In this demonstration, we show how ETF5.x software may be used to optimize low frequency acoustic response of rooms through trial and error measurements with various speaker placements. High frequency response control is acheived through the placement of foam absorbers at key locations within the room. 

This document was written to be read and understood, rather than followed as a step by step instruction set. Each room is unique and this document illustrates how to take & interpret measurements using ETF5 software to optimize reproduced sound quality.

The software provides all of the measurement taking functions required to optimize room response using the methods outlined below. The software and a few foam absorbers are the minimum requirements to complete a project as shown below.

In our experience methods of optimizing loudspeaker position can be more effective, practical and quite a bit cheaper than using various low frequency corrective devices such as helmholtz resonators.

This experiment has two purposes:

1. Illustrate effective methods that create audible differences not only for the "golden eared" professional, but for the ordinary individual who normally may not be interested in optimizing a sound system.
2. Illustrate methods that are both easily understood and cheaply implemented through features in ETF 5

In setting up the acoustics of home theater systems or two channel audio systems, better speaker placement leads to smoother & more balanced frequency response across the hearing bandwidth leading to better sound quality. Proper placement methods applied prior to further corrections such as passive treatments and equalizers can reduce the requirement for these additional devices. 

The measurements required to optimally place the loudspeakers do not take long. Once a few hours are spent experimenting with ETF a level of comfort with the program leads to taking all of the measurements in each section shown here in less than an hour.  

We used the following equipment for this demo:

• ETF5.x software and & calibrated microphone

• Bryston listening room as shown in the example.

• PMC MB1 main speakers on 20 inch stands.

• Two Bryston 7B power amplifiers, one for each channel.

• Bryston 10B-LR sub woofer crossover.

• Bryston 3B Mono for the sub woofer.

• PMC SB100 sub woofer

• Bryston BP25 preamp

• Denon CD player

The main speakers are capable of more low frequency output than the sub woofer, this subwoofer was used for placement tests and will be replaced with a larger unit. The main speakers in this system often change to other PMC models.

Note: Most of the measurements in this example can be done with an uncalibrated microphone. A calibrated microphone is necessary for EQ adjustment for high frequencies. Most omni directional electret condenser microphones, such as the one sold by Radio Shack as an SPL meter are very flat below 500 Hz.

The low frequency response is primarily determined by the dimensions of the room and the placement of the sub woofer within the room. Optimum sub woofer placement testing using ETF5.x will first be illustrated.

Before you begin:

• Only a single sub woofer should be used for this measurement.

• Main speakers should be turned off or disconnected.

• This measurement set takes approximately 1/2 hour to complete.

• Any crossover circuits used for the sub woofer should be disconnected or set to the highest possible cutoff frequency.

• Masking tape for testing markers of speaker placements.

• Furniture should be left in place, as it can affect the room response and may skew results if moved around.  Movement by people and pets will also skew results.

• Low Frequency bandwidth should be selected for this measurement since we are only interested in frequencies below 200 Hz. SPL should be calibrated to allow comparison of levels between the various placements.

Test positions for speaker location should include all aesthetically possible locations. These test locations are normally between 1 and 2 feet apart.

In Figure A below, letter sets 1, 2,..,8 & 1,2,..,8 represent a set of sub woofer test locations used for this example. ETF allows 8 measurement files to be overlaid for easy comparison of the response for positions (1...8) in each color of locations.

Exact placement of these locations can be recorded using masking tape on the floor. This saves taking precise measurements of location for each point. The exact location of the best points can be measured after the best position(s) are found.

Fig A: Sub woofer Test Grid

The highest frequency for placement optimization is 140 Hz for the sub woofer used in this test. It remains omni directional in radiation to frequencies much higher than 140 Hz. The sub woofer can therefore be placed in the listening position and the microphone can be moved around the various test points! The sound sensitive end of the microphone is placed where the cone center for the sub woofer would normally be. This saves the heavy lifting of the sub woofer and provides very similar test results as if the sub woofer was moved around the test locations with the microphone stationary at the listener.

The grid spacing for the points 1,2,..,8 & 1,2,..,8 will be 2 feet. 

The sub woofer was placed on the sofa with the cone within 1 foot of the normal listener ear location for the following tests. The microphone was placed at the various test positions to generate the measurements below.

