WATSON - Stereo Enhancement Loudspeakers

The formation of a sound stage, of a phantom image, of an Auditory Scene in front of me when I close my eyes and listen to two loudspeakers, is a fascinating experience of the brain at work. The brain uses memory and takes cues from the air pressure variations at the eardrums to form an auditory picture from the recording of an acoustic event. The picture has width, extending at least from left to right loudspeakers and sometimes has depth. The scene appears at and behind the two loudspeakers, which seem to define the minimum distance between me and the virtual sources. When listening with headphones the frontal part of the image tends to be located inside my head, just behind my forehead, even for binaural recordings.

When, as an experiment, the same program material is simultaneously reproduced by loudspeakers and headphones and I lift the headphones a few inches from my ears, so that I hear more of the loudspeakers, then the image moves out of the my head and assumes greater expansion in depth. But I also hear the headphones whispering. So a more full-range source than the headphones-at-a-distance seems to be needed. A pair of PLUTO loudspeakers came handy. I could confirm with them that a second pair of loudspeakers close-up, will expand the Auditory Scene, both in depth and width, created by the primary ORION loudspeakers provided that the volume from the PLUTO loudspeakers was set below the level where they become audibly dominant. Most significantly, localization accuracy of virtual sources is enhanced and intelligibility of speech and choral voices is improved. The AS is sharpened, not diffused. The expansion of the AS adds spatial separation between virtual sources. All this is, of course, very program material dependent. It works best with binaural, with sphere and with near-coincident microphone recordings. Coincident microphone recordings have clarity but are spatially less believable. Multi-microphone recordings tend to be spatial disasters, but there are exceptions. In general, I can tell a lot about the microphone placement and mix of a recording, especially when the volume level of a particular source is not consistent with its perceived location in the overall ensemble of sources. For the recording/mixing engineer such extra pair of loudspeakers might be a powerful tool to analyze and improve his stereo recording techniques. It could improve recordings for standard two loudspeaker stereo playback setup.    

First experiment using PLUTO-2.1 loudspeakers to support ORION-3.3 loudspeakers in expanding the perceived Auditory Scene.  With proper level balance between the two pairs, the four loudspeakers and the room disappear from perception. The illusion remains for a short time even after the PLUTO speakers are turned off. Unlike with headphones one can look at different parts of the AS by turning the head sideways.

The volume level required from the close-by loudspeakers is relatively low so that smaller, less obtrusive speakers could be used. Also the bass response is dominated by the ORION and less capability is required from the support speakers. It even sounded occasionally that the deep bass reproduction this close to the ears shows up the nearby loudspeakers. Consequently a pair of small speakers was constructed using a SEAS FU10RB 3" driver in a 4.5" x 4.5" x 6" box, on top of a 33" long, 3" ID pipe. 

First WATSON prototype with a 3 inch driver in a 4.2 liter volume box/stand configuration. The quick indoor frequency response measurement confirms a useable high frequency extension and early low frequency roll-off at 12 dB/oct due to a  high Qts of 0.96 and Fs = 102 Hz.   The maximum output is limited by Sd x Xmax = 15 cm3. In practice I discovered that a flat frequency response makes the speakers too easily noticeable and I had to drop the response above 1 kHz considerably. This was easily accomplished with a passive line level RC network. But the low end of the response required greater output volume than the driver and 40W amplifier could provide comfortably.

Consequently a slightly larger driver with bigger motor and higher efficiency was used, the 3.5 inch SEAS W12CY003. It rests on top of a 4" to 4" pipe coupler, fires upwards and stands 34" tall on its 6.4 liter enclosure. Due to a low Qts of 0.36 its low frequency roll-off starts at a 6 dB/oct rate around 180 Hz and turns into12 dB/oct below 40 Hz. Its Sd x Xmax product is 22.5 cm³ and it is also 3 dB more efficient than the 3" driver.

The WATSON loudspeakers are listened to a 900 off-axis. Their frequency response rolls off above 1 kHz conveniently and no additional response shaping seems necessary. The steep 24 dB/oct roll-off of the PLUTO woofer by itself above 1 kHz loses too much high frequency information. The overall response of WATSON and its sound at 900 is more like that of a good telephone than a high fidelity loudspeaker. Regardless, the speakers are highly effective in expanding the AS and in disappearing. Compared to the ORION their output level is low.

