This is an old page! Many tweaks and corrections to these speakers can be found elsewhere on the site.
This full range speaker system is a bit different to anything I’ve seen with a brand name on it (or the meaningless words “Reference” or “Signature”). 😆 It’s in a way an attempt to design and build a “super hi-fi” system to out-perform the tri-amplified Jamo Oriels in something a little more compact (well they’re smaller, but around 25% heavier at 100Kg or so each). The three-way system is designed specifically for an external active crossover. Four integral sealed 300mm drivers provide astonishing bass performance which can actually be felt in a large room way down and just into the infrasonic range.
October 2013 update
Just for fun and as an indication that these speakers are not simply all about the bass, I made some on and off-axis near-field (1m) in-room measurements of the mid and upper range output using a calibrated Earthworks microphone. Of course the measurements were not taken in anechoic conditions, but I’ll post them here in a few days. I think they look pretty reasonable. Now here.
The Bass Sections (16Hz – 150Hz)
I should mention that higher SPL in the bass department could have been achieved with a bass reflex (ported) design. Indeed simulations of a single identical driver produced a similar SPL curve to that achieved with as little as 50W input. However years of experience told me that the tactile effect that I was aiming for could only really be achieved by exploiting the linear excursion capabilities of large sealed drivers. The shallow (second order) low frequency taper is good, but it sets in relatively high (compared to some other options). To compensate for that, I decided that the “ultimate compromise” might be a Linkwitz Transform (LT) applied in stereo. Open baffle dipole “velocity sources” or “flappers” are next to useless for proper enveloping super LF performance, so weren’t considered.
Why cross so low? Actually, I wanted to cross even lower at around 120Hz, but the midrange would not stretch to it. The reason is simple: It completely eliminates the need to consider box resonances which would typically be in the 200-300Hz range in large enclosures like these. Many people cross at around 300Hz and this to me is not a good idea.
As far as I understand it, an LT (or any other electronic EQ) cannot “boost” the bottom end capabilities of an existing subwoofer system. It can however tame the upper range of the bass output to blend well with a midrange driver.
103dB of acoustic output at 16Hz is far more than anybody needs, but the simulations provided it and that was the target.
After looking at many sub-woofer drivers, the Morel brand seemed to stand out as a no-nonsense product of good design suited for sealed enclosures, but were not available in Australia, so I ordered four of the 12-inch “UW 1258” units directly from the factory in Israel and these were made to order. Within their linear range, two of these displace an equivalent volume of air as a single 21 inch drive. They have external voice coils (the magnet is inside the voice coil former) for good cooling and high power handling.
The box dimensions were chosen to limit cone excursion. This resulted in a Qtc of 0.86. There is no broad hump in the response however since the LT compensates for it.
Where to mount them? A wide front baffle could muddle the midrange and treble performance and ruin the stereo image and my cabinetmaking skills don’t really extend beyond “square”, so it seemed side-firing was the way to go. Below the crossover frequency of 150Hz the wavelengths are longer than any dimension of the cabinet and the radiation pattern is essentially spherical, so this is fine. Ideally the woofers would mount back-to-back to cancel cabinet vibrations, but again that would require a wide cabinet to allow space for the huge magnets. Instead the cabinets were built as a mirrored pair with the drivers stacked one above the other at the respective “outside” baffles. This would be less likely to excite room modes – similar to placing two subs symmetrically outboard of “main speakers” rather than relying on one larger sub. The room is quite open and wide on both sides of the speakers and the listening position is at the “cenre line” so the wave fronts arrive together which is good for tight/fast bass. Also, the cabinets were built quite heavily with substantial internal bracing and a combination of internal damping materials.
Anyway the bass works really well as this will demo:
Sub-woofer Plate Amps
Expensive brand name “monster power amps” such as the old Krell, Mark Levinson, Simaudio, Musical Fidelity and the like might be an acceptable compromise for passively crossed speakers (if you don’t mind throwing all that expensive and imaginary “detail” out the window via the crossover inductors), but for bass alone in an active system, they’re not getting my money. I reckon all you really need is adequate current capability, quiet operation, a high damping factor and well implemented protection circuitry. So called “pro” amps are not suitable due to their noisy cooling fans, but Hypex DS 8.0 Class-D plate amps (800W into 4Ω) tick the boxes and their low output impedance ensures effortless bass. They are compromised on serviceability by their extensive, but necessary use of SMDs, but you can’t have it all and I certainly don’t have the ability to build super high-powered amplifiers.
