) firmware that is present in their SSPs as well as the AVRs which have a  reduced filter bank size. In Part 2, we will discuss the addition of a subwoofer and how it is integrated into the ARC program using the full SSP version of ARC.

By far, the most significant sonic issue for acoustic music when a subwoofer is deployed occurs around the crossover from the subwoofer to the main speaker. Adding a subwoofer to a system using a standard bass management system without a room EQ in the loop (sub and main channels) degrades the flatness of the response in this critical area. At a minimum, the in-room transition band of the low-pass and high-pass filters should track the shape of a fourth-order Linkwitz–Riley (4th order LR – more on these two designers work is below) filter to about 20dB down. With a 4th LR at 80Hz crossover, both speakers are active between 60Hz and 110Hz (-10dB points), a frequency range that is populated with a variety of common instruments playing throughout the score.

A typical bass management system provides only a second-order filter for the main speaker, which will not sum to flat, even in an anechoic chamber. Developers of standard bass management systems (circa 1995) assumed the main channel speaker will rolloff naturally with a slope of 12dB response exactly below the crossover frequency, so they added only a 2nd order filter section to create the fourth-order filter.

This will never happen. Room EQs must provide additional correction for the main speaker channel to have the correct -6dB point and transition band associated with a 4th-order LR high-pass filter. Different main channel speakers require a custom filter synthesized by the adaptive room equalizer. Once the EQ performs this function, the system is flat in an anechoic chamber, but this is only a small part of what the room EQ must do.

Without additional electronic equalization, the room effects (uneven sound pressure preservation using Tom Nousaine’s terminology; see my Technical Article Subwoofers: A Brief Look at the Effectiveness of Using a Subwoofer in a Music System) corrupt the shape of the low-pass and high-pass filter transition bands. The correct in-room high-pass filter transition band shape of the main speaker is as critical as the in-room low-pass rolloff of the subwoofer. Only an advanced electronic room EQ in a Pre/Pro or AVR can provide the required 4th order LR response at the listening seat for the main channel.

The crossover frequency must also be under the user’s control. The user must select a crossover point high enough that the subwoofer enters before the main channel speaker’s distortion starts to increase. At the same time, the crossover should be as low as possible to prevent localization of the subwoofer.

I have found lower crossover frequencies starting around 90Hz (but not much lower than 70Hz as explained in my article cited above) result in smoother response at the join between the main channel and subwoofer. This observation is room dependent. 100Hz – 130Hz tends to be an area with a lot of closely spaced peaks and dips on both the main and subwoofer channels. Given the different spots for the peaks and dips for the main and subwoofer channels (they are in different places in the room), matching becomes more difficult. If the EQ filter bank is IIR filters, we could run out of IIRs to fix all the response deviations of the subwoofer at the higher crossover frequencies in the critical area that both the main channels and subwoofer are active. I also note that speaker-to-room boundary interactions (Allison Effect) begin at about 100Hz, adding to the complexity of the main channel curve.

Some electronic room equalizers will not allow the crossover frequency to be adjusted. Since no room EQ system makes distortion measurements, the automatic selection is often wrong, with the setting far too low. Too much low frequency energy is applied to the main channel woofer, which will overdrive it.

Below I have zoomed in to the portion of the ARC Target Customization panel that involves subwoofer selection.

For this panel, only an L/R Front speaker (stereo configuration) was used, but you can see that you can select the crossover frequencies for each speaker pair and subwoofer separately. This provides a different high-pass frequency (HPF) for the fronts, center, sides, and rears. I call front, center, sides, and rears the main channels speakers to distinguish them from the subwoofer.

The bass management graphic user interface on the TV screen does not have to be set up. When ARC is enabled, the bass management filters are automatically selected per the panel above.

The subwoofer has a separate single low-pass filter (LPF) frequency. All the full range signals from the main channels are summed to a single mono signal, and the single low-pass filter follows before the signal is applied for the subwoofer.

If you set some speakers to a higher crossover value than the subwoofer LPF, you will get a hole in the response from the point the HPF crossover has rolled off the main speaker to the point the LPF allows the subwoofer to becomes active. This may be unavoidable if you are using small rear and surround channel speakers, with larger speakers for the left, right, and center.

If you set any HPF frequency lower than the subwoofer LPF frequency, you will get an overlap and a peak in that range between the LPF value and the lower HPF value. You should never be setting the panel to create this condition.

