SP15Development

This recounts the process of developing the first Parallel Audio loudspeaker.

Application

The intended application is surround and center channel speakers for multichannel audio, home theater, SACD, and DVD-A.

Designing a speaker is first about defining its usage. Second about making trade-offs. There is no ultimate perfection.

After four years of playing around with 5.1 home theater audio I've made numerous missteps. Mislead by marketing and my own ignorance I've gradually gained enough knowledge to be dangerous. I started my home theater with some wall mounted dipole speakers for the surrounds. They were $450 a pair and sounded like a cheap portable radio. Those didn't last long. They were replaced with $1300 a pair quasi ribbon hybrids that were a bookshelf version of the stereo mains I was using at the time—overkill because the demand on surrounds, for either movies or music is modest—but they worked fairly well. They had good horizontal dispersion (yet extremely limited vertical dispersion) providing a wide soundfield. These were not optimal for surround use, but better than most. I sold the mains and later I decided to sell the surrounds too. In their place I put a pair of older speakers. These older ordinary 2-way floor standers did not suffice. Their dispersion was too directional—definitely not good for surrounds. So, I needed to find a solution, one that did the job right and that didn't cost thousands of dollars.

Keep in mind I am neither an electrical nor audio engineer. I don't claim to be an audiophile, tweaker, or audiofool who changes equipment with the seasons. Most of what I've learned is available on the internet. It does require a good deal dedicated searching. It is extremely important to find corroborating sources that don't reference each other. There is so much half-baked information floating around in the audio world, most of it based on many common misconceptions, insidiously faulty logic, specious pseudo-science, or over simplification, that digging for several different sources is absolutely crucial.

Good surround speakers need to create a diffuse soundfield. Listeners should not be able to easily locate the speakers. The sound should wrap around the listeners and envelop the listening space, filling the room with a non-directional, or more accurately, multi-directional sound field. Surround speakers do relatively little work on most movies, and often not a lot more on music. They don't need to be full range nor capable of very high SPLs. However, they do need to provide clean, low distortion sound at moderate to medium-loud levels and spread the sound widely and evenly. When this is accomplished, the surround channels will barely be locatable. You'll hardly be able to tell where the sound is originating.

THX specifies dipoles for surround speakers. They have the advantage of a bi-directional (dipole) dispersion pattern, and when the null is aimed at the listening area, the ability to 'disappear' so that one cannot easily locate the speaker. The sound emanates forward and backward, wrapping around the listener rather than coming from a specific point in space. Omnidirectional speakers also do this well, in fact, very well. Omnis have an even wider dispersion pattern, and with wall reflections can also disappear, in fact disappear to the point that at times the ear is fooled into thinking the sound is coming from somewhere else other than the speaker. (No speaker can disappear completely. Our ears are extremely good at locating the source.) If designed as a single driver speaker, this can simplify construction, lower cost, and allow more to be invested where it will deliver the greatest return : in the driver.

Design Parameters

The design goal : An omnidirectional, single driver speaker capable of continuous 90 dB SPL with low levels of distortion, a frequency response of +/-3 dB from at least 80 Hz and up to 10 kHz, with no crossover or equalization. (Fundamental frequency of adult spoken voice is approximately 85-255 Hz, vocal bandwidth including harmonics extends to 4-5k; the absolute minimum necessary range for center/surround.) Many surround and center speakers only extend down to 100 Hz, some of the little satellites may barely do 200 Hz. It is clearly insufficient for good vocal reproduction to rely on the subwoofer to produce the fundamental frequencies. Furthermore, it hinders localization cues if the fundamental is originating from one location and the harmonics from another. Very low frequency sounds, below 80 Hz, become more difficult to locate, so origination from the same point is somewhat less critical. (However, there is no exact frequency at which the ability to locate low frequencies stops and it can vary greatly under various conditions.) Other important criteria are, a modestly sized, rigid, low resonance, sealed enclosure (for critical damping and bass definition), simple construction from easily sourced parts, and reasonable cost. Eliminating the need for separate stands or wall mounting by making use of the enclosure as the 'stand,' and by doing so, the enclosure places the speaker at the correct height for seated listening and dispersion. The footprint is to take no more space than that of the little cubes or other diminutive box satellites on their stands. With these design restrictions, the speakers can be easily moved or repositioned, unlike permanently wall-mounted speakers.

