----- Experience in your own room the magical nature of stereo sound -----

  ---   Photos/Comments  ---  DSP challenge  ---  System  ---

Ever since I started building loudspeakers with cone-type drivers on open baffles several decades ago, I tried to find the optimum shape for those baffles given the drivers available. For ORION the drivers were selected primarily for their volume displacement and low distortion capabilities when used in a 3-way, active system. The LX521 evolved from experimentation with minimal width baffles, which can provide a more uniform dipolar radiation pattern at higher frequencies, if also suitable drivers are available. Inevitably this leads to a 4-way design, which I mostly tried to avoid in the past.  The shape of the midrange/tweeter baffle was arrived at empirically for the chosen SEAS drivers and after many acoustic free-field measurements. The two SEAS woofer drivers are housed in a V-frame baffle, which exhibits reduced resonance above the operating range of the woofer and some force cancellation.  A bridge over the woofer isolates the midrange/tweeter baffle from woofer cabinet vibrations. Woofer and midrange/tweeter baffles can be angled independently from each other. The bridge could also be built taller and the midrange/tweeter baffle tilted downwards to aim at the listener in front of a mixing desk. The woofer baffle must rest on the floor for ground plane reinforcement of its output. The LX521 Analog Signal Processor splits the broadband line level input signal into woofer, mid and tweeter frequency bands. It equalizes driver and baffle response for each channel and filters it with LR4 response. The midrange signal is split after the power amplifier by a passive crossover filter into lower mid and upper mid driver inputs. Frequency response on upper midrange axis and horizontal dipolar response in the frontal hemisphere are designed for neutral timbre of the the stereo phantom scene, when the loudspeakers are placed in the room as suggested. The aural scene is rendered with clarity and detail, both spatially and tonally.

Rear-view of prototype

Baffle prototype

ASP prototype

The Listening Room

The specifics of the phantom acoustic scene that is rendered by a pair of loudspeakers and perceived by a listener's brain depend upon the radiation characteristics of the loudspeakers, their location in the room, the reflective, diffusive and absorptive properties of the room and the listener's location. The first requirement is for lateral symmetry of the loudspeaker and listener setup with respect to large reflecting surfaces. Secondly, the loudspeakers must be placed at some minimum distance from those large surfaces in order to delay specular reflections by more than 6 ms. This allows the brain to give primary attention to the earlier arriving direct sound from the loudspeakers, if the reflected sound streams have similar timbre and spectral content as the direct sound. This in turn requires loudspeakers with constant, frequency independent radiation characteristics in order to illuminate the room spectrally neutral. The LX521 represents such loudspeaker to a high degree. The typical box loudspeaker is omni-directional at low frequencies and becomes increasingly directional as its baffle and driver dimensions become larger than 1/4th of the radiated wavelength. The power response typically changes by more than 10 dB between low and high frequencies. The power response of the LX521 is 4.8 dB lower at low frequencies and nearly constant up to 7 kHz before it decreases. 

Top and bottom baffle of the LX521 can be rotated independently of each other. The orientation of the bottom baffle determines the degree to which the woofer couples to different room modes. This provides some flexibility in dealing with a problematic room mode by turning the woofer a certain amount. Of much greater significance is the ability to minimize the first side wall reflections L' and R' by orienting the top baffles. The optimum amount of toe-in is a function of loudspeaker distance from the side walls and distance to the listener. In the drawing on the left the right dipole is positioned at 0.5 m from the wall and the left dipole at 1 m. In both cases the plane of zero sound output is pointing at the first reflection points L' and R'. The toe-in axis is at right angle to this plane and points to A or B in front of the listener.  More toe-in is required for the right dipole that is closer to the side wall. Toe-in has to be increased slightly for either dipole if the listener sits farther back. The right dipole is too close to the wall to meet the 6 ms delay between reflected signal R' and the direct signal R, but attenuating R' by proper toe-in greatly improves the precision of imaging in a narrow room. Toe-in also reduces the magnitude of the first reflections L" and R" from the wall behind the dipoles. The level of reverberant sound in the room and the ratio of direct to reverberant sound at the listener are insignificantly affected by toe-in. You can make a drawing for your room dimensions by using your setup distances and corresponding image sources L', L" and R', R'' to determine the optimum toe-in point.  A listener would see the top baffle from the side, if mirrors were placed on the wall at L' and R'.

