Coffee Table Subwoofers

Sleeping beauty (for now)


Concealed 18 inch Cerwin-Vega 189-JE 8Ω “Junior Earthquake” drivers seal a pair of relatively small coffee table substitutes for something out of nowhere. 😈  With very large piston areas and low excursion these drivers provide vanishingly low distortion at ordinary listening levels, but for full-on use watch out – they’re fearsome!  Mounted horizontally with their magnets hanging externally toward the floor for cooling and ease of service, the drivers get “the electronic boost”, are operated over a narrow bandwidth and have very steep (6th order) electronic bandpass filtering.  Driving them directly is a stereo UcD400 Class-D amplifier (≅200W/8Ω)¹.  No servo-control or other fancy-pants treatment – just crude “industrial driver” subs doing their thing.  And since subwoofer measurements are meaningless without showing distortion², I’ve shown at least the harmonic content near the bottom of the page.

  1. The Hypex modules were measured later to deliver only around 150W into 8Ω with a 40-0-40 transformer.
  2. Try telling them that at the forums.

Intended use

These subs are intended to complement the front L&R stereo speakers as a switch-in option.  They’re not intended to cross-over with mains.  Those front speakers extend even lower than the new subs on their own, however since the pre-set analogue bass EQ applied to the front speakers will not properly equalise the bass at the listening position with the new subs firing as well, a different approach is to apply JRiver Media Centre’s parametric EQ to all of the speakers when the subs are used³.

3. Superseded (September 2017) – see the Subharmonic Synthesis page.  The analogue EQ has proven superior to JRiver’s DSP equivalent when split to all four channels.


These are sealed subs.  IMO all other types are too badly compromised, from mildly dumb and over-sized “bass reflex” designs with complex resonances, to supremely and inexcusably dumb open baffle AKA “dipole” designs that as mere “velocity sources” cannot pressurise a room!  The more stupid something is these days, the more they get talked about.  If that’s your idea of “good bass” then go back to!

Also, there is too much talk and time wasted at certain audio forums about motion feedback or “servo” controlled high-excursion subwoofers.  The marketers push “high excursion” drivers as something great and the “believers” therefore want them, but it’s ridiculous.  Why bother with all the distortion which they must produce? Why fight something that you don’t need in the first place?  Large surface area sealed subs in multiples provide more than adequate SPL with just a couple of mm of cone excursion due to their inherently superior mechanical-acoustic coupling with air in the room.  By contrast, small diameter woofers require significantly more excursion to displace an equivalent volume or air.  BUT – and this is the point that no one seems to get about woofer drivers that are presented cone face-out in boxes:  For small diameter/long stroke, there is increased particle flow and turbulence across the cone surround which equals diminished mechanical-acoustic coupling for a given volume displacement.

Basic configuration

I haven’t seen this configuration anywhere (the closest I have seen is at Fig. 4-27 of Loudspeaker Handbook by John Eargle, but its driver faces down more conventionally).  Mine are like this:

That’s not to scale.  The shading represents the shape of  the sealed enclosure.  The “escape area” between the four feet is just over half the piston area (SD) of the drivers for a kind of “slot loading” to the floor in which additional damping and mass are effectively added to the driver to extend LF response and smooth overall response variations.

  • Aside: people have asked me “why did you hang the drivers upside down like that?”  Because:
  1. Professional drivers like these have the gasket at the front,
  2. The magnet is external to dissipate heat from the voice coil,
  3. No lateral forces on the driver cone due to gravity, and
  4. I like it that way. 🙂


A couple of these rebuilt drivers had been sitting in the back of a cupboard since 2006 – although they’re a lot older than that.  Ancient and crude yet of “industrial” build, the drivers have their roots in Sensurround.  They’re basically the same as the original 189E drivers that were horn loaded in multiples for earthquake and battle effects in cinemas in the 70’s.  Some might consider them lo-fi, but they needed to be used, so I figured I’d take a risk and that with limited bandwidth, some bottom-end boost (to compensate for the sealed 2nd order roll-off) and heavy infrasonic filtering, they’d manage OK without horn loading as long as I didn’t push them to extend to crazy lows, or drive them too hard generally.