The following measurements illustrate test results with the microphone in the positions shown in Figure A and the sub woofer located at the listener location.

The third octave measurements provide an indication of frequency balance while filtering out many room effects. This smoothing highlights the response variation due to wall, floor and ceiling boundary interference.

The natural response of the room (modal response) is a characteristic of the room itself rather than actual speaker positioning and will be examined in a subsequent section.

Figure A1: Locations 1,2,..,8 cone center (mic pickup end)= 9 inches

Figure A2: Locations 1,2,..,8 cone center (mic pickup end)= 24 inches

The 1,2,..,8 measurements of both cone center heights show that position 5 (curve 5) has the flattest frequency response.

Figure A3: Locations 1,2,..,8 cone center (mic pickup end) = 9 inches

Figure A4: Locations 1,2,..,8 cone center (mic pickup end) = 24 inches

Optimal Results without Parametric Equalizer The rear corner placement, (curve 1 / location 1) in figure A3 has the most energy (level in dB), locations closer to the listener provide flatter response (curve 7 / location 7). Placing the sub woofer closer to the listener as opposed to the best position from 1,2,..,8 resulted in better response. Curve 7 is almost flat to 23 Hz. The cutoff frequency for this sub woofer was 40 Hz at 12 dB/octave. 

Optimal Results with Parametric Equalizer: If an approach of tuning a parametric equalizer for the best response is to be taken, the best woofer position is in the rear corner (curve 1, location 1, figure A3). Equalizers can be employed to reduce excess levels and therefore cone excursion and distortion. This optimum position can result in huge energy (cone excursion) savings after equalizer correction.

ETF now has both sweeps and a PSD/Sweep test signal that offer superior resolution in low frequency room tests than these methods. For example measurements, see Help -> Performance Enhancement Package -> PSD/Sweep in the demo download. New capabilities provide far greater resolution in these measurements.

The unsmoothed response shows the sharp variations in response that are associated with room modal response.

Figure A5: Position 2 (cone center 9 inches above floor)

This illustrates no significant room mode spikes. Boundary effects are comparatively small due to the near field effect. This position requires a cable to be run from the system across the floor to the sub woofer, a difficulty in some instances (such as high traffic across the cable).

Figure A6: Position 5 (cone center 9 inches above floor)

The best position at the front end of the room shows quite a bit more of the room effects in the response, but still a remarkably even room mode excitation. Notice the deep nulls that show boundary effect cancellations (low frequency comb filtering). This effect is much more visible than it is audible.

Figure A7: Position 7 (cone center 9 inches above floor)

The rear corner position appears to be one of the least desirable positions for sound quality. It is one of the best positions to choose when using a parametric equalizer because the actual output is greater at lower frequencies and therefore easier to correct without using EQ boost and the subsequent increase in cone excursion. Compare the actual sound levels in Figure A7 with Figure A5 & A6.

Figure A8: Overlaid Response

Figure A8 & A9 highlight the characteristic room response. Notice the dominant spikes at 21.5 Hz, 31.6 Hz, 56 Hz, 66 Hz and 108 Hz in all curves in figure A8. These modes are slightly over excited, and these spikes will exist to varying degrees at all measurement locations.

Figure A9: 3D Graph Of One Measurement Location

The extended portions in time in the above 3 D graph show resonances that ring for a long period of time at the above mentioned spike frequencies. The overlaid measurements are best to find exact resonant frequencies, while the 3 D graph shows the relative sharpness of the resonance. A high Q (sharp) resonance will decay slower with time. Similar behaviour with respect to decay time of resonances would show, irrespective of measurement microphone location because this is a natural characteristic of the room.

Low frequency room correction devices must be narrow band due to the relative small size compared to the room and the required effectivness. These devices are used to correct overly excited room modes and take the form of Helmholtz resonators or quarter wave traps. This is a complex topic and requires lots of experimentation to get to work in practical rooms. This is not recommended for ameteurs.

The above tests illustrated three possible optimum sub woofer placements.

Position 5 gives the best response for conventional placement. Position 2 gives the best placement for near field conditions that were tested. Other positions may be tested around the listening position using the methods for further experimentation in near field sub woofer placement.

