|
|
Can a parametric equalizer really replace bass traps?
Comments and analysis by Ethan Winer I've never been a fan of using equalization to correct room acoustic problems. My background is professional audio and recording studios, so I know well that professional studios tried - and ultimately rejected - EQ back in the 1980s. Today, very few professional facilities use monitor EQ. While it's clear that the raw low frequency response can be made flatter with EQ, modal bandwidth and ringing can become worse as you'll see below. Therefore, EQ must be applied carefully and in moderation, and using proper tools while adjusting the EQ is mandatory. You must be able to see the modal bandwidth and ringing times while adjusting, to avoid making things worse rather than better. And don't obsess over fractional dB deviations, especially in the lowest two octaves. Far more important is the "speaking range" for bass instruments from about 80 to 300 Hz, where having a wide bandwidth and reduced modal ringing matter much more. Equally important when assessing the graphs below is to consider the dB span from each peak to its adjacent null, rather than absolute levels across the range. Click HERE to see my first example showing 17 MondoTraps alone versus 17 MondoTraps with EQ. This pair of graphs shows clearly that even at the Front Center location, where the EQ was optimized, the EQ added significant ringing near the 92 Hz marker; it also raised the Q (the peak is more narrow) at that frequency. The null above 128 Hz was also made worse after adding EQ, and the peak just below was made narrower. The main problem with high-Q peaks and ringing is they foster the problem known as "one note bass," where all bass notes tend to sound the same regardless of their pitch. In fairness to EQ proponents, these anomalies are above the 80 subwoofer crossover frequency used in most systems. However, it seems to me these artifacts can and will occur at lower frequencies too if that's where the EQ is applied. Now click HERE for the second example that shows the Right-Rear location with no bass traps, with and without EQ. The EQ was adjusted while measuring the Front-Center, and it definitely helped there. But here, only a few feet away in the room, EQ clearly has made the null near 30 Hz much deeper. In contrast, adding bass traps improved the response and reduced the ringing at all locations. The third example HERE shows what EQ alone can do versus bass traps alone. In my opinion narrow peaks and excess ringing are far more damaging musically than a skewed response. Even at 60 Hz it is clear that bass traps have reduced ringing substantially compared to EQ. In fact, I see no evidence that EQ has improved ringing even a little at any frequency, which was the main point of these tests. Although a mode may appear to have decayed more quickly, the lower level due to EQ just makes it look that way. Look at the final example HERE where the displayed lower level limit was reduced from 60 to 40 dB. You can now see that the ringing is still present, lowered only by the amount of EQ cut that was applied. And EQ has clearly made the ringing worse around 90 Hz in every measured location. Indeed, the entire range above 60 Hz is far smoother with traps alone than with EQ alone. Added January 4, 2006: As I explained in my fourth example above, it looks like the ringing was reduced by EQ because the peak levels were pushed off the bottom of the graph. That example shows more of the ringing by reducing the lower dB limit, but it's still not as clear as it could be because the peak tops are not aligned. So I added a fifth graph HERE with the display range changed to put the levels closer to each other. In addition, I changed both Time displays to a full second, to better see more of the decay. I also added Terry's EQ settings directly on the graph, so you don't have to go back and forth to understand what's being shown. Note that the frequencies Terry adjusted differ slightly from what ETF shows. I assume this is due to small differences in the programs he and I used. I made no other change besides display ranges, and both graphs show a total span of 40 dB to allow making an accurate visual comparison of decay slopes. Starting with the peak at 42 Hz, it is clear that the overall slope is the same in both graphs. If "minimum phase" is the mechanism by which EQ supposedly negates modal ringing, that theory clearly fails in practice. Indeed, this peak starts about 3 dB lower on the lower graph with EQ, yet the ringing extends for even longer. The next peak is at 71 Hz, and again the basic slope and decay times are identical. Likewise for the peaks at 156 and 174 Hz. Once the peak heights are made similar, it's clear that the decay slopes and overall times are the same. What happens at 141 Hz is less clear, but it may be due to applying so much cut in two bands very close together. You can also see a new decay added by EQ between the peaks at 141 and 156 Hz. I can't tell if that's due to EQ or not. If more than one band seemed to exhibit reduced ringing I'd be more inclined to accept that EQ is effective for that. But as is shown very clearly in these graphs, EQ did not improve ringing even a tiny bit at any other frequency. Speaking of ringing, I also highlighted two places at the bottom of the lower graph where the EQ boost clearly added new ringing. Not only at 91 Hz as I pointed out in my initial commentary above, but you can also see ringing added by the EQ boost at 116.5 Hz. Finally, it's important to understand that even with a zillion bass traps - EQ or not - no small room can be made perfectly flat or have all of its ringing removed. At some point you will treat your room as best you can, accept that it will never measure perfectly (though it can sound