Despite their limited Xmax (around 5 or 6mm), the large piston area of one of these drivers displaces a volume only equivalent to a 12 inch driver running at around 11mm of excursion.  That might seem pretty ordinary by modern “standards”, but as noted above large pistons couple with the air far more efficiently and the limited excursion actually contributes to lowering distortion.  Anyway if they turned out to have poor performance, or on the other hand be just too insane for music in the long term, they’d get thrown into the “Sensurround room” (here) which already has four 18 inch horn-loaded drivers – for an even half dozen for movies such as Battlestar Galactica (1978) now on Blu-ray in DTS HD-Master Audio pseudo-Sensurround!  Whoopee doo! 😆  Yes – still living in the past!

But they needed power and I didn’t want to waste too much money on this

The price of decent plate amps has gone up lately with the dwindling Aussie dollar (2015).  And plates wouldn’t have fitted nicely onto my oddball encosure design anyway.  Fortunately I still had this DIY UcD400ST stereo amplifier, but it was being used here.  Well it was way too powerful for background music anyway, so a simple solution was to buy a cheap integrated amp and make a swap.

So what about the subs?

As mentioned (or not), the C-V drivers have very low Xlim so cannot be used in bass reflex enclosures.  4th order bandpass was an option since a small sealed side would provide the necessary excursion control, but it was too difficult without T/S parameters and I didn’t want any port noises.  Hyperbaric was considered briefly, but with two 18″ drivers in a single enclusure, I’d never get it up the stairs.  The drivers can’t even be used in mid-sized sealed enclosures, so small sealed it was.

The drivers pre-date the “invention” of T/S small signal parameters, so actual Xlim and Xmax data is unavailable.  I didn’t need T/S data anyway.  What I did know was that they just seemed to be designed to go “clunk” (voice coil former striking the steel back plate).  But if the drivers are driven predominantly below a conservatively estimated resonance frequency in tight sealed enclosures, things can be “fixed” with a heavy signal manipulation anyway. 😯

Well you can’t just build subs and expect them to work, so I’ll get to them in a minute.

First the signal controller

This old crossover chassis was no longer being used, so I figured I could stick a couple of P48A subwoofer controller PCBs in to compensate for the natural acoustic roll-off of the sealed box design below resonance.  Unlike say a Linkwitz Transform where T/S parameters are required, P48A doesn’t even ask for them.  The R/C circuit components were chosen to position the peaking response at 27Hz with a 12dB/octave roll-off to either side and unity gain at 100Hz.

But prior experience with these drivers gave me two big warnings:

  • They really don’t want “anything” above about 100Hz else they get pretty ugly.  The shallow -12dB/octave low pass slope of P48A at the top end won’t be sufficient, and
  • The shallow -12dB/octave high pass provided by P48A at the bottom end wouldn’t be sufficient to prevent the drivers from clunking below 27 Hz in the box volume that I chose.

So I figured a way to convert a P09 active stereo crossover PCB into a “stereo” subwoofer bandpass filter.  These are meant for LR4 active crossovers (4th order slopes), so I chose -6dB points of 11Hz (infrasonic) and 110Hz (low pass).  These slopes would superimpose over and add to the 12dB/octave peaking response of P48A to provide a nicely tilted narrow electrical pass band with 6th order (36dB/octave) roll-offs at both ends.  Acoustically this ought to translate to a reasonably flat pass band between around 30 and 90Hz.

For each channel, the “hack” simply involved back-feeding the output of one filter section to just past a track cut in front of the other filter section in which the R/C values are different:

Track cuts and back-feed links

(pixelation due to ESP purchase agreement)

A few photos of the modified chassis:

Controller chassis

I had a heap of leftover OPA2134 opamps, so they were used in all 11 sockets.

The two P48As and a double-pole muting relay (seen in the above photo) are tucked in under the modified P09:

PCBs in controller

A stereo 100K Alps RK27 pot is shared by the P48A PCBs.  I made a dumb mistake in wiring the offboard pot as a rheostat per the schematic.  In that case it only trimmed the level through the second integrators and could not attenuate to anywhere near mute.  Actually it only shifted things by about 10-15dB.  I subsequently rewired it as a grounded potentiometer to enable full attenuation.  That kinda worked, but caused a jagged taper which was subsequently addressed – see below.

