Nuovo1



	To Dream the Impossible Dream..... Developing a DIY speaker inspired by the Wilson Audio X1 SLAMM Jon Mark Hancock Introduction In 1994 when I first read about the Wilson Audio X1 SLAMM, my curiosity was piqued and primed. X1 stands for first eXperimental model, and SLAMM for Super Linear Adjustable Modular Monitor. The upper range driver configuration, with D'Apollito configured 7" midbass/midrange drivers coupled with an inverted dome titanium dioxide Focal tweeter was appealing, while the bass module, with a Focal 12" driver AND 15" driver in each cabinet was clearly over the top- just the sort of wretched excess that I thought would render a speaker project "upgrade proof". After reading reviews of them, I was intrigued and committed to auditioning them. Since auditioning them, I was captivated- and very frustrated. There just wasn't a spare $68K for purchasing these anywhere in my audio budget - or any other part of my budget! Seemingly, end of story. But, at this point, some heretical ideas began percolating in the back of my head. I've built subwoofers using a slot loaded design which requires very low Q drivers, and still had on hand some Audax 13" and 15" professional series drivers. Recently, (ca. 1994-96) I designed and constructed conventional and D'Apollito configured two way systems using Scanspeak 7" kevlar woofers, and gained some experience in optimizing the performance of such two way systems, including addressing issues of dispersion and balancing axial and power response. Hmmm, why not build on the drivers I have on hand and investigate a combination of reverse engineering and new engineering to create as much of the X1 SLAMM experience as possible? Some characteristics of the X1's, such as their cabinet construction from high density phenolic materials, were obviously well beyond my means. Nevertheless, was there a chance that with a lot of work and some luck I might be able to achieve 50% or more of the "SLAM" experience, at a small fraction of the out of pocket cost? Even a target of 5% of the cost of the now $75,000 X1-SLAMM Mk II would give me a target budget of nearly $4,000. After some reflection and calculations, plus more than a little pigheaded determination, I figured, "What have I got to lose?" And so, the quest began, leading me on a voyage of discovery to the creation of an interesting pair of not quite clones of the Wilson Audio X1 SLAMM, which tongue slightly in cheek I've named the X1- SLAX, for reasons which will become clear. Measurement system A major factor in the success or failure of any speaker design project is the measuring system. Up until this time I has used a 1/3 octave RTA measurement analyzer, with fairly good results, considering it does not discriminate between early and late room arrivals, and has relatively limited resolution, both in frequency and amplitude. A project such as this would require detailed impedance measurements, quasi-anechoic driver measurements, and the ability to generate data files which would interface with cross over filter design programs such as Sound Easy and LSPCAD. After researching a number of available products, and consulting reviews, it became clear that the best choice in performance for the money was the CLIO system, so I acquired a CLIO-LITE system for doing my measurements, using it with my B&K microphone. Description of Wilson Audio X1-SLAMM The Wilson Audio SLAMM is a 3 or 4 way dynamic loudspeaker which includes two cabinets for each speaker: a ported bass unit housing a 12" and 15" Focal woofer, coupled with an upper range module housing dual 7" midwoofer/midrange drivers and a primary front radiating tweeter in an MTM (D'Apollito) configuration, and two rear firing "ambiance" tweeters. The ambiance tweeters, the 4th range of drivers, are intended to