III

Research & Development

Little did I know, these seemingly reasonable and modest requirements are not as easily achieved. This explains why most modestly priced surround or satellite speakers are so poorly designed and severely compromised. They may look good and superficially seem convenient, but in practice fall far short of their marketing promises.

Starting with a search for appropriate full range drivers, I went through dozens of possibilities, assessing driver specifications, T/S parameters and their applicability to the design goals. Most full range drivers are designed for ported, transmission line, or back loaded horn enclosures. Most are also 6.5” or larger. Larger drivers generally have lower free air resonance and better LF (low frequency) response, but also narrower dispersion in the HF (high frequency), plus they don't meet size requirements for the design parameters set above. Smaller drivers have better HF dispersion, but higher Fs (free air resonance). Finding a small 4” or 5” driver with a low Fs that is suitable for sealed enclosures greatly narrowed the field. For achieving an omnidirectional dispersion pattern the driver will be mounted in an up-firing orientation. This means the listening position will be at a 90° angle off-axis and the HF will be dropping off considerably at greater than 30°. In order to avoid equalization of the high frequencies, a driver that has a HF response extending above 10k and tilting up through the HF is also needed. These requirements brought the choices down to five possibilities ranging in price from a low of $25, to over $250 each on the high end.

Prototype 1

For the first prototype, I chose the least expensive driver for two reasons. One, this was my first attempt at designing a loudspeaker. Two, to keep the initial cost low. After all, it was an experiment that might end in a bust. If things worked out reasonably well, then I'd try other drivers for comparison.

The concept for these speakers is based on the ambient recovery surround speakers design by Siegfried Linkwitz of Linkwitz Lab and Linkwitz/Riley crossover fame. His design used 4” PVC pipe for the enclosure and a 5” mid-woofer for the driver. The driver was a premium line Vifa, but it wasn't a full range driver. The frequency response starts dropping in the upper midrange and plummets above 8 kHz requiring considerable equalization to bring up even the low-HF. The 4” pipe has a volume of about 7 liters, relatively small, and needed an 8”x8” wood base for stability. The design was ingeniously simple. It demonstrates how low the demand is on surround speakers for they proved satisfying in their ambient recovery role. Using his design as a springboard, I wanted to see if it could be simplified further, and maybe at the same time, improved.

I chose to use 6” PVC pipe instead of 4”. The added width and weight would provide stability without the wood base, and more importantly, double the enclosure volume allowing for a lower Qtc. (Lower Qtc improves damping, and LF extension.)

Next, parts availability needed to be assessed as well as calculating Thiele/Small parameters for the enclosure. I set up a spreadsheet for calculating the properties of different drivers in different enclosure volumes. Many hours were spent weighing advantages and disadvantages of different drivers and enclosure sizes. After determining the enclosure size and parts needed, a detailed drawing was made using a simple CAD program. Once all the pieces seemed to be in place, the parts were purchased, and the process of logically stepping through the construction and testing began.

The first prototype was encouraging and disappointing at the same time. The LF extension was better than expected. A good 50 Hz was produced (in room measurement—the calculated F3 was 77 Hz), but there was a dip at 63, then the expected nearly flat at 80. At 90° off-axis the HF was also much better than expected. It rose through the lower treble, was near flat at 6.3k, dropping some at 8k followed by a continuous roll-off, however, an easily audible 12k was produced by the driver, then by 16k virtually nothing. Not a problem because the design goal doesn't require the top octave.

Listening to music proved mixed. Jazz and pop sounded okay, orchestral music was terrible. The overall sound was clean under 80 dB, but by the mid 80s, audible distortion was evident. This driver is definitely not capable of low distortion at 90 dB SPL. The midrange seemed to suffer the most, and the elevated 2k-6k region, even at 90°, made them sound bright, hollow, and screechy. I started with only 2 ounces of polyester fiber in each tube. Doubling it helped to smooth out the irregularities in the LF, but not enough to balance with the HF. Perhaps still more internal damping might be needed to absorb the back-wave and the internal enclosure resonances.

Sand is sometimes used to dampen enclosures. I tried adding about 2 kg of sand to each speaker. Upon listening I heard no immediate difference and so, no new measurements were taken at the time. Containing the sand was a challenge. Using nylon stockings to allow air flow proved only marginally effective. It contained most of the sand, but the finest particles sifted through. Scratch the sand.