Since the chassis once housed a stereo active crossover, there are two sets of RCA outputs.  Now the bottom pair provides the raw peaking P48A responses and the top pair adds the P09 bandpass filtering.  This gives a choice, but the bottom pair probably won’t get used.  Here are the responses as measured (not a simulation):


(another mistake – this time around the P09 output buffers – caused that extra 6dB in the top trace – since corrected)

Here is a wider view also showing that there is a maximum distortion (just short of clipping) of 0.029% at the boost peak – predominantly 2nd and 3rd harmonics:

SPL and Distortion

That’s with the volume control fully cranked.  You’d never play it that loud.  At 12 O’clock the distortion falls to a peak of 0.013%.

So that’s the EQ up front of the power amplifier (in “stereo”).


The active filtering provided by the controller injects phase lag to the signal and this is a good thing because the subs are much closer to the seat than the main speakers with which their pass band overlaps.  The delay will align the bass with that of the mains for maximal impact.

Amp plus pre-amp (or “controller”) ready to go:

Controller and power amp

rear of pair

That power amp looks pretty timid, but looks can be deceptive. 😉  The controller gets the full frequency stereo outputs from a DAC that also serves the main stereo system.  Fortunately it had a pair of balanced analogue output XLRs that weren’t being used, so a pair of single-ended coax interconnects were wired up to pins 1 and 2 of a couple of XLR plugs.

The pre-amp sits next to the lounge chair for easy reach to adjust the level of the subs independently of the main speakers.  An option is to use JRiver’s internal 64 bit volume control (using the fantastic BubbleUPnP’s Android remote controller) to adjust the level digitally ahead of the DAC.  This would enable the level of the subs to track that of the main system – with the “relative” levels set by the dial on the preamp.  See – I have it all worked out. 😀

OK – the subs

There’s nothing quite as boring as a typical subwoofer with its “impressive” driver on the front face.  Anyway, these are ugly drivers, so they got concealed.  Also, large diameter drivers such as these can sag over time when mount conventionally on a vertical baffle allowing the voice coil to scrape, so these are mounted horizontally in the hope of remaining centred.  They might even produce less distortion this way since gravity cannot bias the voice coils and cone off-centre for asymmetric motion.

On the subject of “sag”, of course a driver can sag axially if mounted horizontally, but these old Cerwin-Vega drivers have a high free-air resonance and as such will display very low sag (Olsen 1957: x(mm) = 247/f2).

The sealed enclosures (including the baffle cut-out and cone are small at around 37 litres which is slighty larger than a volume known to prevent the drivers from bottoming out in an earlier P48 (alone) set-up.  In that set-up, P48 was set to peak lower at 18 Hz so power requirements were much higher.  The present and more conservative 27Hz peak together with slightly less powerful amps and the ultra-steep bottom end filter slope gave me a little leeway on box volume.

Here is one enclosure under construction (upside-down):

CV 189JE in tight box

The usual 32mm MDF, but with 25mm plywood for the driver baffle.  The top face of the baffle (photo below) has tee nuts to mount the driver and screw-in eyelets for stringing some long fibre wool stuffing up off the cone:

Underside of baffle

The baffle fits into these internal slots in the side walls and is held by PVA and 8 screws:


Lots of biscuit joiners, PVA, more screws and some maple bracing (not that it’s really needed with 32mm MDF):


The bolt holes match these actual drivers very specifically.  They are so roughly manufactured that the holes are not spaced evenly around the basket!  They only fit one way, cannot be rotated and they certainly can’t be swapped!  Actually although both drivers are labelled “189 JE”, one is much older, came out of a different factory and has a slightly smaller magnet.  Just as an aside, here to the right is a later model 189JE-II.  It is obviously better cast, but has nowhere near the meat!

Dust everywhere

They were however both rebuilt with identical voice coils, cones and spiders in 2006 in the USA.

The small tee nuts in the top panel hold supermagnets into recesses in the top surface:

Supermagnet ring

These are to help to hold down a steel top plate.