Serious Addendum – June 2020
With the great benefit of hindsight, my choice of these amplifiers was a very big mistake! Never again will I buy anything branded “Hypex”. These amplifiers have unwanted and damaging phase shift filtering networks that cannot be bypassed and as a result the LR4 crossover to midrange was bastardised. The amplifiers failed too. An infinitely superior result came from my own DIY plate amps that replaced these in May 2020.
Instead of connecting the standard bridged (BTL+ and BTL-) outputs to the drivers as recommended by the manufacturer, the modules can be “unbridged” into a pair of drivers simply by using the connecting tabs provided with one driver reversed to maintain phase (there is no “BTL” switch to flip the phase of one module):
This reduces the demands on both modules and the power supply.
In effect that’s one amp per driver. Being 8Ω nominally, the drivers each have 210W¹ available just before amplifier clipping (420W per stereo channel) and Xmax being reached simultaneously. The amps are therefore fundamentally incapable of pushing the drivers beyond their linear range in their sealed enclosures.
- June 2020 correction. Those figures were gleaned from the published specifications, but are a gross over-statement! The amps only ever delivered about 150W into the speaker loads.
The above reduces the gain by 6dB, but that can be had back, by tying the input sockets:
The “hat” is one of those idiotic things marketed by nitwits to “block interference” from unused (out of circuit!) input sockets. 😆 I use them to disable unwanted input sockets and simplify things at the back of amps where I can’t see while plugging in interconnects.
The on-board signal filtering is disabled with the exception of the 4th order Butterworth subsonic filter and adjustable phase shift network which are non-defeatable.
Midwoofers (150Hz – 2.2kHz)
Open baffle/dipoles were again dismissed outright for their obvious in-room response aberrations. Whilst certainly useful as home theatre surround speakers, they are not IMO suitable for stereo hi-fi! That certain highly regarded individuals extol their hi-fi “virtues” is completely beyond me. There was also no need to consider line arrays, D’appolito arrangements, or other options with complex horizontal dispersion patterns either.
Because the overall SPL is limited by the attenuated upper range of the sub-woofers, a single sealed driver would provide more than adequate SPL and the simple lobe patterns would not be unnecessarily complicated. So with only one driver required per channel, I thought I’d buy something decent like Eton Hexacones for example. I ended up with quite inefficient Scan-Speak Illuminator 18WU/8741T00 8Ω “midwoofers” and mounted them conventionally in sealed enclosures for good transient response. Keep it simple.
They have compact neodymium magnets and cooling holes in the voice coil former.
At 2.2kHz (before a pronounced peak followed by the inductance taper) the manufacturer’s on-axis response curve is up from the mid-band level by about 1 or 2dB:
Being a standard anechoic “large baffle” measured small signal response, box loading and damping, baffle shape and room reflections are not accounted for in that graph. Anyway I aimed at avoiding the big “warning sign” there around 3-6K with the sharp roll-off filter of the active crossover at 2.2kHz. Distortion of the upper frequencies is not evident.
For fun, I also simulated the so-called “baffle step” response which ignores diffractive surfaces other than the edges of the front baffle (and they are everywhere), but it is interesting nonetheless because it predicts a 6dB general slope in the response from around 1kHz (in open space anyway) down to the lower crossover point:
The drivers are at unequal distances from the side and top edges of the front baffle and the above graph takes that into account, but even so I had never really observed or noticed an in-room baffle step response so simply took that simulation as a warning sign for possible future compensation.
The drivers are quite inefficient, but are powered directly by the lateral MOSFET amplifier which has ample power (150W) to compensate and superb control.
The chosen crossover points bound the simulated in-box impedance deviation to no more than 2Ω (6-8Ω ) which is practically as good as a resistive load and therefore naturally ideal enough.
Tweeters (2.2kHz up)
Why cross so low? Well, apart from the midrange driver limitations mentioned above, if there are any phase/amplitude anomalies brought about by poor acoustic summing of the driver responses around this critical crossover region, I don’t want them right where the ears are most sensitive (3-4kHz).
The tweeters are Scan-Speak Revelator D2904/7100-03 soft domes – the best I could find to approach the obsolete Dynaudio Esotar T-330 D. Ribbons and the latest beryllium tweeters with their wonky response curves weren’t considered. They are set at or slightly above ear height for the seated listening position.
Here is the manufacturer’s published response:
There is the expected off-axis response taper (green and red lines). Rather than mucking around with makeshift waveguides for the tweeters (they have shallow ones anyway) attempting to compensate for that and probably introducing other problems, the speakers could be toed-in to bring the black line to the listening position.
The tweeters have a nominal 4Ω impedance and are powered by the Class-A power amp which can deliver up to 9W without any crossover distortion (around 100dB!).