To prevent the problem of creating an overlap, or a hole, four LPFs would be required. One set for the front channels summed to mono, one for the center channel, one for the mono summation of surrounds, and one for the rears. Each of these four LPFs would be set to the value assigned to the corresponding HPFs selected in the control panel above. I have never tested an AVR that did not mix all main channels followed by one LPF. Adding the three additional low frequency LPFs requires the DSP to have six additional 2nd order IIR filters. This would require a more powerful DSP, or if the DSP cost is to be kept constant, a reduction in the size of the filter banks for the room EQ.

Yet another LPF to the subwoofer should be present for the LFE (low Frequency Effects) input (sometimes called subwoofer-out on a DVD or Blu-Ray player). The issue is the LFE channel has a spectrum up to 100Hz. In Anthem products, the LFE is summed with all the other channels before it enters the subwoofer LPF (note no selection box for the LFE channel in the panel above). If the LPF is selected lower than 100Hz (we want to match the main channels HPF to prevent overlaps discussed above), then information in the LFE above the selected crossover will be lost. Some AVRs have an extra LPF at 100Hz just for the LFE, which then bypasses the LPF for all the other main channels, and solved this problem. This is shown in the panel above.

If you want the full LFE frequency range reproduced, you will need to set the subwoofer to 100Hz to hear all the sound effects (music will never be in the LFE channel if the mix engineer is minimally competent). As explained above, this comes with a really significant cost. An overlap between the L/R/C and the woofer will resultant in peaking in the overlapped region. With ARC, the Cinema mode may become a useful workaround. This allows for using a subwoofer crossover of 100Hz for movies but matching it to the lower value selected for L and R for music.

Another way around the problem, exclusive to Anthem, is to use the only speaker setting you cannot control on the ARC PC display. In an on-TV screen setup menu generated by the Anthem AVR, there is an option called LFE Bypass Xover, which sends the full LFE channel to the subwoofer (this works under the assumption the LFE channel has been band-limited when the movie was mastered, as should be the case). Since the PC is gone for normal operation, you also need the menus to turn ARC on and off.

 

Setup for Performance Verification Testing of the ARC System with a Stereo Configuration that has a Subwoofer

To test the ARC system with a subwoofer, I used a subset of the main channel speakers I used for the tests without a subwoofer presented in Part 1 of this review. The subwoofers used for these tests were from NHT and Sherwood. The sample measurements below are for a pairing of the Infinity C336 as the main channel and the NHT B10d as the subwoofer. The latter was recently reviewed in Secrets. The Infinity appears here so as to compare results with my review of the HK 990 in Secrets and the Sherwood R972 with Trinnov printed in the Boston Audio Society Journal Speaker (BASS v33 n2). While I am showing only one brand of speaker in one room I replicated the result shown below other speakers and rooms.

The left and right channels speakers were placed in different parts of the room to produce different frequency responses below 200Hz. This was done to see how ARC worked under different conditions. The positions were both reasonable places a speaker would be set in the room. I did not attempt to create an artificial condition by placing the left channel in the corner.

The target panel settings are shown below for all the measurements presented in this document is part of the ARC review.

You can see 80Hz was used for both the L/R fronts and the subwoofer. Failing to match mains and subwoofer frequencies renders everything below invalid. I did verify that the use of other matched crossover frequencies will yield similar results to what is shown below. The ability of ARC to produce exemplary main channel – subwoofers blending with a user selectable crossover is a significant advantage of this system.

While the test below was for a stereo setup I verified that the center and surround channels produced the same results (5.1). The rear speakers (6th and 7th speakers in a 7.1 system) were not directly accessible from jacks at the back of the unit.

Below I have expanded the view of the target panel to shown two of the optional advanced setting.

A 500Hz Max EQ frequency prevents the EQ from applying any correction above 500Hz. A Room Gain of 0dB produces a flat low frequency target. The room gain setting was discussed in Part 1 of the ARC review. The white space cuts out a section of the control panel containing text that supplies some information to the user setting the panel.

Why the filters Siegfried Linkwitz and Russ Riley invented are important to you

Below is the ARC PC display after making a 9 point measurement with the ARC microphone.

As with the full range speaker tests in Part 1 of this review the points are around an 18 inch square and center as the starting point. This the first time I am describing how the ARC PC program shows multiple speakers. This is a 2.1 system. Eight frequency response plots would be shown for a 7.1 system.