Richard Taylor has analyzed the required room dimensions and speaker setups to obtain a >6 ms delay for the first order reflections. He also determines the toe-in angle for equal strength of side wall and front wall reflections in a small room with a dipole source . I consider it more important to minimize the side wall reflection and to diffuse, not to absorb, the front wall reflection for optimal imaging.

James Heddle contributed a spreadsheet, which calculates the strength and delay of first order room reflections. Enter your speaker's distances from the walls and adjust the toe-in angle to minimize the reflected energy from the closest side wall. Suppression of this reflection widens the sweet spot and this top baffle orientation places the aural scene between the loudspeakers even for far off-center listeners. Compare the calculated dipole reflections to those of a monopole.

A room that is not open or acoustically dead behind the listener is likely to cause problems due to rear wall reflections. If the distance d_r between listener and wall is l/4, i.e. when the reflected wave has traveled l/2 relative to the incident wave, then incident and reflected waves will cancel. The frequency response at the listener's location therefore has a notch at frequency f = 340/4*d_r . Ideally this notch is at a low 10 Hz, which means that d_r = 8.5 m (28 feet), a very long room. A more common listener-to-rear wall distance of 1.5 m (5 feet) would put the notch around 57 Hz, right into the bass range and be very problematic. Thus every attempt should be made to attenuate the rear wall reflection. Use a variable low frequency sinewave signal source to hear the notch frequency range at your listening place. Pink noise is not a good test signal for this. 

The ideal listening room acts like a waveguide with the loudspeakers at some distance from the diffuse (live) end of the room and sound traveling past the listener to the open (dead, absorptive) end of the room (see the drawing below). Sound reflections from the wall behind the listener should be attenuated as much as possible, particularly below 200 Hz., the frequency range where discrete room resonances tend to dominate the bass distribution in the room. The LX521 being a dipole has the advantage of minimally exciting lateral room modes due to the null in its radiation pattern.

A sound processing room is usually designed to be quiet and acoustically dead. It is a work environment and not at all representative of the typical living space where people listen to a stereo recording for enjoyment but also pursue other activities. The processing room has to be dead to ensure a direct to reverberant sound ratio of no less than -6 dB at the work place and to minimize the influence of reflections and reverberation due to the colored illumination of the room by the typical box-type monitor loudspeakers. Close-field monitors at short listening distance relax the reverberation time requirements. They approach headphone listening. Headphones are completely unsuited for judging the spatial rendering of a stereo recording that is intended for loudspeaker playback. Headphones are optimally suited for analyzing tonal artifacts in a recording but completely distort distance perception. Recording engineers often claim that they "can hear through the flaws of their monitors" to the real sound. Then why are there so many technically poor recordings? They would be well advised to listen to their work/mix via LX521 Reference Loudspeakers.

An acoustically small dipole radiator in a room with T60 = 750 ms will have the same D/R ratio as an omni-directional radiator in a room with T60 = 250 ms. One room is dead, the other very live, but at the same distance from loudspeakers the room contributions to the sound at the listener's ears are equally subdued compared to the direct sound coming from the loudspeakers. The dipole loudspeaker reaches by a factor 1.73 = sqrt(3) deeper into the room. The two graphs below and the Listening_distance.xls spreadsheet show what this means in actual numbers. Note in particular how steeply the required amount of wall absorption must increase to obtain a 250 ms reverberation time. My preference is for a reverberation time of around 450 ms, which also provides a pleasant environment for talking and reading in addition to critical listening. The wall behind the loudspeakers should be diffusive in order not to lose the rear radiated sound from the LX521. Specular reflection from the side walls must not be attenuated as this reduces high frequency energy in the room. The wall behind the listener should be lossy to attenuate room modes. In addition, cloth wall hangings, rugs, pictures, upholstered chairs, open cabinets, plants and other decorative elements are all that is needed to interface a dipole loudspeaker with the room regardless of whether it is intended for work or pleasure.