The tweeters are positioned directly below and as closely as reasonably possible to the midrange drivers.
A 27μF SCR MKP protection capacitor is in series with each tweeter.
Primarily the capacitor helps to protect the tweeter from amplifier fault DC, low frequencies and any switching transients caused by dodgy pre-amps and certain digital processors, CD players and especially power amps. Everything up front of the speakers in this system has proper muting relays, and the tweeter amplifier is incapable of such insults due to its dedicated protection and muting circuitry. So the cap is an insurance policy in case “someone” connects something else by mistake.
At 27μF, the capacitance was tuned with the tweeter’s impedance curve to a -3dB corner at around 650Hz (inaudible due to the sharp input roll-off by the active crossover) where the impedance is about 9Ω. The tweeter impedance at the crossover frequency is the nominal 4Ω figure. As a result, the capacitor attenuates the tweeter’s response by 1.6dB at the crossover frequency. This is generally undesirable, but it seemed a good idea at the time to compensate for a slight peak in the anticipated summed output of the two drivers brought about by the published response rise of the midwoofer at around 2Khz mentioned earlier (the tweeter’s unfiltered response just starts a shallow roll-off below 2.2kHz). Of course the active crossover attenuates the input signals by 6dB at the crossover frequency anyway.
The protection capacitor also causes a phase lag of about 34º in the amplified signal to the tweeter at 2.2kHz. If the acoustic centres of the two drivers were on the same vertical plane, this would tilt the central lobe downward by about 8º. Here is a rough simulation of the lobe pattern (with 4th order crossover):
That central lobe “fattens up” with decreasing crossover frequency and some compromise was of course needed in determining it. These and many other factors interact and the room furnishings probably scramble the actual response away from the simulations, but a little thought before building never hurt anyone. 🙂
The capacitor also reduces a very slight 100Hz mains-related noise from the Class-A power amplifier’s output (which could be heard with ear to a midrange test driver earlier).
The above didn’t really work out so well. In-room impulse response measurements found problems and these are dealt with elsewhere on the blog.
Having no regard for expensive speaker cables (or should I say the crooks who peddle in them), the internal cabling is nothing fancy. The tweeters and midwoofers connect to the Speakon sockets with a simple double-insulated high current “automotive” 48 strand copper cable. The woofers are connected to the plate amps with ordinary 11AWG multi-strand copper “Soundlink” speaker cable purchased off the roll.
Constructing the Cabinets
The cabinets are fairly tall at about 1280mm including the feet. Depth front-to-back is around 500mm and overall width is around 275mm excluding the metal base plate and feet.
They are basically 32mm MDF boxes built as strongly as possible with biscuit joiners, lots of PVA and glued-in screws (to withstand the pressures of the sealed design) with a timber veneer applied after construction to conceal the screw heads and butt joints. I decided on such thick MDF for one main reason: panel stiffness to increase natural panel resonances out of the chamber pass bands. Thinner boards would have necessitated substantial internal bracing to avoid resonances. Extra bracing would rob from the internal volume and the speakers would have to be bigger to compensate. They are much more heavily constructed than commercial speakers! Although the MDF surfaces didn’t have to be absolutely perfect, automotive body filler was applied and sanded flat around the edges and over the countersunk screw heads prior to applying the veneer:
A plunge router with a special base plate called a “Router Buddy” was used to form the speaker rebates including the small one for the tweeter, after which the centres were removed to form the apertures. A few photos:
The holes for the mounting bolts were drilled after marking through the speaker bezels and metric T-nuts were installed at the back before assembly:
I like T-nuts because they fix the drivers by applying a compressive stress to the MDF rather than shear stress of directly tapped threads or machine thread inserts – more secure over the long term.
Just an earlier test-mounting:
The required enclosure volumes put the tweeter right on the partition:
So the partition was filed-out to clear the T-nuts:
The resultant air leak between the midwoofer and subwoofer chambers was sealed by gluing a few layers of cardboard to the other side:
Despite the large parallel side panels, the number of parallel internal walls was minimised by setting the partitions at odd angles. This probably achieves nothing, but it doesn’t matter. The inside is lined in places with Bostik adhesive bituminous (automotive) pads:
That obtuse partition in the middle divides the subwoofer chamber into substantially equal volumes, but deliberately does not reach the front and back just in case the volumes ended up slightly different. Simulation-wise one 86 🙂
The box is held together with industrial strength PVA, biscuit joiners and 60mm enamelled decking screws which were screwed into countersunk pilot holes with PVA as “thread lubricant”.