Next try was rigid fiberglass. It's not flexible enough to bend into a nice cylindrical shape to fit in the PVC, so cutting several narrow strips to fit in the inner circumference worked, sort of. I didn't notice a drastic change in the sound, but there was some difference. Measurements showed a very similar response indicating that still more experimentation with damping materials was needed.

More listening at various levels confirmed initial impressions. The HF emphasis made instruments and voices sound bright and nasal, but also the lack of lower midrange/upper bass authority would need to be remedied to avoid the cheap small speaker sound. I began to think that maybe the logic of a rising HF response is counterproductive; that the midrange is far more important than HF extension; and that a very solid upper bass (80-160 Hz) is critical. Surround channels carry full bandwidth information, the more they can deliver the better, yet most of that information is in the 80-5k range. Rethinking the driver's response, along with room interactions, lead me to realize that the rising HF may help compensate for the HF loss in the off-axis listening position, yet on the contrary, it creates a very bright reverberant soundfield, and because of the dispersion pattern, we hear a high ratio of reverberant to direct sound. That overly bright reverberant field makes the composite of direct and reflected sound very unbalanced even if the HF 90° off axis response is attenuated. Balance is the key to any speaker system, and the reverberant field, created by a speaker's dispersion characteristics, contributes greatly to the balance.

I went back to the drawing board for a second look at drivers. I decided to reevaluated priorities and find drivers with a different set of characteristics; larger diameter, for stronger low end and a more solid midrange; flatter HF response, for a more natural and balanced tonality; and higher quality, for lower distortion and higher power handling. A closer look at other drivers brought two promising, and not too expensive alternatives to the forefront. I ordered a pair and the necessary parts for the next prototype.

Prototype 2

The larger driver proved to be a significant improvement, offering much lower distortion, less emphasized HF, and far greater volume displacement. However, there were still problems. Although the HF was better and the LF potential far greater, there was still a lack of upper bass. Trials with different damping materials and quantities still didn't solve the dip in response around 100 to 200 Hz, a critical octave. Obviously, still more research into the damping material was needed. The larger size of the driver necessitated using a 6”x 5” PVC reducer; hard to find and more than twice the price of a 6”x 4” used on prototype I. This reducer also had other issues. The more rounded ends were a problem for mounting and sealing, plus a 5” pipe insert was needed for mounting the driver. Five inch pipe is nearly impossible to find. I could not source it. I reluctantly decided, despite my original intentions, that a wood mounting baffle would be the only solution. The search for a wood working source that had a computer controlled cutting machine (known as a CNC, Computer Numeric Control) only took a few phone calls. After describing the part I wanted, I got a quote that surprised me. It wasn't nearly as much as I had expected, and far less than the cost of the 6”x5” reducer. Additionally, the wood baffle turned out to be more aesthetically pleasing than the ugly reducer, and allowed lengthening the 6” pipe for more internal volume—two advantages. Encouraged, I drew up the baffle, redesign the enclosure, and proceeded to work on a center channel version also.

Prototype 3

The baffles turned out perfect. MDF was chosen for its lower cost, and ease of machining for cleaner, more accurate cuts. I also ordered 2 yards of industrial felt. Research indicated that felt is a excellent damping material. Two new prototypes were constructed. One lined with felt on the inside only, IIIa, the other with both inside lining and an outside felt jacket, IIIb. Measurements showed no difference in the frequency response, but there was an audible difference with the addition of the outside jacket. With a blind A/B comparison between the two, the one with the outer jacket seemed to have less audible pipe resonance.* There was a distinctly smoother sound from the jacketed version. High frequencies were rolled-off equally on both, but the 'warmer' sound of the jacketed speaker made for a more balanced presentation. Yet, another questioned arose; was the inner felt liner beneficial? Further tests indicated that taking out the inner liner made no measurable difference. Blind A/B listening test with music also proved no distinguishable audible difference, therefore the inner felt liner is not needed and the cost savings significant.

(*The PVC pipe is extremely rigid. There should be no wall resonances, unlike the flat panels of a box. So, theoretically, the external felt shouldn't be making any audible difference. What I suspect is happening, is not cabinet resonances or sound penetrating through the walls, but rather the reflected sound off the hard exterior surface of the PVC. The external felt absorbs and reduces the unwanted reflections.)