The design is such that if the subs proved disappointing, the drivers could be removed.  The plywood baffle would then simply form a brace inside a larger enclosure, and a new floor/baffle could be added to mount a more modern driver with the cone facing down.  Those angled braces would clear the magnet. 😀

A maple web in each bottom corner provides bolt-down points for the bottom plate:

Corner web

No clamps were required.  Clamps are completely unneccessary when using screws, and the screw heads if countersunk can be filled-over and painted, or veneered-over.

Tassie Blackwood veneer being applied:

Veneer over sticks

PVA was applied by roller to both the MDF box exterior (over all the filled-over countersunk screws) and the under-side of the veneer.  It’s allowed to dry, then the veneer is placed over some satay skewars which are removed as pressure is applied by hand from the centre towards the edges, then it’s ironed down with plenty of steam through an old tee-shirt then quickly pressed while still hot using an old telephone book:

Ironed down

One side per day.  The other box is used as a “press” overnight:

Other box to weigh down

Carefully trim and sand one panel at a time:

One nice edge

It takes days, but they start to look nice after a few coats of satin varnish:

A few clear coats

Top panels sealed with automotive high-fill primer and more varnish (magnets concealed with “plastic wood” and primer):

High fill primer on top

Long fibre wool stuffing inserted.  String was threaded through the seven eyelets and a single coil spring tensions the whole “star” (sorry string barely visible):

String suspension

A second star of string was added later.  It passes right through the coil spring.  If the first star/spring comes loose, the second string catches the spring so it can’t fall into the driver cone. 😀

Drivers installed (with 8 spring washers so the gasket doesn’t pull away from the baffle and leak over time):

One driver in

Aside:  This inside-out configuration might lend itself to these old drivers which (unlike say the more modern JBLs) have no cooling holes which might create unwanted turbulence noises which would ordinarily be suppress inside an enclosure.

Steel plates back from powder coaters.  6mm top plates have bevelled edges:

One top plate

One of the 8mm bottom plates is placed temporarily to mark through for drilling the bolt holes:

One bottom plate

Like this:

Oversized hole for bolt

The hole is deliberately oversized to allow the plate to be aligned perfectly.  There’s a tee nut on the other side of each of those corner webs.

Some egg crate foam to dampen the bottom chamber (maybe suppress any turbulence noises and unwanted out-of-band harmonics a bit).  Dodgey cardboard masking protects the drivers from adhesive overspray:

Spray adhesive masking

These little holes were easier than cutting the bolts down (yes- very fussy):

Bolt cut-outs

Bottom foam in

Neutrik Speakon sockets connected plus-to-minus because the drivers are in “backwards”:

Speakon on

Solid steel feet (powder coating hanger behind):

8 feet

Nice eh:

1 foot

Now for a little more fussing.  I was gonna put flyscreen mesh across the bottom opening, but it didn’t work nicely, so I went for speaker grille cloth instead (not that anyone would ever see it, but at least it might stop spiders from setting up home in there).  Some very thin 3M double sided adhesive tape on the upper face of the bottom plate around the opening:

Double-sided adhesive tape

Waiting for speaker grille cloth to arrive and lighter without the steel plates attached, the boxes walked themselves upstairs:

Upstairs sans metalwork

Neoprene skins (for underneath the top plates) cut to size:

Neoprene cut to size

OK got the cloth and lightly stretched it over a melamine board then plonked the bottom plate onto it so the double-sided tape can stick:

Stretch first!

The tape is time and pressure-activated!  No kidding, so I left some gym weights on top over night:

Gym weights

Trimmed and attached to box.  Tape/cloth sandwich ensures no rattling (a minor bow of about 1mm in the bottom plate actually helps):

One sub done

Spring washers and fibrous washers to prevent damage to the powder coating:

Fibrous washer

Yeah.  Over (or under?) engineered.

“Lids” on (held by “supermgnetism” through the skins – oh and gravity):


Power amp shoved down the back:

Perhaps a little less conspicuous than the subs?