fabulous), then sit back and enjoy the music! Comments and analysis The frequency domain statistics showed that EQ with or without bass traps improved the frequency response over all the listening positions. Not only was the standard deviation (a measure of flatness) improved with parametric equalization, but the minimum dip and maximum peak were improved as well. Click HERE to see the tabulation of all the frequency response results, and HERE to download the data with graphs in Excel format (1 MB). We used 7 test locations and 4 quality parameters, for a total of 28 cases. In 22 out of 28 cases, EQ alone or 17 MondoTraps plus EQ was superior to no treatment or just MondoTraps. Of these 22 cases, about half (13 cases) were 17 MondoTraps plus EQ. This suggests that the presence of the 17 MondoTraps offered little or no advantage for the frequency response when EQ was present. In the average over all locations, EQ alone or 17 MondoTraps plus EQ won. This is a surprising result. The parametric equalizer was tuned for only one of the seven listening positions, yet had a positive effect on all of them, as well as on the averages. Click HERE to see the EQ settings that were applied. Use of 17 MondoTraps improved the decay time from about 70 Hz to 200 Hz, as shown in the waterfall charts. This large quantity of bass traps also broadened the peaks over the same frequencies. The presence of only 2 or 4 Mondo Traps offered no clear improvement at any frequencies. EQ showed no obvious improvement in the waterfall charts within this frequency range. However use of 17 MondoTraps demonstrated no waterfall chart improvement below about 70 Hz, which are most of the subwoofer frequencies for the typical crossover frequency of 80 Hz. EQ appeared to show an improvement in decay time at these low frequencies. While most musical instruments do not extend this low, many do, notably cello, double bass, bassoon, piano, organ, and harp. And of course, subwoofer frequencies are critical to the accurate reproduction of movie sound. One of the 12 EQ filter settings, a 6 dB boost of the narrowest possible bandwidth at 91 Hz, was set in error. It created a resonance at all of the listening positions, which Ethan observed and pointed out in his analysis. I could have corrected this by simply disabling that particular filter on a second pass of adjustments, had I performed one after seeing the time domain response. This does illustrate the potential problem with applying a narrow-band boost, even a modest one, with a parametric equalizer. It is also interesting that this "ringing" of the equalizer was indistinguishable from a resonance produced by the room. A second order filter is a second order filter, whether it is being generated by an equalizer or by a room. This is what gives parametric equalizers their amazing power to cancel room modes. |
By
Ethan Winer On October 1, 2005 RealTraps partnered with acoustics expert Terry Montlick to settle an age-old question among audiophiles and home theater enthusiasts: How useful is equalization as a way to improve room acoustics? Room treatment, and especially bass traps, are by necessity large and visually intrusive. Many people would like very much to believe that a small electronic device can replace acoustic panels, or at least reduce the need for them. Top The position of RealTraps is that EQ is much less useful than acoustic treatment and bass traps for many reasons: EQ cannot reduce modal ringing much if at all, and any correction is localized and potentially makes other places in the room worse. Likewise, nulls cannot be improved much since that requires extra power and can add ringing, and echoes and early reflections cannot be improved at all. Further, EQ has the potential to make things worse if you don't have a way to measure the room accurately while making the adjustments. Like most pro acousticians, Terry agrees that EQ does not replace room treatment, and is best used in conjunction with it. But he has suggested that EQ can reduce modal ringing at least to some extent, and that the improvement EQ offers can extend over a usefully wide physical range. Therefore, the crux of this issue is mainly whether EQ really can reduce ringing, since even EQ advocates accept most of EQ's failings I listed above. (However, a few EQ proponents do believe that EQ can be a satisfactory substitute for all room treatment.) Top For these tests we used RealTraps MondoTraps, and for the equalizer Terry used a Behringer Feedback Destroyer Pro model DSP 1124P. Although this particular equalizer is not generally considered a "high end" device, it is inexpensive and thus very popular. Many people use it to equalize only their subwoofers, usually from 80 Hz and below. So in that context its quality could be considered less important than if all of the audio were passed through it. However, this is opinion only (and certainly not mine) so it should not to be taken as criticism of Behringer products, or any inexpensive audio products for that matter. Terry used all 12 available EQ bands, applying both cuts to reduce peaks and boosts to improve nulls. However, he limited his boosts to no more than 6 dB; applying more than that is impractical because it increases loudspeaker distortion and is potentially dangerous to subwoofers. Applying EQ boost also creates ringing, and is most likely why all of the "with EQ" graphs show a pronounced ringing around 92 Hz that was not present without EQ. Indeed, having a wide modal bandwidth and low ringing time are audibly at least as important as a flat response. This is why we present 3D waterfall plots below, in addition to simple frequency response graphs which otherwise hide those important attributes. Top Note that we measured from 20 Hz up to 200 Hz, rather than only at the