Tests also revealed, much to my surprise, that there was little difference in the frequency response between the taller 36” version and the 27” center channel. The upper bass dip was slightly higher with the 27”, yet the mid bass was within 1 dB from 80 Hz down to 50 Hz. Voices were clear on both—a very pleasing finding which means the 27” center channel would perform well as a center speaker for dialog. Its response and sound was very close to the 36”.

Compared to prototype I, the IIIb is much fuller sounding and tonally better balanced. It is far less bright, and doesn't exhibit any shrillness or distortion when pushed to 90 dB. Orchestral music is still a challenge for the IIIb as expected. After all, it is only a single driver.

I was still not happy with the upper bass dip. I was about to give up and considered the need for equalization until I tried measuring the response near a wall and in a corner. In practice, the speakers will be placed in such a configuration anyway. Corner placement, with proximity reinforcement, smoothed the upper bass more than expected. In fact, it nearly cured the problem. At 12” from each wall, corner placement, it added too much bass and created an unacceptable 'boxy' sound. At 24” there was not enough boost, but at 16” it was reasonably flat. (Amazing how just a few inches makes such a difference.) I still preferred the more open sound further out from the walls, but the response at 16” from each wall of a corner worked. Considering that surround speakers are usually placed near walls or corners, this is actually better from a practical stand point. A speaker with perfect anechoic chamber response may have too great an emphasis in the mid to upper bass when placed near a wall.

More experimentation also proved that too much internal stuffing hurt the bass response, especially that vexing upper bass dip. With too much stuffing, the sound was also muffled; too little stuffing, audible pipe resonance. With the proper amount, the upper and mid bass were better. Corner response stayed nearly the same, but this time, the flattest bass was achieved with the speakers out in the room rather than in the corner, confirming my preference for the sound with more delayed wall reflections. Actually, corner placement was a trade off between mid and upper bass, mid bass better in the corner with a sacrifice of the upper bass, and vice-versa. Since surrounds are commonly placed near walls, their sound will not suffer much, but more open placement is better if one has the space. The optimal amount of damping material clearly helped the bass issues and resonances.

I had to keep reminding myself that these are small speakers for surround and center use. They will never be suitable as mains for music. Considering their purpose, they do a very satisfying job, and better than most commercially produced speakers. Their sound is clean, power handling very good, transients and dynamics are excellent (no crossover network hindrance), and they nearly meet my original design specifications with better bass extension, 50Hz compared 80, yet a slightly more rolled-off treble, -3dB at 8k instead of 10k. All things considered, they are pleasingly close to original design goals. Sacrificing a little treble extension for more than a half octave of bass extension is a beneficial trade-off that results in better mid to upper bass and allows a lower crossover point to the subwoofer.

The cost went up more than I was hoping, but still reasonable. The 5.5" driver is not cheap, nor as expensive as others. Maybe a lesser expensive driver could do the job, but I haven't been able to identify anything more moderately priced of comparable quality. (See update.) The felt really bumped up the cost, but its contributions were more than acoustic. It eliminated the need to paint the pipe, and it looks great. I am even more please with the aesthetics than I had hoped. Not only is the sound clean, but the looks are clean too.

This entire project has been very humbling. Speaker design is quite complex and far more difficult than might be expected. Despite all the science, there is still a good amount of art (really trial & error) in designing speakers. Frustrating trade offs are unavoidable. Not only are there hundreds of hours of research and experimentation that go into a well designed speaker, every little step to improve something increases the cost. Balancing the response and controlling the back-wave is a formidable task. And keeping aesthetics in mind adds another dimension of complexity and expense.

(II) UPDATE: The 5.5" version of the mid-bass driver used in the SP18M is the answer. Not only is it half the price, its overall performance is better, with equally low distortion, lower Fs, reduced 5k peak, and lower Qs.

Conclusion & Critique

After hours of extended listening with many different recordings and types of music, I find these speakers meet my goals. With most studio recordings and pop music they sound good. Unless you want pounding bass, their 50 Hz is adequate. More importantly, the bass is very well defined, about as good as any speaker I've heard. They can handle a soprano belting, and brass screaming without adding their own edginess. Very high frequencies are rolled off making their sound a bit less 'airy' than would be if the top octave were there. Still, cymbals and bells sparkle nicely. Sibilance is neither exaggerated nor muted. Their sound is well resolved and reasonably neutral. With most recordings they are respectable.