Testing time

When first tested, the right channel didn’t work and the left basically played at near-full volume and thankfully without clunking!  I had made a mistake in the volume control as mentioned above.  Even so I was immediately aware that my pessimistic expectations had been completely unjustified.  With just the left sub firing, the bass was apparently far more powerful, solid, low-distorting and enveloping than the quad of modern long stroke sealed Morel 12 inch drivers of the front speakers running a 16Hz Linkwitz-Transform!  Indeed I could not tell where the bass was coming from and did not know that one wasn’t firing until I stuck my hand under each to feel that the grille cloth wasn’t moving on the right sub.  So I (sort of) fixed the volume control and investigated the cause of the missing channel … Some idiot hadn’t properly tightened one of the cable-fixing lugs in a Speakon plug! 😳

So they’re both working and that means an extra 6dB and OMG!  The whole house shudders and as far as your body goes, the sensation is fundamentally visceral.  It goes right through your belly.  It’s probably a good thing those top plates are heavy, else these things might just launch themselves! 😆

There’s no way I’d ever use modern long-stroke/small diameter (12″ or less) woofers again.  Compared to these “industrial” cinema drivers from the 70’s, they’re just high turbulence, low efficiency “air pokers”.  No kidding.

One of the few projects on this site that really put a smile on my face and far exceeded performance expectations.

I’ll run a sweep later and post the results below.  I don’t expect any miracles of “flatness”, but that wasn’t the aim of the exercise anyway, but it ought to be interesting nonetheless to see the distortion traces.


Well there were three:

  1. The volume control still wasn’t right.  The 100K pot was positioned after the first integrator, and before the second.  Revisiting the circuit diagram, I figured that this is no place for a log taper volume control!  It was only meant to be a set-and-forget rheostat/trimmer.
  2. The upper corner frequency of the bandpass filter (modified P09 PCB) was set at around 110Hz which just “sounded” a bit too high.  Changing a bunch of resistors from 6k8 to 8k2 would reduce this to around 90Hz.
  3. Although the left sub is completely impossible to localise (which is good), the right sub did actually tell me where it was somehow.

So there were some things to do …


Well I swapped out the upper corner frequency resistors for 90Hz and got the new volume control done.  First I had to replace the pot for VR1 on each P48A with a fixed resistor or maybe a wire link.  Tried 25K5Ω as a first guess:


Then I had this odd little PCB for an Alps RK27 in the junkbox, so I thought I’d put it to good use.  The new volume control will take the P48A outputs (100Ω output impedance) and feed the inputs of P09 (100KΩ input impedance).  The existing 100K log taper pot coudn’t be used as its taper would be screwed.  A 10K pot is better suited because the pot value should be at least 10 times greater than the preceding output impedance and at least 10 times less than the following input impedance.  I also thought “why not try ESP’s P01 just for fun?”  This takes a linear pot and converts it to near log taper using a resistor.  Of course this is “stereo”, so dual gangs are needed.  Here is the PCB with 1K2 resistors attached between the wipers and ground:

1K2 resistors

The stupid Chinese PCB has dumb terminal through-hole spacings for which headers are unavailable, so I used PCB pins for the wiring:

pain in the arse PCB pins

A right pain:

New volume control PCB

10K pot mounted to the PCB and installed with a new direct ground wire to the PSU:

10K B from back

As it turns out it’s a good thing I couldn’t find screw terminals for that little PCB as it’s too close to the side wall of the chassis anyway. 😀

Looks smooth on the scope with good taper and attenuates to a flat line.

Then a rethink of the overall gain structure of this and the main front left/right speaker system lead to a 10dB overall reduction of output from the W4S DAC-2 that’s up front.  This is really an aside, but the W4S DAC2 had been set on a fixed volume of “70” which is its maximum – giving an output voltage maximum of some 2.6V which is quite ludicrous.  There was a suggestion on some random forum (allegedly from a W4S insider) that “60” provides the best SNR, so I reduced it to “60” for something more sensible at a little under 1V maximum, upped the gain of the front left/right preamp by about 2.5dB (to get its analogue volume control to hover around the 12 o’clock position), replaced the temporary 25K5Ω resistors from the P48As (seen above) with wire links for an extra 5dB or so, and reinstated the earlier mistake at the P09 output buffers to add back that 6dB.  That was OK, but the volume control still had to be cranked beyond 12 o’clock, so I upped the gain provided by the output buffers again (by another 2dB or so) by swapping a couple of grounding resistors from 10K to 6K8.  What an ordeal. 😥

Anyway, the levels are all good and the pots are being used around their most “linear” ranges (10 – 2 o’clock).  Now the controller just clips at the 27Hz maximum boost point with a 1V sin wave input.  Ready to make some sweeps…


I set the Earthworks calibrated microphone up at the listening position between the two subs and did this at the highest level that my sound card (Focusrite 2i2) would allow without clipping and here’s what I got.



