Introduction

More than ten years ago now, when I decided to setup a large front projection Home Theatre style system, I needed some kind of subwoofer for the surround sound system to make use of all the THX Laserdiscs I had been collecting in Atlanta GA  while I was working there with BT/Concert.  I ended up buying a lot of equipment form Jeffries Hi-Fi in Eastbourne because that was the nearest place I could get the Seleco projector that I wanted.  I asked what type of subwoofer they would use for a large system like the one I was building and they suggested REL who made quite a range.  As I was somewhat dubious about adding any kind of subwoofer to the system I asked if I could borrow one.  After a while they got me a demo unit from REL which was a Stentor model and was the largest they had ever sold at the time.  In those days, the Stentor was a 200 watt MOSFET system with a 10 inch Volt down firing driver in a 72 liter bass reflex enclosure.  It looks like the latest version is very similar but now with a 300 watt amp.  I connected the Stentor into my system that also consisted of my Velleman K4000 tube amp and my Quad ESL-63 speakers and started to set it up as REL describe in their manual.  This consists of the setting of three controls on the amp of the subwoofer.  Two of them control the upper crossover frequency (from 30Hz to 120Hz in 10% steps), one course adjustment and one fine.  The other control is for the gain.  The Stentor has a line level input that I connected to my Pioneer SP-700D ProLogic processor full range mono output or if you need to it also has a pair of speaker level inputs to connect from the power amp output directly if needed.  It took quite some messing around with the controls but after a while I got it so it sounded quite good although I ended up setting it to the minimum value of 30Hz.  Its cool to listen to CD's you've been listening to for years and suddenly hear deep bass that you never knew existed, so I was quite pleased with it!  It's also worth pointing out that most CD's I listened to where I expected to hear something different, sounded just the same.  Next I cranked up the system and watched the Peter Gabriel Real World tour concert on Laserdisc with digital audio.  Wow!  Holy cow!!  It was fantastic, so I cranked it up some more to see if something would break before Susan told me it was too loud!  Luckily, Susan likes Peter Gabriel so the REL was in for a real test of guts.  Well, the thundering dynamics of Tony Levin's fantastic collection of bass guitars very shortly proved to be far too much for the REL, and the Velleman K4000 tube amp and the Quad ESL-63 speakers were just blazing past it!  The poor Volt driver was popping all over the place and I was sure it was going to bust a gut if I were to continue!  Sorry REL Stentor, audition failed!  Having said that, the REL Stentor did have a good sound and still receives good reviews (if you believe those), but to be beaten down by a tube amp and electrostatic speakers is nothing short of embarrassing for any subwoofer!  Even Susan asked why I was turning it down!  Not only that, they go these days for about $5000 which I consider to be totally ridiculous considering the performance.

Reverse Engineering

Well, seeing as the REL Stentor was just lying there with it's feet in the air, I decided that some open heart surgery was in order to see what made it tick before I returned it to the shop!  Unfortunately, there were of course no digital cameras back then, so I'll just have to describe what I found as best as I can remember.  Inside the layout was much as expected.  Class B MOSFET amp built into the back plate with an active low pass filter, a well built and heavily braced cabinet lined with acoustic absorbing fiber with a single bass reflex port and the 10 inch Volt driver.  As I indicated above, it looked like the Volt driver was the weakest component although some of the blame lies with the front end active filter in that it was only a low pass design and did not provide a band pass so that the subsonic frequencies that were popping the Volt could have been avoided as there is no loading for the speaker cone from a bass reflex cabinet at such low frequencies.  The bass on the digital audio sound track of the Laserdisc was so good that you could actually see the Volt bass driver slowly moving under the pressure of Tony Levin's thumb resting on the strings of his bass even when no notes were actually being played in between the songs!  Looking at all the guts of the REL laying there it struck me that it wouldn't be hard to build a much bigger and better system myself for a fraction of the cost!  MOSFET's, operational amps, bass guitar speakers & wood are not that expensive to buy!  Of course, replicating the high quality wood finish of the cabinet would be hard, but I was more interested in the quality of the sound than the appearance.  With that in mind, I set about investigating the main mysterious component of the system: The adjustable active low pass filter which REL proudly refer to as ABC (Adjustable Bass Control) as this did not appear to be an item I could buy off-the-shelf.