Two types of music really reveal their weaknesses, solo piano and orchestral music. The generally rising response through their bandwidth leaves the upper bass of the piano relatively weak. Orchestral music challenges any speaker, and these are no exception. They just can't reproduce all the forcefulness of a symphony orchestra. It's just too much for a single driver to handle.

Vocals are quite good and realistic. A cappella soprano/alto/tenor/bass sounds well balanced. Soloists are easily understood. For voice, the sound is lifelike, spoken or sung. Listening to dialog on movies with the 27” center is distinct, male voices are full, and female are natural, neither chesty nor nasal. It should be noted that in the past, I have been satisfied without a center channel, relying only on the phantom image created by the stereo mains. In fact, I much prefer the phantom image to a center speaker that is too far below or above the screen which makes the voices sound disconnected from the image, or one with too little bass response that makes voices sound unnaturally thin. However, having a well placed center without a doubt improves the cinema experience. The 27” height was chosen to position the speaker right at the lower edge of a rear projection screen or just at the bottom edge of front projection. The up firing driver reflects off the screen to help create the illusion of the dialog coming from the screen, as it should, though still not perfectly centered with the image. There's no solving that unless the center is placed behind an acoustically transparent screen. I haven't tried, but I wouldn't recommend placing the 27" center behind a sonic transparent screen. A direct firing center would be best for that arrangement. (The SP18Mc is designed for behind screen use.)

A listening technique I've found helpful, if not revelatory, is to leave the room, completely out of hearing range for several minutes while the speakers continue playing. Later, go back in to listen with 'fresh' ears. It's amazing how different things may sound. Your hearing will be far more objective. You'll learn how quickly and easily our ears adapt to deficiencies in the sound, particularly linear balance. It's similar to our eyes adjusting to the color balance between daylight and indoor florescent or incandescent light. When we are listening to only one pair of speakers, our ears begin to adjust to the tonal balance between low, mid, and high frequencies. If the bass is a little weak or the treble a little accentuated, we'll soon do a psychoacoustic re-balancing and adapt. (This is likely what happens during the alleged break-in period many manufacturers claim their products need. With time we start hearing what was at first missing and then get used to the maladjusted voicing of a system.) Every time I do this with these speakers, I notice their weakness in the upper bass and lower midrange. Interestingly, I'm not at all bothered by the rolled-off top octave. This reinforces the importance of a solid low midrange for presenting full, natural, balanced sound. It's the only weakness that recurrently reminds me that they are simple single driver speakers.

Their strongest attributes are soundstage, imaging, and best of all, the wrap-around dispersion. Close your eyes, and with a good recording, the musicians will be floating in space right in front of you. The omnidirectional dispersion creates a believable soundstage. When I finally set them up as surrounds, rather than listening to them as a stereo pair, my reservations about the upper bass/lower midrange evaporated. They blended completely with the main front channel speakers and took on their duties as ambient speakers without attracting attention. There is little coloration that would make them separate from the mains to create a conspicuous mismatch. They are neutral enough to allow for good integration. I could hardly noticed they were working unless I switched back and forth between stereo mode and 5-channel mode. My expectations are based on full range high performance speakers. It is unrealistic to evaluate the SP15 as stereo mains, as I have done. It's not their purpose, and far too critical. When they are functioning in their designated role as surrounds, they meet the objectives set.

(II) UPDATE: The SP15 blended very well with the Piega C8 LTD that I was using for the mains. After the C8s were replaced by the SP18M, the blend was not as good. The C8s have a peak in their response in the mid-treble. The rising response of the SP15 'matched' that peak and balanced well, making the comparison between stereo and 5.1 mode nearly inaudible with the C8s. The SP18M, which is linear throughout its entire range, reveals the rise in the SP15's response. Although this is not noticeable on movies, and definitely not an issue for the center dialog channel, on music the imbalance is exposed. For critical multichannel music applications it would be advisable to use the SP18M for the center and surrounds. Considering the limited availability and popularity of multichannel music recordings, perfect matching is not yet a high priority.

back to the Development page

on to the Project SP18M page

(|| ) Audio Home