No room nulls as anticipated, but a known (from previous measurements of the main speakers) room mode at around 40Hz.

THD is as follows:

  • @ 10Hz 0.98%
  • @ 20Hz 6.7%
  • @ 30Hz 1.1%
  • @ 40Hz 0.47%
  • @ 50Hz 0.43%
  • @ 60Hz 0.6%
  • @ 70Hz 0.16%
  • @ 80Hz 0.24%
  • @ 90Hz 0.13%
  • @ 100Hz 0.24%

Awesome!  Thrashing the quoted figures that I have seen for any commercial subwoofer!

Note:  These are not anechoic chamber or outdoor measurements.  Things were rattling and shaking around the room at the bottom part of each sweep, so the mic would have been picking that up as well .


What I’ve done here is to compare the phase responses of the subs (blue trace) with that of the front speakers (red trace) measured some time ago with the same equipment from roughly the same mic position:

Phase overlay

Although it makes no perceptible difference down in the modal region, the curves align better for the bottom half of the shared band, but not so well for the top half.  This actually stands to reason given that the subs are closer to the seat (mic position) than the front speakers by about 2.5m.  A 7ms delay to the subs might just align those traces very nicely in the upper part of the shared passband.

Group delay:


Unexpectedly flat (not swooping up too much at the bottom end) and hovering around audibility thresholds (about 1 cycle).

It all looks pretty respectable, however…

Long-term listening and right sub localisation

Well after a week or so, the incredible power of these things and moreover the ability to localise the right sub had become overbearing.  The novelty kinda wore off, so I had a re-think after re-measuring a few things and decided to follow my own experience and do something a bit “scientific”. 😮

Some time ago I measured the bass response downstairs and added more subs into the measurement.  There’s a dual 18 incher up front and a pair of 18 inchers at the back corners.  Basically the SPL curve flattened out as the second and third subs were fired up.  Indeed it completely removed the need to EQ the front sub.

At present the front L&R speakers are full range (actually extending lower than the coffee table subs) and are quite finely EQed to my favourite seat from 125Hz down.  This is of course not optimal anywhere else in the room.

Instead of having both of these new subs right next to the seat to shake the living crap out of me, they could be spread out more randomly in the room and have their levels tamed for very low THD.  The front L&R graphic EQ could then be reserved for “pure” stereo listening (front speakers only) and bypassed (by flipping the bypass switch) when the subs are wanted.  A parametric EQ table could be generated using REW and keyed into JRiver Media Centre’s DSP Studio for application to all 4 subs.

Aside:  My experience in EQing the bass sections of the front L&R speakers told me that exacly the same EQ must be applied to all of the subs.  Any attempt to individually EQ the subs would only be a waste of time and effort.

So I moved “Sub 2” (the right sub that could be localised) to another position:

Moved Sub 2

So all 4 subs (if you count those built into the front speakers) were then a bit more randomly spread within the strange geometry of the room.  I thought that might make for a very nice response as one sub is nearby the mains as recommended by “one of those” experts on the subject.

However a quick test with some music immediately revealed that to be a really bad position.  Sub 2 was still easy to localise – maybe this time for being too close to the front left speaker?  Damn!  Was there something wrong with that sub?

No!  The next day I decided to move it here:

Best spot

Nowhere near any internal corner and basically alongside an open stair well.

And not only is it impossible to localise aurally (just a metre or so from its original position), but it’s even more insane than before!  Now things shake in a rear bathroom some 15 metres away and down a long corridor!  And they kick you on the chest – big time!  They actually thump you!

All four subs at once – measurements and EQ

OK, I was planning to EQ to a spatial average around the room, but canned that as stupid considering the only people who use the system sit at the two seater, so I took 10 readings with the mic angled differently (left/right/up/forward – yes I know it’s omni-directional) around the two seater and they look like this:

Multi Sub Spatial

Click for a bigger view.