If equalization is to be used, the test position with the highest SPL (sound pressure level)at lowest frequencies should be chosen. The equalizer can be used to optimize response. The end result will be less cone excursion for a given SPL. (Position 7)

It has been shown that the difference in response between the best and worst possible sub woofer placement can easily be as high as 10 dB over the range of interest. Careful placement using these techniques can result in a far superior bass response than may otherwise be achieved with arbitrary placement.

Before you begin:

• Only one of the main speakers will be used for this test.

• Sub woofers should be turned off.

• The main speakers should be operated at full bandwidth (no crossover circuit employed) to measure the response throughout the expected crossover region for the sub woofer. Integration of this response with the sub woofer will be discussed in part 3 of this experiment.

• The unused main speaker will be placed as a mirror image on the opposite side of the room after the optimal location is found.

The main speakers in this system will be required to operate above approximately 70 Hz. Their placement should be optimized for frequencies as low as 50 Hz for smooth crossover transition to the sub woofer. This results in a very large square area for the grid but this size is often reduced due to other practical considerations. Upper limit of placement optimization is dictated by the thickness of absorber used to control high frequencies.

Practical considerations must include the fact that Blumien stereo works best when speakers are placed approximately 60 degrees apart, (the distance between speakers is equal to the distance between the listener and either speaker).

The lower operating limit for the absorber operation is determined from the absorber thickness or actual tests on the absorber. The formula below is a rough calculation for the lowest frequency of effectiveness based on absorber thickness:

The response of the main speakers must be optimized between 50 Hz and 565 Hz using the same grid method as for sub woofers above. The response of the region below 565 Hz cannot be controlled with foam placement.

The lower frequency of operation dictates the total size of a square grid for measurement locations including all points:

Practical limitations on speaker placement dictate that this should be reduced to a 2 foot square grid. The point spacing will be determined by the upper frequency region:

Due to practical placement considerations, speaker placements will be tested on a 2 foot by 2 foot grid with 6 inch point spacing. The chosen grid area for test measurements is shown below.

Sets 1,2,..,8 & 1,2,...8 show locations of the center front baffle of the loudspeaker relative to the floor.

Figure B: Main Speaker Test Grid

Two overlay graphs, each consisting of 8 measurements will be generated with the main speakers at each grid location. This can be done very quickly using the ETF5.x One Shot feature to measure and a subsequent file save after each "Shot". The measurements can all be added to an overlay graph after they are completed and saved.

The actual main speakers will be moved so that the front center baffle coincides with each point on the floor labeled with masking tape. A stationary test microphone is placed at the listener position.

Note: The mic - speaker interchange method should not be used for main speaker placement as it is for low frequencies due to the shorter wavelengths involved over this range of interest. 

"Full Range" bandwidth should be used for this measurement. The range of interest is between approximately 50 Hz and 550 Hz.

The following measurements were taken with the loudspeaker center front baffle at the positions indicated in Fig B.

Position 6 (curve 6) provides the flattest response below 500 Hz of the set 1,2,...8. Positions 1 & 2, (curve 1 & 2) should be avoided.

The 1,2,..,8 positions all give rapid fluctuations in the response below approximately 100 Hz. This may lead to poor main/sub woofer integration. Optimal placement should be chosen using Figure B1.

Placement should be chosen from positions 5,6,7 in Figure B1. These provide a relatively smooth response in the main speaker / sub woofer transition region as well as relatively smooth response up to the upper frequency limit of 565 Hz.

From figure B1, a closer examination shows that position 5 should likely be avoided as well. Position 6 gives the best response between the limits of 50 Hz and 565 Hz. It may work well to choose a crossover frequency higher than 70 Hz for the sub woofer for a smoother transition if 5 is chosen.

It is interesting to note that the user had chosen position 6 before this experiment took place. This would otherwise require many hours of careful evaluation by a highly experienced listener without using ETF.  Position 6 also provides a very flat response in the crossover region used.

The lower mid range is the most difficult region of the human hearing range to optimize because wavelengths involved are short enough to generate response that is highly dependent on listener position. The wavelengths are too long for this frequency region to be controlled with absorption. Listener position sensitivity of response increases with increasing frequency over this band of frequencies.

In this section of the example a single main speaker and the subwoofer will be operated at the same time to evaluate the integration of the response between the two units. Various methods for tuning this response are outlined.