	fill in the room power response above 10 - 12 kHz. where the response of the primary tweeter narrows in dispersion, due to the relationship between the wavelength of the radiated frequencies and the tweeter diameter. In actuality, the top "module" is a bit more complicated than this brief description might lead you to believe, being comprised of three separate enclosures housing the midrange drivers and tweeters, which are mounted in a framework which allows adjusting the front to back positions independently. Also included is a separate module for the crossover. The time offset " between the primary tweeter and the 7" drivers is adjustable, as is the offset of the 7" drivers to each other, allowing customization of the response through the crossover region, and facilitating vertical aiming of the response lobe in the crossover region. Depending on your perspective and the measured results, this is either an interesting feature for system tuning, or an opportunity for problems; the reality seems to be a little of both. The crossover appears to follow a classic D'Apollito practice of using an even order all-pass filter; in this case 2nd order, which requires inverting the phase between the midrange drivers and the tweeter. This is how the X1-SLAMM's are configured. The bass module is relatively compact considering the driver area. The port tuning of the Wilson X1 is centered at 24 Hz, as indicated by the impedance curves of the system. With this much cone area and this port tuning, high output with low distortion over a wide range of bass frequencies would be expected, and is delivered. All of this information is available from published reviews, but only represents a starting point for attempting to understand the issues and trade-offs in this design, and how to develop a similar speaker. If you want to build copies of some of the old "classic" speakers of the past, in many cases it's not hard to find some detailed analysis and plans, including wood working, and even crossover schematics. With the X1-SLAMM, the situation is quite different; the only comprehensive technical information available is from published reviews. Understanding these speakers' strengths and weaknesses required obtaining as much
	detailed review information as possible; in this regard, the review in the December 1994 issue of Stereophile Magazine was a key resource for detailed measured performance data. Since it would be impractical to recapitulate their review in it's entirety within this article, I'll instead bring up a few points which caught my eye because they seemed to reflect areas in which a different set of design tradeoffs might optimize the speaker better for my preferences. Your mileage may vary! By discussing my preferences in goals and methodology, I invite you to optimize this or similar projects using your own criteria. Let's take a detailed look at the upper range module first, then examine the bass system and the integration of the two. Upper Module The upper module uses conventional but high quality drivers in several frequency ranges. The upper module reportedly uses two custom Dynaudio 7" drivers with mineral filled polypropylene cones from about 125 Hz to 3 kHz, where they hand off to the primary tweeter, a custom Focal unit with an inverted titanium dioxide dome construction, dual magnets for high efficiency, and ferro-fluid damping



	on the typical Focal 20 mm voice coil. The two ambiance tweeters are titanium dome/surround units with smooth extended response to over 30 kHz; these aren't brought into play until about 10 kHz to 12 kHz, with what appears to be a 2nd order Bessel high pass network. The tweeters appear to be Audax DTI01 titanium dome tweeters. An examination of the response of the midrange drivers shows the gradually rising response from 500 Hz to over 2 kHz typical of a narrow cabinet profile without baffle loading compensation in the crossover network. Somewhat compensating for this is the response roll-off of the 12" and 15" drivers, which is only at 6 dB/octave, resulting in some fill in from these drivers up to several hundred Hz. The acoustical highpass transfer function of midrange drivers appears to be a combination of the 2nd order sealed box roll-off combined with a 1st order high pass network, resulting in a net 18 dB/octave highpass characteristic. The primary tweeter position maybe adjusted front to back, and the vertical alignment of the 7" drivers may also be adjusted, giving flexibility in aiming the typical D'Apollito crossover lobe towards the primary listening position. Some excess energy is visible in a broad region around 1.8 kHz, possibly due narrowing of the radiation pattern from the 7" drivers or other aspects of their behavior. There are measured irregularities in the tweeter axial response at 2 meters which apparently are not due to the tweeter's inherent performance, but rather to reflections from the overhanging midrange module combining with the direct radiation from the tweeter. As setup by factory personnel, the acoustic origins of the midrange drivers and tweeters seem to be aligned closely, requiring significant setback of the tweeter relative to the midrange enclosures. This would seem to indicate that Dave Wilson preferred using a physical alignment of acoustic origins rather than a lattice network delay filter for the tweeter to assure the intended transfer function through the crossover region. Would it be possible to achieve a satisfactory crossover region response while eliminating some of the diffraction errors in the nearfield tweeter response in the version I would build? At face value this upper range module is a fairly sophisticated design, yet there were some aspects which my past experience suggested could be reexamined, and some measurements and subjective auditioning by reviewers raised additional ones or confirmed the ones I held. It's a classic issue that if your design priorities are different, you may prefer different tradeoffs in the secondary choices of what is fundamentally a similar design. One point which bothered me was the use of a plastic cone midrange driver as high as 3 kHz. A point which a friend of mine stressed repeatedly in the development of the early Avalon Acoustics speaker designs was operating as much as possible only in the pistonic region of all drivers. This is difficult to do with cone drivers, particularly with bextrene or polypropylene, because most of them are nominally designed to gradually decouple the center of the driver from the cone edge with rising frequency, and the cone materials are chosen as much as anything else for their self damping (low Q) characteristics. Though this technique can produce an overall smooth response when measured with pink noise, it may show up as some resonance ridges in a waterfall plot, and subjectively may vary between a subtle loss of clarity, to out right cone cry on some program material. This was evident to a mild degree in both measurements of the X1's and in reviewer comments on the reproduction of some program material. On the other hand, the 3 kHz crossover used is probably required by the high goals for power handling and acoustic output that the X1 aspired to. The crossover design choice, which seems to be a 2nd order Linkwitz-Riley down 6 dB at the crossover point, further facilitates this by reducing the power handling requirements for the tweeter and midrange units by 3 dB in the crossover region and beyond. Which raised another interesting question: did I need or want 120 dB from my clone project, or could some ultimate acoustic output capability be traded off to benefit other aspects? Another design choice I questioned was the use of a 2nd order D'Apollito configuration