What is shown in the green curves in the three graphs above is the response of the individual drivers based on the correction filtering ARC has calculated for each channel. ARC cannot predict the complete response with the main channel and subwoofer both operative, only the individual drivers.

What we are most interested in with these curves are the characteristics of the highpass filter in the main channels and the lowpass filter in the subwoofer path. These should represent a 4th order Linkwitz–Riley (LR) filter set. With an LR LPF (main channel) and LR HPF (subwoofer) at the same crossover points and filter order, the signal path should sum to constant amplitude. The key attributes of the LR filters are the 6dB down point at the crossover frequency and the slope of the filter below the crossover. The LR filters should show no peaking in the passband, and after the crossover point it should become asymptotic to 12dB /octave (2nd order) or 24dB octave (4th order).

The graph below is from Wikipedia:

http://en.wikipedia.org/wiki/File:Linkwitz_vs_Butterworth.svg

It shows a pair of 2nd order LPF filters (LR and Butterworth). A corresponding set of HPF filters are also shown. The summation of the LPF and HPF filters for the Butterworth and LR are the fifth (Butterworth sum) and sixth (LR2 sum) curves. The 2nd order Butterworth filter has a 3dB peak at the crossover point, but the LR filters sum to a flat response.

Since the graph above shows 2nd order filters, the slopes are at asymptotic to 12dB per octave.

Technically a 2nd order LR is two 1st order Butterworth’s in cascade. The 4th order LR is two 2nd order Butterworth filters in cascade (often called Butterworth squared). LR filters are even-ordered only. Odd-ordered crossover filters do exist that sum to unity gain but may have issues the even-ordered LR filters do not. This is beyond the scope of this review.

Below I have shown magnified sections (how to magnify an ARC plot was discussed in Part 1) of the frequency response plots for Left and Subwoofer channel.

We need to verify if the post correction -6dB crossover points and asymptotic slopes are correct to form a 4th order Linkwitz–Riley (LR).

The upper curve showing the highpass filter response (upper curve) was expected to match the lowpass filter response (lower curve) in the lower curve (-6dB at 80Hz with an asymptotic slope of 24dB per octave). The principle designer of ARC, Peter Schuck, explains the reason the upper curve displays different data:  “ARC displays the target highpass response without the 2nd order highpass filter that is added by the bass manager. Thus it displays a 2nd order highpass target (Butterworth with -3dB point at nominal crossover frequency). The final (acoustic) response with the crossover turned on will be 4th order Linkwitz-Riley (Butterworth squared) at -6dB at the nominal crossover frequency. For the subwoofer ARC displays the target 4th order LR response that is provided by the bass manager.  This may seem a bit odd but makes sense from the following point of view: the difference between the 4th order LR subwoofer target and the subwoofer measurement is the ARC target EQ response. Similarly the difference between the 2nd order Butterworth highpass and the speaker acoustic measurement is the ARC target EQ response. The lower graph shows the subwoofer response on the ARC PC display, and has the correct 24dB octave low-pass LPF rolloff”.

 

Measurements

Measured data using my independent room acoustics measurement system of the individual main channel speakers

We now turn to the key question: Does the acoustic response of each speaker, after ARC electrical correction, show flat response in the passband, the correct -6dB value at the crossover and the 24dB octave rolloff?

The graph below is my acoustic measurement of the left channel alone with the ARC room equalizer enabled.

It is measured using the same technique outlined in Part 1 of this review (9 point spatial average). Each point my measurement microphone was placed corresponds to a place the ARC microphone was placed for the room measurements. In theory, the curve should be down 6dB at the crossover point (80Hz), and since it is a 4th order filter, the final slope should be 24dB per octave. The curve should show no peaking across the full frequency range.

To verify the cutoff frequency and rolloff slope of the left channel, I needed to expand the span of the graph to 33dB and limit the upper frequency to 500Hz to accurately measure the slope. Using 0.33 octave smoothing eliminates small ripples in the graphs that obscure the asymptotic slope. The measured results closely correspond to the expected 80Hz -6dB point and the 24dB octave slope. The asymptotic slope of a low- or high-pass filter with no transmission zeros is quantized to 6dB per order of the filter (6dB for a first-order, 12dB for a second-order etc.). There is no such thing as 3.6 order filter. Since we are looking at the curve of an acoustic measurement in a room from a real speaker, small deviations from the theoretical quantized values is inevitable. No other room EQ I have tested is capable of producing a curve with a near textbook shape as ARC has done here.