Example of a 100 m3 volume room for which EBU specifies a 250 ms reverberation time

Listening distance for -6 dB D/R in the 100 m3 volume room for different reverberation times. Required open window area in % of total room surface area.

See also:  Sound Field Control for Rendering Stereo

• Outside dimensions: 49.25"H x 16"W at bottom x 15"D,
  3 parts: Woofer baffle, Midrange/Tweeter baffle and Bridge
• Bridge can be built with increased height to raise Midrange/Tweeter baffle, if needed
• Weight 67 lb est. (30 kg)
• Open-baffle, dipolar radiator, < 20 Hz to 20 kHz,
  Acoustically small in the horizontal plane through the upper midrange
• Response -3 dB at 30 Hz (Q < 0.5) on ground plane, free-field
• Dipolar response over +/-60⁰ horizontal, < 120 Hz to 10 kHz 
  (see Sound Field Control for Rendering Stereo - 2)
• 4-way loudspeaker system with woofer, lower midrange, upper midrange and tweeter
• 3-way ASP crossover/equalizer with LR4 filters at 120 Hz and 7 kHz,
  assembled on two ORION ASP printed circuit boards
• Passive 1st order crossover between lower midrange and upper midrange at 1 kHz,
  located in base of midrange/tweeter baffle
• Tweeters - SEAS 27TFFNC/G, H1396-04, coated textile dome,
  front and rear of midrange/tweeter baffle
• Upper Midrange - SEAS MU10RB-SL, H1658-04, Curv cone
  42" above floor
• Lower Midrange - SEAS U22REX/P-SL, H1659-08, Curv cone
• Woofer - SEAS L26RO4Y, D1004-04, Aluminum cone 
  Push-pull mounted in V-frame baffle of 24"H x 13"W x 15"D
• Minimum amplifier power: 8 x 60 W (e.g. ATI model AT6012)
  60 W for each woofer, lower & upper midrange crossover, two tweeters in series
• Room size: >240 ft² (>22 m²) area,  >8 ft ceiling
• Speaker placement measured from tweeter: 
  >4 ft from wall behind it, >2 ft from side walls,
  speaker separation >8 ft 
• Listening distance 8 ft to 18 ft depending upon loudspeaker application
• Room acoustics: Fairly live with RT₆₀ of 400 ms to 600 ms for a natural living space
• LX521 as Mixing Monitor: - No need for RT₆₀ = 250 ms to obtain a sufficient Direct/Reverberant sound ratio in the room.
  - More articulated bass reproduction due to reduced coupling to    room modes.
  - Neutral timbre and fast time response
  - Low non-linear distortion and high instantaneous dynamic range
  - Optimal spatial rendering of the stereo phantom scene to judge the mix.

Burning Amp Festival 2012

So I gave in and started to experiment with a 4-way driver arrangement and immediately saw promising results. 

I changed the project name to LX21 as this experiment might now become my 21st completed loudspeaker design. After several iterations I found on 5/21/12 a baffle shape that worked well with SEAS CA22RNX and FU10RB drivers. They were combined with 1st order quasi-Butterworth filters, which are 6 dB down at the crossover frequency. Both drivers then add in-phase over a limited frequency range.  Some frequency response irregularities still had to be cleaned up, particularly around 2 kHz.. During my F3 experiments I had investigated the SEAS U18RNX/P.  I liked its smooth response due to a new Curv cone. SEAS provided me with a 8" Curv cone prototype using a short voice coil. I want low voice coil inductance to have low Le(x) and Le(i) non-linear distortion in the midrange. SEAS furthermore changed the FU10's surround damping to optimize the driver's upper midrange response for my application. With two well behaved and capable midrange drivers of different size, but covering nearly the same frequency range, it became feasible to build a very wide bandwidth, 120 Hz to 7 kHz, dipole midrange bandpass filter and keep group delay variation low in its passband by using a 1st order crossover filter around 1 kHz.  Group delay variation relates to envelope distortion, which in a 4-way design can be kept lower in the critical midrange than for a 3-way.