	with a fairly substantial recessing of the tweeter. Using an even order network in a D'Apollito configuration, the phase response can be matched between the midrange and tweeter throughout the crossover frequency region, and for flat axial response, minimal demands are placed on the tweeter and midrange units, since they are down 6 dB at the nominal crossover frequency. However, the vertical dispersion lobe in the crossover region is relatively narrow. This is looked at by many designers as a benefit, because it helps avoid floor or ceiling reflections in the crossover region, but this is a somewhat questionable benefit considering the dispersion of the tweeter is not so constrained, whereas the midrange drivers in the range below the crossover region still have considerable off axis cancellation until reaching the point where the effective driver diameter of the pair is less than ½ wavelength. Above the crossover point, and below it, the vertical window is not the same, nor is the power response. For flat axial response, there is a dip in the room power response; often a compromise response curve must be used with the axial response slightly elevated at the crossover, and the room response slightly dipped (Bullock filters). In the case of a 2nd order crossover, the affect of the crossover region anomalies extends over a wider region than for 3rd or 4th order networks. Is this the best choice, particularly for closer listening positions than a large room? Now, it's certainly arguable that any crossover is a compromise of one sort or another. My question became, what is the best choice overall for mid to high frequency clarity, flat axial response, and uniform power response? In my case I was interested in optimizing for relatively close listening conditions, say 3 meters. Some speaker designs employ tradeoffs which optimize the total power response over axial response, and this works well for larger rooms and a more distant listening position, where the reverberant room response will be closer in level to the early arrival response. Other problems seem to derive partly from the midrange/treble driver layout. The smoothness of the X1 treble response suffers from diffraction effects from the upper and lower midrange cabinets overhanging the tweeter. This has a substantial impact on the axial frequency response, causing some comb filtering in the measured response by Stereophile magazine. Overall, the power response into the room is fairly smooth, but not what you would usually call state of the art. It's my suspicion that much of the sonic appeal of the X1-SLAM on auditioning stems not from a marvelously flat frequency response (axial or power), but from overall very low distortion, excellent dynamics, and very good transient definition in most of the frequency range. As regards the bass module of the X1 system, it's measured performance indicated little need or possibility for improvement. Very high output levels at low distortion are possible with this speaker. Undoubtedly the low cabinet resonance and excellent audible definition of this system in the low frequencies was due to a combination of the high performance Focal drivers used, and the very rigid enclosure structure achieved with proprietary phenolic laminates by Wilson Audio. The Focal drivers used have different T-S parameters than the Audax drivers I had on hand, and lower nominal efficiency. I suspected that approaching or matching the bass system performance of the X1's with the construction methods I had available would not be easy, if possible at all. Still, I believed that using more conventional cabinet construction methods, a very high performance bass system could be built. How close in performance to the X1's, that remained to be seen. Choosing the Drivers & Configuration When the concept of designing and building speakers similar to the X1 SLAMM first raised it's head, my initial response was that of a typical engineer- try to draw up a list of the performance goals and nominal design constraints, and evaluate quickly if the available resources had a chance of meeting the requirements. In particular, it's important to try to identify early on issues which could become "show stoppers", both to avoid investing in an impossible dream, but also to identify key issues to address early or fundamentally in the design process. Obviously, for building a speaker system,