Moving on to investigate the electrical inverse curve using the pair of curves below

The top plot in this curve set has an enlarged left channel frequency response from the ARC display. As we demonstrated in Part 1 of the ARC review, the ARC acoustic measurement before equalization (red curve in the Anthem plot above) closely correspond to my independent measurements (verified but not shown), so we will use the ARC graphic. Note the span of the plot has been reduced by cutting the display below 40dB out.

The bottom curve in this set is the electrical transfer response from left-in to preamp left-out. The six-channel input was used for this measurement, but any input will yield the same result. The black curve on the electrical inverse correction plot is the bass management system with no room equalization applied. It has a slope 12dB octave, which is what the specification for a standard bass management system requires as discussed above.

The red curve in the bottom graph is with the ARC room EQ activated. You can see how different it is from the black curve. In addition to the added peaks and dips which fill the corresponding changes in the pre-equalized acoustic frequency response (red curve on the top graph), you can see the high-pass filter slope. Note how much faster the red curve on the lower graph attenuates the signal than with ARC not activated (black curve on lower graph). The exact rate of attenuation of the electrical response alone is hard to see because the speaker itself is starting to roll off at about 50Hz. ARC must account for the intrinsic speaker rolloff if the post EQ acoustic curve is to be correct.

That the electrical inverse correction curve is calculated correctly is without dispute, given the acoustic measurement shown on the previous graphic – my independent acoustical measurement of the post EQ acoustic response is shown in the plot just above this one.

Please note again that the green curve (post equalization) on the ARC plot (upper plot) shows only the slope (12dB / octave) for the bass management system, but the acoustic measurement verified the acoustic response has the required 24dB / octave slope. The reason for this discrepancy was explained above by ARC designer Peter Schuck.

As an aside, I have circled the area around 500Hz on the electrical inverse correction curve (bottom graph) in blue. The Max EQ setting is entered in the advanced dialog box. Note the smooth transition to the reference value above 500Hz with no discontinuity. I placed an orange box on the Anthem graphic above 500Hz. Note how pre- and post-correction curves are on top of each other. This is the expected result, since no electrical correction is occurring in this area. This is the ideal result. This was discussed in greater detail in last section of Part 1 of the review but with a different Max EQ value and different speaker room position.

Below is an acoustic measurement of the right-channel main speaker taken in the same manner as the left channel.

The result is also close to textbook standards despite the significant difference in the right channels pre-equalized response.  The curve set below has a copy of the ARC frequency response display on the top for the right channel. The bottom is the electrical correction response from right-in to preamp right-out with the bass management system activated. This pair of plots presents the same data I discussed for the left channel above.

Recall the left and right channels were deliberately placed in the room to produce different response characteristics below 200Hz. Compare the electrical response curve with ARC EQ (red curve on bottom graph) for the left and right channels. Note how different they look responding to the different pre-equalized response of the left and right channels. That ARC can produce such complex yet very different curves dependent on room conditions speaks to the highly advanced algorithms used to build the DSP filters.

Again note the smooth transition to the reference level above the 500Hz Max Eq (blue circle).

Measured data using my independent room acoustics measurement system of subwoofer

My independent measurement of the acoustic response of the subwoofer alone with ARC activated is shown below.

The response was measured using the same procedure as the left- and right-channel main speaker’s acoustic measurements shown above.  For this plot, I used a smoothing of 0.15 octave to better resolve the flat passband response of the woofer, which extends from 30 to 60Hz. I deliberately expanded the span of the graph to 45dB and set the upper frequency limit to 200Hz to better resolve the slope.

The ARC system has done an estimable job in flattening the response in the subwoofer’s passband. The anechoic rolloff of the NHT woofer (-3dB) is about 29Hz. It is not as extended in this graph because of room mode cancellation around 30Hz. The -6dB point, which should be at the 80Hz crossover and asymptotic slope, correspond closely to the expected values. The critical low-pass filter slope is once again near textbook in shape. The small slope offset that occurs -30dB down has no effect, since -30dB indicates the subwoofer’s output has already been decreased to 0.03, and its output is so low as to not be adding to the in-room acoustic response above this point (190Hz).

The curve set below has a copy of the ARC frequency response display for the subwoofer as seen in the top graph. The bottom graph is the electrical inverse correction response of the bass management system activated for the LPF (subwoofer path).