Normally I use the Tracker Pre in conjunction with my ARTA acoustic test software, but here it serves as a convenient delay line (actually the only one I have handy) and volume control. Mute and Stereo/Mono switching are used for experimentation. Some recordings benefit in imaging from switching the WATSON pair to mono. The Volume 2 level in B-C mode is usually fixed. For a few recordings it has to be increased, for others decreased. In some case the SEL could not do anything useful to the recording, which was essentially broadband noise with no meaningful spatial information preserved, scrambled eggs.   In practice the setup is sensitive to forward and backward head movement as it should be, because ear distances to WATSON and ORION change. There are no signs of phasiness or discomfort like I have experienced with DSP based crosstalk cancellation schemes. The SEL system is easy on the brain and untiring, if the recordings are well done, which means that they preserved a strong relation to what a person might have heard from a single point in the recording venue space.

Connection A-C in the diagram above is useful for finding the proper ratio of ORION to WATSON signal levels by setting Volume 2 for a fixed Volume 1. Connection B-C maintains this ratio and and makes it independent of Volume 1 settings.

If this loudspeaker array were to be used for cross-talk cancellation, then the WATSON on the right should be driven by the left channel signal and vice versa. But if I do so, as an experiment, then the AS becomes narrow. If the level of the WATSONs is increased too much, then the AS reverses left and right sides, as expected. This is not correct and points out that SEL is not a CTC scheme, which requires cancellation, but it is a left and right channel signal enhancement scheme increasing D/CT and D/R. Digital Signal Processing based CTC for a single pair of loudspeakers cannot increase D/R for the listener and room. CTC actually injects cancellation sound streams into the reverberant field, which contaminate the reverberant field and which the ear picks up too. This is why CTC works best in very dead rooms or in the anechoic chamber. SEL overcomes reverberant spaces to a high degree at least for the listener in the sweet-spot. 

Zero crosstalk comes automatically with headphones, but there are the issues of headphone/cup frequency response and head turning. Zero crosstalk must actually be counter-acted with cross-feed to achieve naturalness when reproducing recordings that are far from binaural recordings. 

Thus I have come to call WATSON a "Stereo Enhancement Loudspeaker", SEL.  An electronic time delay, an accessible volume control and a stereo power amplifier are needed for its application in a prescribed loudspeaker and listener setup.

WATSON-SEL Construction

Construction of the SE Loudspeakers is simple enough that I will not provide plans. The Schedule 40 4" ID pipe and 4.5" ID coupler, or something similar, should be available at your local hardware store. The base of the pipe is formed by a Closet Flange Hub and a 10" x 10" x 3/4" piece of plywood, which also seals the pipe. The pipe is filled lightly top to bottom with Acoustastuf to attenuate standing waves. I used Poron rubber tape with adhesive backing to form a seal and flange for the driver to sit on. The driver is not screwed in. It is heavy enough to just sit on top of the coupler. Be sure that there is air passage around the large magnet on the inside of the coupler.  

The domestic acceptance factor for WATSON is probably even lower than that for PLUTO, because of its placement in the room. I look at them as being my favorite seat in the house for listening to a concert. When not in use the WATSON SEL and Tracker-Pre will be placed to the side of the room and the chair turned around for a normal living room configuration. For use at a mixing console the SEL should probably be further back or mounted to the chair in order not to interfere with the mixing desk. I think this is possible but requires investigation of the array radiation pattern, which I plan to do. 

Have fun with this project and the enjoyment it will bring you!

Updates

7 Feb. 2012 - I am now using a Behringer DCX2496 to experiment with different amounts of delay in setup B-B. I measure a signal delay of 0.85 ms for the DCX2496 that must be added to its displayed delay in order to get the actual delay through the unit. Gain is set to +6 dB. Volume2 is controlled by a dual log-potentiometer connected to the output of the Behringer.
I am investigating delays for WATSON, which are 500 us or 250 us less than my previous 5.8 ms selection. For comparison, the distance between the ears is about 17 cm or 500 us. The WATSON output arrives at the ear earlier than the ORION. 

26 Feb. 2012 - Upon further listening and investigation I concluded that the new SEAS FU10RB driver has sufficient output capability for my SEL application. It need not be mounted in a small box as above for high frequency extension but can be mounted at the top of a pipe for omni-directional operation in the horizontal plane. This new SEAS driver was shown at RMAF 2011 and should become generally available soon. The small diameter and lightweight column, which holds up the driver is less obtrusive and more easily stored out of sight when not in use than previous versions of WATSON. The SEL should enhance any stereo speaker pair, not just the ORION.