	selection and matching of drivers is pretty fundamental. The drivers used in the S1-SLAMM are custom built for Wilson Audio. In some cases, similar drivers are available from the same suppliers. However, part of my interest in initially pursuing this project came because I had on hand (read, "no out of wallet expense this month") some high performance 13" and 15" drivers- but ones not intended or purchased for this application. If this economic incentive was to be maintained, then I would have to figure out a design which could use these Audax drivers. The tweeters didn't seem like a big problem area- Audax DTIO1 are readily available, and since they were padded down relative to the main tweeters, efficiency was not a problem. For the main tweeter, the X1's use a special dual magnet version of the Tioxid T120, with ferrofluid damping. A single magnet version of the Focal T120 Tioxid tweeter was available when I began planning this project in 1995 which I hoped would be up to the job. It was not quite as efficient, but was still one of the most efficient direct radiating tweeters on the market. With the addition of some ferrofluid damping to improve power handling, I expected the power handling would be adequate. The titanium oxide coating on the diaphragm makes this tweeter very resistant to breakup modes in the audible range. The Focal T120dx2 became available which seem to be very similar to the one supplied to Wilson Audio. Initially I was convinced that I would use Scanspeak 7" kevlar drivers 18W/8544, or 8545 for the midrange- I have a fair amount of experience with these drivers in MT and MTM systems in the early 90's. However, recollection of my listening experiences with several Avalon speakers, and investigation of the measured data led me to look closely at Eton, and in the end select the Eton 370/32, which as their 7" drivers go, appears to be optimized for the midbass/midrange performance. It combines a moderate compliance with a very low mass cone (13 grams), resulting in a typical Fs of 47 Hz, with fairly high efficiency, typically 89 - 90 dB. Though it doesn't have power handling as high as some of their other 7" models, it compensates by having a very smooth and extended top end- an asset for optimizing the midrange to treble crossover and performance. Though I'm used to working with driver's quirks and expecting to have to use some crossover equalization, I do like to listen to midrange drivers with a full range signal to get an idea of their character and identify the ranges where they start to get beam, or when they get out of control. The 370/32 is a most remarkable driver, combining the upper midrange clarity and definition of the best 5" drivers with good power range response and reasonable efficiency usually only available in 6-1/2" or 7" drivers. Active Bi-amplification approach- issues and advantages Another significant design constraint for three way systems managing the transistion between the bass and midrange elements. An important point in driver selection is matching the driver efficiencies. Because of the unusually low crossover point between the bass system and midwoofer/midrange, in order to maintain good electrical damping of the midbass and lower midrange, it would be undesirable to have to use any driver padding for the midrange. Furthermore, if I chose to use the Audax HD38S100 15" and HD 33S66 13" woofers, the problem I might have on my hand would reverse- these drivers have very large magnetic systems, 4" and 3" voice coils, respectively, and nominal efficiencies usually only encountered in Pro series speakers- because they are! Though both are low Fs drivers, at 25 Hz and 27 Hz respectively, the nominal efficiency of the HD38S100 is 98 dB, and for it's 13" younger brother, 96 dB. However, because the driver Qts is so low (0.12 and 0.15 respectively), their low frequency response is over-damped, and begins rolling off below 200 Hz. Examination of the measured or simulated response of these drivers suggests that by using a 125 Hz crossover point, the nominal efficiency in the region below this crossover frequency is in the ball park. Designing a passive crossover to deliver such a crossover point without interaction with the driver characteristics and with sufficiently low insertion loss might be a challenge,