So much for certain internet “experts” (with stupid panels all around their corners and cornices selling them as “bass traps” that cannot possibly work!) who say that the bottom end varies widely with small movements of the microphone.  Well the above traces prove that it doesn’t and why should it?  That’s over a 2m area and is real!  The only significant changes are from around 50Hz up!

That’s with all 4 subs (including those in the front L&R speakers – actually 6 drivers in total) firing.  The pink noise levels were matched between the mains and the subs to -20dB prior to taking the sweeps.  A phenomenal in-room low extension from sealed drivers!  Note the cursor position in the next picture is at 12Hz!  This is completely unexpected since the subs are electrically peaked at a conservative 27Hz and the front are Linkwitz Transformed to 16Hz.  The modal response of the room is what’s doing it and it’s a fluke (not genius – that’s for sure)!

Not surprisingly though, distortion is lower than with just the subs measured on their own.  Here is a typical one of the distortion traces (one of the 10 sweeps):

Low Distortion

Under 1% THD at 12Hz!  More up around the boost peak of 27Hz of course.  Interestingly the 3rd harmonic predominates the 2nd harmonic only below about 17Hz!

Also interesting: At 20Hz the 40Hz second harmonic is 30dB down (about 3.2%).  When you take equal loudness curves into account, this means that the second harmonic is “quieter” than the fundamental.  Similarly at 40Hz the 80Hz second harmonic is about 45dB down (0.46%) and that’s prior to the big 9dB EQ drop to compensate for the 40Hz room mode – see below.  At 30Hz the second harmonic is 36dB down at around 1.5%.

The software-averaged SPL looks like this:

Lounge Spatial Average

Quite clearly some EQ is wanted to counteract that big room mode at 40Hz and maybe smooth out the upper part (up around 100Hz) a bit, but assuming that the Earthworks microphone is actually accurate that low (precision seems clear enough) the bottom end looks too good to touch.  Here is a first trial EQ to the above average as generated by REW.

Multi Sub EQ

The target line seems to be geared toward subs that just don’t go that low, so a little tweaking here and there shifted the target curve closer to the reading and reduced the number of adjustments from 6 to just 2 low Q filters.  These are simply keyed into JRiver’s “DSP Studio” for both L&R channels.  Moreover, the same filters are applied to all of the subs and no filter was needed to retain that awesome bottom end!

Here is the final tweaked EQ graph (with a gentle rise in the target curve to the left):

Simple Bass EQ

This is not so fussy about ripples (actually a 2dB envelope is good enough) and it’s great because there are no boost filters (that can’t work).  Just -9dB at 40 Hz and -2.5dB at 110 Hz.  Both Q=3.  Simple as!

The 9dB Q=3 cut filter at 40Hz will dramatically reduce the likelihood of amplifier clipping and will reduce cone movement and voice coil heat during regular playback at ordinary levels.  Also, when playing music with JRiver (which does the parametric EQ) it’s pretty much guaranteed that distortion will be inaudible.  I’m gonna see if I can run REW via JRiver’s WDM driver to confirm this graphically later …

… OK well this will be easy.  Here is a line-level test sweep using REW and a loop-back cable between the Focusrite 2i2’s output and input sockets.  JRiver’s WDM is set as the default Windows audio device:

loopback test

That was just to see if I could get it to work.  Well it does and it shows that the JRiver DSP Studio parametric EQ is doing exactly what it’s supposed to, so the next step is to connect the microphone again and see how the EQed system actually measures in the room.

Done.  Final in-room measurement:

Harmonic Distortion All Subs

Show me a “bass trap” that can do that!  Note 1/48 octave smoothing that hides nothing.  Not a single bass strap in the room!  And the microphone was at the seat that’s up against the back wall.  So much for the silly “38% rule”.  It’s just more internet nonsense.

The seat is some 3 or 4 metres from the front speakers and just for fun, I measured full range – front stereo bass, mid and tweeters included – at a level safe for the tweeters.  REW’s prediction and JRiver’s EQ work brilliantly.  There’s no EQ (apart from a simple analogue baffle step compensation circuit) above the modal region.  Adding any “room EQ” above say 200Hz would be a complete waste of time.  The main speakers cross over from bass to midrange at 150Hz, but there’s nothing in the measurement to give that away.  The bass is within a 2dB envelope from 12Hz to just over the crossover point where it starts to become a bit grassy from room reflections and both speakers were firing so the grass is not representative of their near field responses taken individually (and to which digital EQ can be applied).