The top of each cabinet has a powder coated 6mm steel plate with bevelled edges (shown here before powder coating):
The plinths are 8mm steel and give the cabinets a wider footing. The speakers are to sit on a carpeted timber floor, so I gave them wide round feet. The legs are solid steel and are high enough to allow a vacuum cleaner head underneath. 😀
The heavy steel top plate is “magnetically suspended” on neoprene padding directly upon the top of the midwoofer chamber:
The idea was to add mass to stabilize sideways reactive movements of the enclosures against action of the woofers. 😎
These super magnets are inset in the top panel and are anchored by internal T-nuts:
These were shimmed up with cardboard rings to ensure the bolt heads were as flush as possible, then concealed with automotive filler:
And a few coats of acrylic “primer putty” undercoat (“high-fill primer”) and the magnets were “gone” :cool::
“They” don’t mention it, but the Hypex plate amps are not airtight at the sockets (and in the case of the DS 8.0 – around the square pot shaft as well). Even if they didn’t leak, plate amps should be isolated from the pressure chamber anyway, so special amplifier compartments were made:
This was a tricky part of the cabinet design. Being wider than the for the bass drivers, a small overall footprint, heavy 32mm MDF construction and a wide plate amp didn’t necessarily go hand-in-hand. Also 32mm MDF is a lot more expensive than two lots of 16mm MDF, so I decided to PVA-laminate two 16mm boards to form the large side panels. Both boards were cut to the same size of the shop, but these corner cutouts were made at each inner board before laminating:
So the cabinets are 32mm narrower than otherwise. 🙂
In order to completely conceal the joints, the veneer was applied after assembling the cabinets. Use of contact adhesives was dismissed as it is toxic, difficult to clean up and gives you just one chance to align everything because it sticks permanently at first contact – too risky.
Iron-on “thermo veneers” are OK, but are available only in a limited range of timbers – none of which really grabbed me. Paradoxically, speakers that I’ve seen finished in exotic figured and burr timbers tended to look a bit “plastic” and I baulked at the idea of hand-cutting speaker holes though the random grain directions anyway, so I just went for something that suited the furniture. Tasmanian Blackwood is a nice solid looking sustainable timber and is not expensive, so I purchased some in raw 0.6mm sheets and decided to affix it with PVA. The veneer was cut to leave some over-hang:
“Exterior” or “Industrial strength” PVA (which is stronger and less runny than standard “interior grade”) was applied with a roller to both the MDF face and the back surface of the veneer.
Carefully around the driver apertures not to allow the glue to drip into the rebates:
It was allowed to “clear” as the water evaporated (about 30 minutes – depending on the weather):
Only then were the surfaces brought together. Because the glue is basically dry, the sheet can be repositioned (if any white colour is still present it can grip permanently like contact adhesive at first contact).
Once in position a steam iron was pressed down on the veneer through an old centre out with moderate pressure the veneer affixed perfectly with no funny lifting or other problems. I did just a small area at a time and pressed down with a telephone book straight after lifting the iron and keeping pressure on until the area cooled. Once the heat dissipates from an area just done, the PVA is basically set, but must be left to cure for 24 hours. No need for a press, but just in case, I left reams or paper and telephone books across the whole surface over night. This method was found to be ideal after a few techniques were tested on some off-cuts. It allows the moisture-induced expansion of the veneer to occur prior to fixing down and therefore prevents lifting at the splice joints.
After a day to allow the PVA to cure completely, the peripheral over-hang and excess around the amplifier boxes were carefully trimmed with the Stanley knife:
Special veneer-cutting tools are out there, but this simple “ruler method” ensured against cutting into the edge and left about 1mm to sand using a flat sanding block:
Those speaker wire holes allow the 11AWG insulated wires through snugly without being absolutely air tight. This follows Linkwitz’s recommendation to include a “pin hole” to equalise internal pressure to atmosphere.
The adjoining faces were done by the same method and after a little sanding, the edges looked pretty good:
The only real downside of the post-assembly veneer application method was that the speaker holes had to be cut out very carefully by hand. I used curved-blade, blunt-tipped hobby scissors with the outer wall of the rebate used as a guide for the lower blade.
A little sanding, then achieved a clean edge:
I wanted the cabinets to appear as though formed as a solid timber pillar, so only the vertical surfaces were covered. Grain across the top in either direction might have been a giveaway (and was another reason for having the steel top plate).
Four coats of polyurethane clear satin varnish were applied with a spray gun, sanding in between to achieve a reasonable finish, but some annoying runs appeared in the top coat. Not happy with that “unprofessional” look and not confident in my spraying skills, I sanded the runs out, lightly sanded the whole surface with 1500 sandpaper and rolled on another two coats with a fine mohair roller. The result was excellent.