	though. After taking an initial pass through such a design, I began to consider the possibility of other alternatives. To achieve acceptable performance with a passive network required using a zobel network to damp the impedance of the box/driver resonance above the port tuning, and a very expensive bill of materials with 10 AWG crossover coils and polypropylene film capacitors- rather large ones! Though I've never used active crossovers in my home system, I've designed and built several for friends, as well as assembled some Marchand electronics crossover kits for friends. The possible advantages to using an active crossover in this system began to look very attractive once I thought about them. First, it would allow some flexibility in determining the LF crossover frequency without any significant cost penalty- a few precision resistors to change the frequencies, instead of expensive new inductors and capacitors. Second, it would be easy to implement the low frequency room equalization between 200 Hz and 25 Hz which I envisioned the system may require. Third, and hardly insignificant, with direct connection to amplifiers, the LF damping in both the bass and midbass would be as good as could be hoped for, avoiding the series X_L and X_C
	inherent in passive crossover networks. This approach might also simplify the final "voicing" of the system, as an amplifier most congenial to the midrange and high frequency range could be used for the upper cabinet, while an over endowed low frequency monster could supply the power, control, and depth of response necessary to optimize the bottom end. Cabinet Design & Construction Because of the decision to use an active crossover, and the preference for a lower acoustical radiation height, the bass cabinet, though overall similar in dimensions for the acoustical enclosure, is several inches shorter, since it doesn't require the sub-enclosure for the passive crossover in the base of the cabinet. The internal volume is approximately 180 liters, after accounting for the drivers. A pedestal base with spikes is attached to the main bass cabinet in the final design. Unlike the wide range of adjustability which is built into the original X1 upper range module, my preference was to realize a single cabinet module, with fixed driver spacing, minimum diffraction issues, and a crossover optimized for the cabinet construction and my requirements for a closer listening position then might be common for typical rooms which the X1 has been reviewed in Though I wanted to preserve the basic acoustical configuration, including the use of auxiliary tweeters to fill in the room power response somewhat, everything else was up for grabs and reconsideration, as will be seen in the discussion of the electrical design. Figure 2 shows a wireframe view of the final cabinet design, showing driver orientation. Some points of the final realization, such as the tweeter backing baffles which prevent acoustic loading



	from the midwoofer/midrange drivers on the tweeter frames, are not shown in these drawings. Internal volume of the of the upper enclosure is approximately 18 liters. All cabinet panels taking the midwoofer/midrange are a minimum of 1.5" thick MDF, with all tweeter holes having a 3/4" MDF backing panel. The rear portion of the enclosure is used for housing the crossovers, away from the acoustical and magnetic influence of the drivers. Enclosure Construction An extensive description of the enclosure construction is beyond the scope of this brief paper, for which the intent is to focus on measurements and their influence on the design and performance of this speaker. I'll show a few pictures to illustrate briefly the scope of the work involved. All enclosure walls are a minium of 37mm thickness, with woofer front panels being 63 mm. Tweeter mounting panels are have plates behind them to eliminate the influence of the pressure wave from the midwoofers. Extensive bracing is used in the bass cabinet, which is tuned for a box Fb of 27 Hz. Figure 3 shows the pile of cut and partially glued subassemblies before starting assembly of the speakers. Figure 4 shows initial clamping and assembly of a bass cabinet after assembly of the front panel formed from three layers of MDF. Figure 5 shows the back panel sub assembly after clamp up and mounting of the 150 mm diameter port. Figure 6 shows the clamping for the complete bass cabinet assembly, after adding the rear panel subassembly and the second side. At this stage, the bass cabinet weighs over 90 kg without drivers.