This pair of plots presents the same data I presented for the left and right channels above.  This electrical inverse correction response measurement (bottom graph of the pair) is made by applying the test signal to the left six-channel input (not the subwoofer input, which is for the LFE channel) and measuring subwoofer preamp output level. An identical result would be found driving any other channel since, as discussed above, only one LPF is used for the subwoofer channel in Anthem products.

Displayed on the bottom graph are the response of the bass management system alone and the response with the ARC EQ in the signal path. The traditional bass management system has the correct 24dB octave slope, but the curve with the EQ active is different, as the room effects are canceled. As expected, the dips in the lower red curve corresponding to each peak in the upper red curve on the ARC graphic. Note these corrections are required well into the crossover transition band. The correspondence is easier to see with the subwoofer, because the subwoofer is relatively flat to ~150Hz. In the corresponding plots for the left- and right-channel main channels, the speakers were rolling off at the same time the electrical rolloff was being applied. This obscured the complete rolloff slope.

Again, I am not showing my own acoustic measurement to verify the accuracy of the upper red curve (Anthem pre-corrected subwoofer response), but I did do the measurement and confirmed that they matched closely.

The acoustical measurements above verified the left, right, and subwoofer crossover frequencies were very close or at the -6dB level, and slopes were very close or at the 24dB / octave value with each speaker playing alone. The key question is what happens when the subwoofer and one of the main channel speakers play together? The PC display for the ARC software cannot show this, since it requires another set of measurement after the filter coefficient have been downloaded to the Anthem Pre/Pro or AVR.

Measured data using my independent room acoustics measurement system of the of the complete system

Below I show my acoustic measurements for the complete system after all the equalizer filter coefficients, bass management filter settings, distance, and level correction have been downloaded to the Anthem D2 Pre/Pro.

This plot highlights the remarkable success of the ARC system. The upper curve is the nine-point averaged acoustical measurement with the left main speaker and subwoofer operative. The lower curve is the right main speaker and the subwoofer playing together. Note the span has been reduced to 21dB, which is the same value I used when verifying the performance of the ARC system with a full-range speaker in Part 1. The smoothing is 0.15dB/octave, and the maximum frequency of the graphs is limited to 200Hz to better examine the response near the crossover. Above 200Hz the subwoofer is down 35dB.

For the left channel only, a -2dB drop at 70Hz is noted. For the right channel, a slightly larger drop, with a maximum of -3dB, is seen from 75Hz to 95Hz. Note that the frequency variations outside the crossover area are of the same order of magnitude. From the natural rolloff of the subwoofer to 200Hz, the frequency response changes only ± 2dB. This is a remarkable result, considering the complex modal pattern of the room, in addition to required blending of the main channels and subwoofer that are many feet apart.

Note that like almost all room correction systems, ARC cannot tell if the woofer is in or out of phase. If the subwoofer inverts phase, you will get a dip in the crossover. A test disc with 1/3 octave bandwidth noise will tell you if things are correct. Only connect the left channel speaker and play the noise on the test disc centered at the crossover. Now reverse the polarity of the subwoofer and listen again. The result with the loudest level is correct.

 

Conclusions on the Performance of ARC with a Subwoofer

I have presented a significant number of graphs in this section. They represent only a small population of the measurements I made to confirm the ARC system is able to join main channels (four different speakers tested) and a subwoofer (two models tested) seamlessly. Of the room correction systems I have tested, only the Anthem ARC has the ability to create the correct filter shapes for each channel correctly, with the corresponding flat total response curve. This result is achieved at any sensible crossover frequency chosen by the user – an option not available in some room correction systems. A few room-independent room EQ components selling for well into five figures for eight channels may be able to produce these results, but I have not tested them. Such independent components introduce a redundant ADC and DAC into the system that could affect sound quality. This is why it is critical that the state-of-the-art room correction system, such as ARC, be internal to the AVR or Pre/Pro

Many other reviewers have tested ARC-enabled products and provided subjective impressions, so I will provide only a simple observation. Before the Anthem ARC, I had never heard a system with a subwoofer that had not been degraded in the crossover area. To me, this degradation from other room correction systems I have had access to, was far more significant than the extension of the bottom end. I am not the only one who appears to have come to this conclusion. How many equipment reviewers who work in the domain of two-channel music systems are using subwoofers?  If these reviewers had access to an Anthem product with ARC, some minds might be changed.