subwoofer frequencies from 80 Hz and below. Some of the worst low frequency problems in most rooms lie between about 80 Hz and 300 Hz which is the "speaking" range of bass instruments. Below that is more the fullness range, where bass instrument clarity and articulation are less important. The tests described here were performed in the RealTraps lab, whose dimensions are shown in the drawing below. The room is exactly 8 feet high, so the microphone was placed 36 inches above the floor to conform to the "38 percent rule." We used the ETF software on a PC running Windows XP, a pair of Mackie HR-824 powered monitor speakers, and a Carver Sunfire subwoofer (11 inch cubed) set to cross over from the Mackies at 80 Hz. Since many listening rooms and home theaters have more than one seat, we measured at seven locations as shown in the drawing. The main location was considered the "Prime Seat" and this is the only location for which Terry adjusted the equalizer. Then we measured 2 feet away to either side, and again at the same three locations behind to simulate the back row in a home theater. We also measured 4 inches to the right of the Prime Seat, to see how much the response varied across that small of a span. As you'll see in the data below, even over such a tiny distance, and at very low frequencies, the response still varied a lot! Top
All of the data below ranges from 20 to 200 Hz, and the 3-D waterfall graphs show the top 40 dB with an 800 millisecond Time setting to see enough of each room mode's decay. The Prime Seat was used to calibrate the level for ETF with the room empty, and all subsequent tests are relative to that initial test. Top While RealTraps prefers the ETF waterfall graphs, Terry considers the steady-state frequency response graphs also generated by ETF just as important. "The 2-D low frequency response graph shows how the room responds to continuous sound," according to Terry. "ETF can write this frequency response data into a file, and this allows computation of such fundamental quantities as maximum peak, minimum dip, and standard deviation, a simple measure of overall flatness." ETF 2-D frequency response graphs follow each set of 3-D waterfall graphs. A gate time of 1150 milliseconds was used to make the frequency data as accurate as possible over the 20 to 200 Hz range. The top 40 dB is plotted, just as for the waterfall graphs. The mean dB value, range (maximum minus minimum sound), maximum peak above the mean, minimum dip below the mean, and standard deviation are given below each of these graphs. The best number for each test location is shown in bold. Top If you own ETF and would like to examine the data, perhaps to create overlays comparing different data sets side by side, click the Download link at the left of each graph to retrieve the ETF file for that location and setup. The links to the left of the 2D graphs are the same as those to the left of the equivalent 3D graphs. (They are repeated for your convenience.) Each file is Zipped for efficiency to reduce its size from 789 KB to about 130 KB. Or click HERE to download all 43 ETF files in a single Zip (5.7 MB). Top The first sets of graphs below show the response measured at each of the seven locations with the room empty. Following that are pairs of seven sets of data for each location. Each pair is 3-D then 2-D, and each set shows the measurements for six situations:
While we had the 17 MondoTraps set up without EQ, we also measured at the Prime Seat with the subwoofer along the front wall about three feet from the corner, to see how that changed the response and ringing. (Sunfire subs emit out of both sides, so it was oriented facing left and right.) That graph is shown at the end, next to a repeat of the graph for 17 MondoTraps without EQ with the subwoofer in the corner, to make it easier for you to compare the response at both locations. Finally, it's important to explain that the ETF waterfall graphs below do not give hard numbers. Rather, you will look at each graph to assess the flatness of the response and the length of ringing time manually. The raw response is along the rear wall of the graph, and the "mountains" come forward over time to show modal ringing. In my comments at left are links to display graph pairs I consider relevant. Click those links to open a new window for each graph pair, then click the graph (or anywhere else) to close the window automatically. Top Added March 8, 2006 ETF author Doug Plumb has created a greatly enhanced room analysis program called R+D, and one of its features is the ability to load ETF data files and display low frequency ringing above 200 Hz. Being able to see the entire bass range makes it easier to see the improvement from adding bass traps, since the "clarity range" for bass instruments extends to at least 300 Hz if not higher. Therefore, I used R+D to analyze three of the ETF data files from the original tests described in this article. All three images below represent data measured at the Front Center location. The first image below shows the response and ringing with the room empty, the second shows the result of adding 17 MondoTraps with no EQ, and the bottom graph is for no traps and EQ only. Note that at this time R+D offers no way to scale the vertical axis so the range shown is 54 dB, much larger than the original ETF graphs shown below. Click the images to enlarge them.
|
ALL 7 LOCATIONS, ROOM EMPTY WITH NO TRAPS OR EQ
|
ALL 7 LOCATIONS, ROOM EMPTY WITH NO TRAPS OR EQ
|
FRONT CENTER
|
FRONT CENTER
|
FRONT CENTER, OFFSET 4 INCHES TO RIGHT
|
FRONT CENTER, OFFSET 4 INCHES TO RIGHT
|
FRONT LEFT
|
FRONT LEFT
|
FRONT RIGHT
|
FRONT RIGHT
|
REAR CENTER
|
REAR CENTER
|
REAR LEFT
|
REAR LEFT
|
REAR RIGHT
|
)
)
)
)
)
)
){kind=link}
){kind=link}
){kind=link}
){kind=link}
){kind=link}