Crossover to the tweeter was first planned for 5 kHz. The crossover has to be steep, because the separation between upper midrange and tweeter tends to be acoustically large, which causes irregularities and narrowing of the vertical polar response. The LR4 crossover affects the frequency response for +/-1 octave around the crossover frequency. Below 5 kHz the dispersion from a 1" dome tweeter is wide and influences the dipole response of the summed drivers negatively. The LR4 crossover was therefore moved to 7 kHz. This also helps to hide front and rear tweeters from each other so that they only interfere in desired dipole fashion at large off-axis angles. This is also the reason for widening the baffle shape around the tweeters.  Above a few kHz baffle design and driver selection become very critical, if one tries to obtain a specific radiation pattern. Consider that a baffle thickness of 3/4" corresponds to 1/4th wavelength at 4.5 kHz or 900 of phase shift between front and rear edge diffraction. Practical baffle and driver dimensions cannot be kept acoustically small at higher audio frequencies and baffle design turns into educated cut and try with lots of polar response measurements. The problem is that drivers are not acoustic point sources, i.e. they are more than 1/16th wavelength in size. This means that for a wide vertical polar response the drivers must be placed as close together along a vertical line as physically practical. For a controlled horizontal polar response the baffle must be narrow and have an outline that optimizes the dipolar off-axis amplitude fall-off with angle for lower midrange, upper midrange and tweeter drivers when measured on their own. Finally the driver outputs are summed with appropriate crossover filters and the overall response is equalized on-axis.

The shape of the midrange/tweeter baffle is the result of acoustic requirements, as is the woofer baffle. The baffles are angular and not hidden behind grill cloth. The acoustic impedance of grill cloth rarely matches the acoustic field impedance near the radiator, causing frequency dependent reflection and transmission loss, which can also be angle dependent. The LX521 is meant to be used without grill cloth with the exception of a light fabric table runner over the woofer baffle to partially cover its front and rear openings.

The loudspeaker's unusual form has evoked a variety of reactions: 
- Interesting, I like it. 
- LX521 recalls the esthetics of Italian futurismo.
- The top looks like a vase.
- Like a lamp on a stand without a lamp shade.
- Aggressive looking, Heavy Metal. I like it.
- My wife is not keen on the shape. Attached is a slightly different design that my wife could live with.
- It's ugly. How can I hide it?

Paint can make a big difference in how dominant a form appears in your room. I like the black silhouette, the red bridge and the angularity of shapes, as does my wife. I envisioned the LX521 as a piece of machinery, an electro-acoustic transducer, a sculpture with acoustic integrity. I cannot help you if the looks prevent you from building them. Sorry.

The two SEAS 10" woofers ended up in a V-frame baffle after I had first tried a W-frame for force cancellation. I did not like the complexity of baffle construction, given my DIY skills for square joints. I also wanted the baffles to use as few wood parts as possible, to eventually provide a low cost flat-pack of parts. The V-baffle also has a less pronounced resonance above the working range of the woofer than a W-frame. Even with the W-frame much mechanical vibration was coupled to the midrange/tweeter baffle when it was placed directly upon the woofer baffle. Therefore a bridge is placed over the V-frame woofer, which detaches the woofer from the midrange/tweeter baffle. 

The frequency response on the upper midrange axis was designed to be flat. The free-field woofer/midrange response was shelved down by 4.2 dB to account for floor reinforcement. When listened to on program material in my room it became apparent that the high end had to be shelved down slightly. I use the same -3.3 dB shelf as for ORION. It merely requires a resistor and a capacitor on the circuit board. The resistor is easily removed to hear a flat response. A flat response is needed when the loudspeakers are used for wave-field reconstruction purposes. For phantom acoustic scene creation, as in 2-channel stereo, a slightly rolled off high frequency response is indicated by the sphere derived HRTF for loudspeakers at +/-30⁰.

The overall result is a speaker that provides a neutral sound, clarity, speed and spatial openness. Whether this is due to the improved polar response, due to inherent qualities of the midrange drivers, or the 1st order crossover, or all of these, I do not know. For its sonic qualities, wide range of applications and an important date in its development, I call it the LX521 Reference Loudspeaker. 

After listening to a great variety of good, bad and so-so stereo recordings I am tempted to call it MAGIC521. The music comes through!