In the ITD frequency range of perception between about 200 Hz and 1 kHz WATSON expands the stereo stage by increasing the delay between the left and right stereo signals at the ears. Phantom source perception is derived from the group delay difference between the ears as a result of amplitude differences between left and right loudspeaker outputs. Left and right eardrum signal levels are equal, there are only timing differences between the ear signals! These timing differences are proportional to the timing differences caused by sound from a real source at different angles in front of the listener. As the signal level for WATSON is increased, the ITD is also increased and phantom sources are spread further apart and increased in 3-D size. The subtended angle of approximately +/-650 for WATSON tends to a maximum ITD of 525 ms when WATSON dominates. The +/-300 angle for ORION generates a maximum 260 ms ITD. Output from WATSON has to be kept low so that it only introduces a spatial effect and is not detected as a loudspeaker. The effect may seem subtle at first but then the brain constructs a much richer AS and desires more realistic volume levels that are appropriate to the transformed space. The effect varies with recordings and therefore a volume control is desirable. Unlike with DSP cross-talk cancellation, which primarily operates in the ILD frequency range above 1 kHz, no harm is done to the loudspeaker and room reflected signals in order to improve the information content of the signals at the ears. It was thought necessary to delay the drive signal for WATSON so that ORION and WATSON outputs arrive at the respective ears at about the same time. A 17.5 cm spherical head model analysis confirmed the subsequent ITD increase. It was pointed out to me and I confirmed that the delay is not necessary to increase spatiality without losing focus. Thus I am now listening with zero delay and I also removed the 2 kHz roll-off. It appears that with the early arrival of the low level WATSON signal at the ears, 6 ms ahead of the ORION signal, the law of the first wave front - The First Sound - comes into play and a wider bandwidth signal is tolerated without noticing WATSON.  I will continue to study the psychoacoustics that are at work here by devising spatial test signals. It is very difficult to draw conclusions from the wide variety of recording styles, microphone configurations and mixing choices. It is clear that the SEL reveals a lot about recording practices. It also can lead to greater appreciation and enjoyment of program material. I strongly recommend to try WATSON and to let yourself be surprised how much more information there is to be enjoyed in familiar recordings. And your listening room and speakers are out of the picture.  This must be the simplest and most rewarding loudspeaker construction project that I can think of. WATSON-SEL Construction: SEAS FU10RB driver glued to the top of a 3" pipe coupler with GE Max5000 Siliconized Acrylic Caulk (clear). The driver rim is at a height of 34.5" from the floor for my chair and seating. The 3" SCH40 ABS pipe is sealed at the bottom and filled with 110 gram of Acoustastuf. The base consists of a 10" x 10" plywood square and a 3" Closet Flange Hub. A passive volume control works for connection A-C above. For connection B-C a preamplifier with gain is necessary to guarantee more than adequate output volume from WATSON. A 2-channel receiver with remote volume control could do the job perfectly.

3 Mar. 2012 - An in-room measurement of the pink noise SPL at the head location and in the 500 Hz octave band, indicated a WATSON SPL about 2 dB lower than the SPL of ORION. This is the correct level for many recordings, but some recordings image more believably with a lower level from WATSON. The measurement does not reflect the higher ratio of direct sound to room reflected and reverberated sound in favor of WATSON. It is a significant contributor to hearing the recording venue and not the listening room. 

When WATSON is moved further away, like at arm's length with stretched out fingers, then the input level for WATSON has to be increased. This could be the situation when listening while seated on a couch. The spatial effects remain, except that the AS cannot be spread as widely to the sides. Thus, the placement of the SELs is not critical, provided they are significantly closer to the ears than the main speakers and their subtended angle is larger.

4 Mar. 2012 - An analysis of relative signal levels from ORION and WATSON in my living room provides further insight into the contribution of the Stereo Enhancement Loudspeaker to spatial perception. With a room volume of 137 m³ and a RT₆₀ of 450 ms in the midrange, the reverberation distance, or critical distance, becomes 1.7 m for the dipolar ORION and 1 m for the omni-directional WATSON. Thus at a 2 m listening distance for position A the direct sound from ORION is 1.5 dB below its reverberated level. The direct sound from WATSON, which is placed 0.6 m in front of A, is 1.5 dB above its own reverberated level. The small circles in the graph below indicate the distance from the ORION, where direct sound and reverberated sound are equal for ORION by itself and for WATSON by itself when placed at 1.4 m and 3.4 m from ORION.

Direct and reverberated sounds combine on a power basis due to their random phase relationship. At listening position A the total sound level generated by ORION is about 2 dB higher than the total SPL generated by WATSON. The two direct signals differ by only 0.5 dB. I found that many stereo recordings were spatially enhanced at such relative sound levels.
At listening position B the direct sound from ORION is 7.5 dB below its reverberated level. The optimum total WATSON level is about 1.5 dB lower than the total ORION level without WATSON. In that situation the direct sound from WATSON is 4.5 dB higher than the direct sound from ORION. WATSON works well for the larger distance B and actually is run at slightly reduced level.