Before you begin:

• There are several possible ways of doing this, all involve tuning of the response while both units are being measured simultaneously - one main speaker and the single sub woofer. This should be done using the "Low Freq." bandwidth setting.

• In many cases the sub woofer may be placed much closer to the listener than the main speakers. In this case, the sub woofer output will have a smaller signal propagation delay than the main speakers due to the different physical lengths between the listener and respective loudspeaker. This results in poor time synchronization between the two units.

• The sub woofer can be up to 25 ft closer to the listener than the main speakers for ETF to capture the response of both units. The sub woofer can be further away from the listener provided that the gate time used in the measurement is large enough to capture both responses (usually the case).

ETF provides a measurement of the loudspeaker - microphone distance and propagation delay time. All channels in a system should be measured for propagation delay. Appropriate electronic delay can be added to channels having the shortest propagation times so that all channels are synchronized in time for a particular microphone (listener) position. The new Bryston SP 1 surround sound processor makes this adjustment between channels automatically.

Note: If the sub woofer and main speaker cannot be time synchronized, only the "Low Freq." bandwidth selection should be made when testing main speakers and sub woofers operating together. 

In cases where a time synchronization is available to make both units more time coincident, it should be set before these measurements are taken. Fine adjustments to this delay setting can take place as in (3 - below) to optimize response more carefully. ETF5.x may be used to measure the speaker mic distance for the sub woofer and main speakers independently.

(1) Take a measurement with the normal and reversed phase connections on the sub woofer. (Reverse the phase by reversing + & - connections on the unit or amplifier). The flatter of the two responses is the desired result.

(2) Many powered subs provide a continuous phase adjustment on the sub woofer. Varying this from -180 degrees to +180 degrees in increments of 30 degrees with measurements taken for each phase setting will provide a set of measurements from which the best response can be chosen.

(3) Fine tune the electronic time delay adjustment if the surround sound processor being used has one. 

(4) Parametric equalization used with the sub woofer or main speaker response can be used to optimize performance in this transition region.

The room was modified to include some additional furniture before these measurements were taken. This changed the responses at the optimal positions for the sub woofer and main speakers slightly. The sub woofer placement chosen was at the room front close to the main speakers. This position was chosen because the SP1 surround sound decoder was not yet available and therefore no timing adjustments between the sub woofer and main speakers was possible.

The measurement below shows SPL calibrated sensitivity measurements of the sub woofer and main speakers. The sub woofer channel had to be decreased in level.

Figure C: Sub Woofer & Main Speaker Response

The 24 dB/octave crossover slope provided by the Bryston electronic crossover made sub woofer / main speaker integration very smooth. Levels were adjusted during a Sequential Acquisition measurement to yield the response shown below.

Levels were adjusted and crossover point was set at 70 Hz.

Figure C1: Sub Woofer + Main Speaker Response

The response obtained by using the frontal sub woofer position was held within almost 6 dB below 200 Hz. A repeated grid test of the sub woofer and main speaker to optimize their response further may have improved this response. The addition of furniture did play a role in changing this response from the optimal response.

Response may have been improved for rear sub woofer placement and the correction of time delay differences between the sub woofer and main speaker.

Before you begin:

• You will need another person to assist you in the mirror placement, as well as a chair/ladder they can stand on.

As explained in Part 2, the main speaker placement grid was used to optimize response below 565 Hz. Careful absorber placement will be used to reduce the effect of the room on frequencies above this point.

Absorbers placed using the mirror trick on the side walls, rear wall and ceiling were used to control the early reflections that occur before 10 ms in this room.  (See diagram below)

The listener is successively seated in each location of the room deemed to be a likely position for a listener. At each  location, the listener observes a second participant move a mirror along the ceiling and wall surfaces of the room as well as any other suspect hard surfaces. 

If the seated listener can see a loudspeaker in the mirror, the mirror is in a spot where sound waves can reflect from the surface to the listening position. These locations may require absorption. 

The mirror trick works because light waves reflected from the speaker reflect from the mirror the same way that high frequency sound waves reflect from hard surfaces.

After the placement of each absorber, an ETF impulse response measurement should be taken to verify the correct position of the absorber and to verify that the absorber is actually reducing the level of the reflection. If the absorber is not necessary, it should be removed.