Used some adhesive foam tape to seal the woofers (mids and tweeters came with it already):
A bit of for the copper wire ends before inserting them into the spring terminals:
Neutrik Speakon sockets attached to tweeter and midwoofer wires. Bottom plate bolted to pre-installed internal tee-nuts and plate amps installed:
The front baffle could have been given rounded vertical edges by routing out a square strip and inserting say Tassie Oak quad. Rounded edges supposedly diffuse diffraction, but to quote from the Linkwitz site “much is hypothesized, little is proven and much is overrated when it comes to diffraction” and I like the square look anyway. 😛
The finished cabinets are rock solid, “dead” and very heavy at around 90-100Kg gross each.
The page was getting too here.
Setting up and Calibration
I guess you can do all the simulations you like with a computer, but they’re based on T/S and other parameters and equations that don’t necessarily translate to reality. I simply used them as a starting point expecting to require some adjustments. equaliser bands to tweak for the stereo subwoofers in a difficult room. Here there is the pre-configured LT curve and nothing other than the overall levels to adjust in front of the crossover output buffers. The ESP-based power amps have the same gain (26dB). Like for most products these days, the Hypex documentation is highly inadequate, but 26dB could be taken as a reasonable guess (edit: they have now published 32dB as the gain of the DS8.0 on RCA inputs), so assuming that without actually measuring it, compensation was made only for the relative driver efficiencies. This table sets it out:
|Quoted efficiency (dB)||87 (1W/1m)||5.4 (2.83V/1m)||94.4 (2.83V/1m)|
|No. of amps/drivers||2 (+6dB)||1||1|
|Average impedance across selected pass band (Ω)||8||8||4 (-3dB)|
|Overall efficiency 1W/1m||93||85.4||91.4|
|equired compensation (dB ref’d to 0)||0||7.6||1.6|
|Equivalent electrical gain (V)||1||2.4||1.2|
That’s for just one speaker and assumes +6dB for the two subwoofer drivers. A function generator and oscilloscope were used to set the P09 output trimmers to the tabled relative voltage levels at the crossover frequencies. The midwoofer may appear to be a bugbear requiring too much input, but the lateral MOSFET amplifier has more than enough power to compensate – with headroom to spare. The tweeter clearly requires very little power and the Class-A amp is more than adequate (even without exceeding it’s 9W pure Class-A limit). Tones swept through the system revealed nothing too odd, but my ears are not a calibrated instrument so who knows. After some experimental fiddling with different types of music, I ended up tweaking the levels a tiny bit.
Since taking that photo the disc player and W4S DAC-2 have been “thrown on the tip“. The DAC-2 was particulary poor in my opinion.
Everything is powered at 226V with a bucking autotransformer/soft-starter (shown at the foot of this page) which feeds this Dell IEC power strip which has its metal body grounded to the rack. No stupid fancy power cables:
Interconnects are all like these and the cables to the speakers are basic Soundlink “figure of 8” off the roll – none of that stupid expensive stuff:
Long fibre wool stuffing (from a Teddy Bear supplier) for the mid-range and bass enclosures:
Better and nicer than fibreglass!
The LT was originally set up for -3dB at 20Hz, but in the upstairs room the speakers could not stretch to the very low notes that I was accustomed to with the Oriels which are in a smaller room downstairs and on a concrete floor. The upstairs room is oddly shaped and quite open:
Also, the floor is only light-weight timber and there’s a large garage underneath to suck bass from the room. Attempting to achieve the really low notes resulted in unbearably loud overall SPL from the mids and tweeters. Anyway, the organ test track induced only about 5mm of one-way cone excursion in the subwoofers which indicated that the plate amps had plenty in reserve (the simulations had full power coinciding with the 12.5mm perfectly flat” by overwriting the calculated resistor values with some close but unusual value metal film resistors that were available from Element14 like 14k, 102k and 60.4k. Here they are (odd black resistors) installed on one of the P71 PCBs:
Instead of lifting the old series “make-up” resistors I just soldered the legs of the new ones across them under the PCB:
Input caps C4 were also increased to 1.5μF using the left over tuning caps. This practically eliminates the shallow on board subsonic filtering in favour of the steep 12Hz subsonic filter in the plate amps which cannot be defeated. 16Hz made things a bit tight there.
It is really quite remarkable what a difference this made. The downside of course is that some really powerful unwanted rumblings unintentionally present on a lot of noisy recordings is lifted to the surface – but it’s kind of “exciting”.
In-room impulse response measurements made later found a few problems and these are now corrected and discussed on the “DSP or Not” page.