Modal region THD with all woofers running together:

    • @ 10Hz 5.5% (out of band)
    • @ 20Hz 0.71%  (2nd harmonic 0.49%, 3rd harmonic 0.29%)
    • @ 30Hz 0.76%
    • @ 40Hz 0.38% (from here on up THD is below the noise floor of the room)
    • @ 50Hz 0.49%
    • @ 60Hz 0.39%
    • @ 70Hz 0.40%
    • @ 80Hz 0.45%
    • @ 90Hz 0.29%
    • @ 100Hz 0.26%

June 2020 Addendum

The controller box got a bit of a revamp.  The plate amps for the bass drivers in the front speakers were replaced with DIY units devoid of any stupid phase shifter circuitry.  A consequence of that was phase interaction between the main speakers and the subs getting screwed up, resulting in a 105Hz cancellation.  After looking at new REW sweeps of the subs and mains taken separately from the seat, it was decided to lower the low pass filter even further from 90 to 70Hz, and to change the peaking frequency from 27 down to 20Hz. It was a matter of soldering “make-up” caps under the existing tuning caps on the PCBs.  The passband of the subs is now more like 12 to 70Hz, but more importantly, additional lag was introduced to phase-align the subs with the mains over as broad a range as possible.  This more narrow band renders the subs pretty much innocuous for ordinary music, but heaven help any unsuspecting sole when the DBX 120X-DS sub-synth is engaged. It is simply astonishing!  I actually frightened myself (in an exciting kind of way) playing a Lalo Schifrin track (“Hunt Down” from the 2017 Mannix re-recording) when the generated sub-harmonics (some of which were verging on infrasound) sent almighty wobble effects through the air and shakes through the floor like it was in Sensurround! Incredibly tight and solid near-silent but powerful vibrations to be felt.  I’ve experienced many commercial as well as DIY subs in lots or people’s homes and there is nothing that comes even remotely close to the performance of these.  Absolutely thrilling sh|t eh? 😀


ZenoI cannot find them.

6 thoughts on “Coffee Table Subwoofers

  1. Hello Ian. Another very good work and lovely finished cabinets , and they are not intrusive, even in the room that they are in. Congratulations, you always do a magnificent work. Kind regards, Elson.

  2. Thanks Elson for your kind comments as always. I’m looking forward to your visit. With any luck I’ll have ironed-out those problems beforehand. Cheers.

  3. Hi, I have one 4 ohm 189JE and I would like to build an enclosure. I know that the right choice would be the horn, but I have so space, and that bass reflex is not good for this driver, so I want to try your idea.
    Please, can you tell me the measures of your box ? Closed volume and other useful details to build the box ?
    Do you think that a band-pass configuration may be better ? Thank you very much, kind regards

  4. The sealed volume is small – around 37 litres including cone volume. I’d need the drawings for dimensions and I can’t find them, but you can do any shape you like. If it’s too big and the driver goes “clunk” stick some polystyrene in there. With such a small box the driver pass band must be limited to below resonance – no more than around 100Hz and it must get a lot of low-end boost per (for example) the ESP P48 project. Without this heavy EQ the sub will have an enormouse peaking response up around resonance and sound really bad and the cone could rip! Band pass (4th order) might work but you’d still need the small box behind the driver. Too hard with ports and a second chamber volume to work out with no T/S data and it would not hit you with the transients that a basic sealed would anyway. Caveat: I haven’t used the 4 Ohm version, so take this at face value.

  5. Hi, I just picked up two original Cerwin Vega Prostax 18B enclosures with the original 189 JE ll driver (serial 27968 and 27954).
    Ragged cabinets but drivers seem ok. Will do some testing tomorrow with an Amcron MA2401 amp. I am thinking of doing something similar as you did. Do you happen to have the T/S specs for these drivers for I have not found them anywhere yet. Thanks in advance.

  6. You’d probably have to measure for the parameters. I wouldn’t trust anyone else’s data since the drivers are so old. The data would not have been published when the drivers were new since they almost pre-date the work of Thiele and Small.

Leave a Reply

Your email address will not be published.