Measurements before placement of any absorbers are illustrated below. Each measurement was taken with the microphone placed at the listener position.

The reflections shown in these graphs that occur before approximately 12 ms are those resulting from surfaces that were later covered with absorbers, except in the case of the floor reflection.

The impulse response is the most popular way of looking at this response, but has a problem in that only high frequency information is visible. Lower frequencies are more spread out in time and lower in level for the same energy, this prevents them from being easily seen on the impulse response. This problem can be avoided by using a band filtered ETC response.

Figure D2: Impulse with no absorber placement

Figure D3: FFT result on Impulse Response.

The effect of absorbers on actual perceived frequency balance and response is minimal. Absorbers improve imaging qualities, particularly when used on the ceiling. The absence of a ceiling reflection makes the room and the sound stage seem much larger and more lifelike. 

Many audiophiles object to a ceiling absorber, our recommendation is to try it. Foam absorbers can be used with double sided carpet tape. The appearance is not as undesirable as one might think.

Absorbers were used as shown below:

• Side Wall: 6 inch thick, 4 feet by 4 feet on each side

• Ceiling: 4 inch thick, 4 feet by 4 feet on each side

• Rear Wall: 12 inch thick, 8 feet by 8 feet

The effect of this amount of absorption was to reduce reverberation time by approximately 50 ms across the range above 1 KHz. The room did not sound different for talking, except when standing under the ceiling absorbers. In this case the room did sound larger.

The band filtered ETC curves taken after the placement of absorbers is given in Figure D5. Note the before (figure D1) and after differences in the 0 - 5 ms section of the graph.

Figure D5: ETC with Absorbers.

Figure D6: Impulse Response with Absorbers.

This measurement can be misleading (D6). It shows that only the high frequencies have been removed. It is difficult to see low frequency behaviour in the impulse response because low frequency energy is more spread out in time in the measurement.

Band filtered ETC or log impulse squared measurements are much more indicative of room behaviour across the spectrum.

The careful set up of this room, often with only one microphone position used may lead some to believe that the listener must be carefully seated for an optimum response.

Tests for listener position sensitivity were carried out by taking measurements at various positions along the back of the couch where listeners would normally be seated. The results are shown in Figure E1 .

Figure E1: Listener Position Variances in Response.

Correct absorber placement leads to lower listener position sensitivity at high frequencies because reflection levels are reduced. The small sensitivity at low frequencies can be explained by the longer wavelengths associated with lower frequencies.

The position change effect on low frequencies is given in Figure E2.

The response between approximately 30 Hz to 20 KHz is held to almost within a 6 dB set of limits for each listener placed on a 6 foot wide couch.

The dip in the midrange response could be corrected by toeing the loudspeaker in toward the listener. The toe in angle was 0 degrees for this experiment.

When the actual geometry of the room is considered, low listener position sensitivity can be predicted. The relative remaining reflection arrival times do not change with different listener positions to the extent one may think. The changes in loudspeaker position do effect the relative reflection times from the various room surfaces to a larger degree and therefore provide a greater degree of change between positions. This is of course dependent upon room geometry.

The resulting "sound" of this room is very dead because of the absorption of early reflections. The level of the early reflections determine the relative "liveness" of a room. Fortunately there is room for experimentation. The ceiling reflection should always be absorbed but there is controversy on the subject of side wall absorbers. Many prefer to use no absorbers on side walls in symetrical LEFT - RIGHT room arrangements. 

Ceiling absorbers should be used if absorbers are to be used on side walls in the same manner as the ceiling absorber.

Overall, these methods provide substantial improvements to the measurements in this example. Exhaustive subjective study was not carried out, it would therefore be inappropriate to make conclusions on the subjective sonic improvement. 

The room will remain set up as done using the methods shown here.

The measured 1/3 octave results of the completed room changed very little when gate times were varied. This would indicate that the measurements are subjectively accurate.

The best measured speaker locations did coincide with the users original placement that was done with many hours of experimentation by a listener with 25 years experience in setting up high performance systems. 

Speaker placement alone have a 10 dB - 20 dB effect on low frequency response. This can be optimized quickly and effectively with only ETF5.x software.