	The upper cabinet assembly is a complicated process done in several steps; Figure 7 shows the final step, adding the rear panels in the crossover area, while Figure 8 shows the raw cabinet shell after gluing before the start of detail finishing. Figure 9 shows the cabinet lamination in early stages, with milling for the laminate covering done in the auxiliary tweeter area. Figure 10 shows one of the prototype cabinets, not yet with all assembly finished, but ready for acoustical testing. For reference, the finished cabinets are approximately two meters tall.





	Measurements and crossover development The measured response of the Eton midwoofer drivers in the upper module enclosure are shown in Figure 11. The Eton 370's are low Q, and might usually be used in a reflex





	type cabinet. In a small sealed system such as this, the low total Qts results in over damped response and the start of the low frequency roll off at about 175 Hz. The dual driver configuration does improve the upper bass and lower midrange efficiency, but the



	narrow cabinet, while aiding diffraction control, removes boundary loading below about 500 Hz. The actual measured impedance curve without crossover for a paralleled set of drivers is shown in Figure 12. The tweeter was also measured, and figures 13 and 14 show the measured frequency response as well as the impedance curve. The unusual impedance curve appears to be due to the combination of the aperiodic phase plug loading of the Td120dx2 and the use of ferrofluid damping. With measurements of the midwoofers and tweeter in the cabinet completed, CLIO's text export facility is used to export the measured data to test files which can be imported into crossover filter development programs. Many programs can read this test





	file; for initial studies I used SoundEasy 3, but for final crossover development I used LSPCAD. Because I desired a wider vertical lobe in the crossover region, I chose a third order Butterworth crossover topology, instead of a Linkwitz-Riley fourth order all pass design. After calculation of impedance compensation zobels and optimization of the network component values, the predicted response is shown in figure two. Figures 18 and 19 show the LP and HP crossovers, which are assembled using polypropylene capacitors and air core coils. Euro style power block connectors are used to facilitate a gas tight connection for wiring, while giving ease of assembly and disassembly for testing.



	The in room MLS response shows a similar dip in the axial response between 2 and 8 kilohertz as noted in the measurements by Stereophile magazine of the actual X1 speakers. The RTA response shows this dip not to be at the same level, due to off axis contributions, I think. The subjective response was a very clean and detailed high frequency performance, so no effort was made to boost the crossover response in this region. The complete speaker plus crossover input impedance is shown in Figure 20. Because the midwoofer drivers are nominal 7 ohm speakers, the combined impedance is low; which dictates some care in choosing the driving amplifier. Aragon 8002 and Ayre V-5 were both used with good results; the Ayre had the smoothest and most detailed midrange and highs, though.



	Only the upper crossover point ( at 2.25 kHz) is realized using a passive crossover; the lower crossover at 150 Hz is implemented using a custom active crossover design. This circuit is shown in Figure 23. The crossover is implemented with a third order state variable filter, at a nominal frequency of 150 Hz. The primary signal path op amps are OPA627, chosen for their clear sound and precision. The output buffers after the level controls implement shelving EQ functions, and provide high current drive for any length of output cables by including high current video amplifiers (AD815) to buffer the OPA627. For the high pass circuit to the upper module, baffle step compensation is included. For the low pass section, low frequency contouring/lift is applied to adjust the response from the low Qts bass drivers. Figure 24 shows the finished electronic crossover, built in a SESCOM chassis. The power supply and audio are configured in "dual mono", with separate audio PCBs and separate regulated power supplies only having the power switch and line cord in common.



	As can be seen from Figure 25, the cabinet achieved the tuning target (Box Fb) of 27 Hz; this is shown by the minima in the impedance curve measured by CLIO. Because these drivers are "over damped" for this cabinet alignment, from about 150 Hz down they have a response which declines at about 3 dB/octave, until reaching the box tuning. This is not the disadvantage is might seem, as flat anechoic response to a low frequency box cutoff will usually result in heavy bass in room, due to the boundary reinforcement which can be up to 6 dB, depending on location in the room. The electronic crossover implements a shelving equalizer which provides a compensation for the over damped alignment (due to Qts of 0.15), resulting in flat in room response down to about 25 Hz.