WATSON-SEL functions by increasing the direct signal level sound streams relative to the room reverberated sound in the 100 Hz to 3 kHz range. This makes the recording venue space more audible, if the two stereo tracks contain such information. The larger subtended angle of more than +/-30⁰ increases ITD and spreads the Auditory Scene laterally, making it bigger and asking for a proportionally larger volume level from ORION. Delay of the WATSON drive signal is not necessary, because any combing and cancellation between the direct sounds from WATSON and ORION is masked by the room reverberated sound and not perceived. 
It seems important that WATSON is omni-directional with a smooth frequency response so that its reverberated and reflected sound field is not colored by the speaker. The same applies to ORION with its dipole radiation. The overall effect is that ORION, WATSON and the listening room essentially disappear from the Auditory Scene, which makes it mandatory to listen with closed eyes in order not to destroy the illusion.

Related material and references:

[1]  WATSON page of the spreadsheet modes1.xls for calculating reverberation distance and direct signal levels at the listening location.

[2]  Room Acoustics

[3]  Brad Rakert, William M. Hartmann, "Localization of sound in rooms. V. Binaural coherence and human sensitivity to interaural time differences in noise", J. Acoust. Soc. Am., Vol. 128, No. 5, November 2010

14 March 2012 - After more experimentation and listening, also using PLUTO as the main speakers instead of ORION, I have come to some conclusions about the applicability and effectiveness of WATSON. The essential parameter appears to be the ratio of the main speaker's direct-to-reverberated sound in the room, whether from ORION or PLUTO. If the main speaker's direct signal level is more than 6 dB below the reverberated level, then WATSON will still increase spaciousness, but imaging precision suffers. Also, and even more importantly, it becomes more difficult to find a level setting for WATSON, where it is not recognized as a separate source from the main speakers, while at the same time providing spatial enhancement.

The direct-to-reverberant sound ratio D/R is inversely proportional to the ratio of listening distance d to the reverberation distance Rx, such that  D/R = 20 log(Rx/d)    dB D/R = 0 dB for d = Rx D/R = -6 dB for d = 2 Rx According to observations, the listening distance d should not be larger than twice the reverberation distance Rx.  Rx depends upon the reverberation time T60 and the directivity gain G of the loudspeaker over a monopole. Rx = Rm sqrt(G) Monopole G = 1:  Rx = Rm Dipole G = 3:  Rx = 1.73 Rm   Conventional box speaker, G = 1 below the baffle step, G = 10 and more at high frequencies The reverberation time T60 depends upon the ratio of room volume V to total room surface area S and to the average surface absorption as in the graph on the left from modes1.xls.

It shows the increasing difficulty to obtain reverberation times below 300 ms. An average absorption of 50% is equivalent to 50%, or half the surface area of the room behaving like open window area, where sound escapes and is not reflected.

The reverberation time T60 is best measured, for example with ARTA test software. With some experience it can also be estimated by looking at the room's furnishings, carpets and wall hangings, since V/S is easily calculated from room size measurements. A measurement of T60 in a domestic size room is meaningful because WATSON operates in the frequency range above the Schroeder frequency where the room behaves statistically. WATSON is of no use in the modal region of the room. Once T60 is known the reverberation distance and the direct-to-reverberant sound ratio can be calculated. It is interesting to note the value for the average surface absorption for future reference. 

Table F on the WATSON page in modes1.xls shows whether the chosen listening distance d from the main speakers will be advantageous for WATSON. 

In general, and also without WATSON, the listening distance should be less than 2*Rx to obtain the best imaging performance from stereo speakers in a given room. At greater distances the room sound begins to dominate. Conventional box loudspeakers with their directional gain variation from 1 at low frequencies to 10 and more at high frequencies, imbed this behavior into the reverberated field and are therefore not frequency neutral transducers. Between low and high frequencies the reverberation distance changes from Rm to 3.16 Rm and the direct to reverberated sound ratio increases 10 dB for flat on-axis, free-field frequency response. This makes the changing high frequency balance problematic and tends to make the loudspeaker more noticeable as such in the Auditory Scene. Frequency independent directivity, as realized with an acoustically small monopole, dipole or cardioid loudspeaker, is necessary for neutrally rendered sound under reverberant conditions. A benefit of increased directivity is either an increased listening distance for the same direct-to-reverberant sound ratio, or hearing less of the room by having a larger D/R ratio at the listening position. When listening to music in a concert hall, D/R changes all the time because different instruments have different directional characteristics. Some excite the hall's reverberation more than others. This is typical. A loudspeaker that changes D/R with frequency in the listening room acts like a musical instrument of sorts and not as a neutral transducer.

==================================================================