The REL ABC filter

Component by component investigation showed that the REL ABC filter consisted simply of three DC coupled operational amplifier stages each using one half of a TL072 operational amplifier device.  I have drawn a simplified diagram of each stage that excludes the power.  The first stage was a variable gain inverting buffer.  The 100K potentiometer shown is the level control knob.

The next stage was a unity gain low pass second order Sallen and Key filter.  The four position course frequency adjuster was a dual gang rotary switch that simultaneously changed the values of C1 & C2 in the filter always keeping C1 double the value of C2.  The four position fine frequency adjuster was a dual gang rotary switch that simultaneously changed the values of R1 and R2 always keeping R1 equal to R2.  When C1 = 2*C2 and R1= R2, it gives the filter a Butterworth frequency response. The mathematics that describe the filter are shown here.

The third and final stage was a unity gain inverting buffer which in combination with the previous stages caused the output to have the same phase as the input and gave a suitable output impedance to match with the input of the power amp.

REL provide this table as a guide to the crossover frequencies applied by the various switch settings and these are inline with the formula shown.  This formula only works when R₁ = R₂ and C₁ = 2 X C_2.  Since it was professors R.P. Sallen and E.L. Key that first described the Sallen & Key low-pass filter in 1955, and it's implementation in operational amplifier based active filters designs is openly described in the datasheets of the companies who produce such devices, there is clearly nothing special or anything proprietary to REL in this design and as TL072's are just 26 cents a piece, the whole filter can be made for pocket change!  Hmm.., now I'm really on REL's case about the ratio of cost to performance with the Stentor!  Also, as you might imagine, changing the settings on those rotary switches while the amp is on causes a frightening pop from the poor Volt speaker!  So you need to power it down each time before changing the setting which can make the whole process somewhat tedious.

Do It Yourself Bass!

Now knowing all about the active filter part, I just had to design my own board, find a suitable driver, an amplifier and a cabinet design so that I could build my own super version of the basic REL Stentor.  That was all as easy as just picking up the Maplin catalog!  Maplin carry an enormous range of components and equipment and I still order things from them now even though I am in the U.S.A. and this is a description of the system I built.

The Driver

I turned to the speaker section and looked for the biggest most powerful bass driver that Maplin had to offer!  It turned out to be a 400 watt 18 inch Big Cat bass guitar speaker made by Eminence in the U.S.A.  It looks like they don't make the same unit anymore but it is a big heavy cast frame speaker with an enormous magnet and a four inch voice coil that is vented through the back of the magnet!  The current Sigma Pro 18 looks most like it although that one is 650 watts.  It looks totally awesome from either side.  The tiny 10 inch Volt driver from the REL Stentor looked like a little toy by comparison!  The Thiele & Small parameters indicated the the speaker had a resonant frequency of 33.5Hz and this was the only parameter I looked at and I just designed the cabinet to have the same resonant frequency.  At the time, it cost 129.99 British pounds (about $195).

The Amplifier

For the amplifier, I decided to buy a K4010 Velleman kit so that I could build the amp and filter electronics all together.  The K4010 is a powerful 300 watt MOSFET amplifier with a large toriodal core transformer and a pair of 10,000 uF capacitors in the power supply.  There is also space to add an additional pair of 10,000uF capacitors to make a total of 40,000uF for the one mono channel.  That's a lot!  Take a look into your main power amp and see what you have for each channel!  This means that the output impedance of the amplifier is reduced even further which will increase the damping factor.  Even without the extra capacitors, the damping factor of the amp is over 600!!  The damping factor of an amplifier is one of the main parameters that effect the quality of bass sometimes referred to as tightness of bass.  The base is tighter (less floppy and fumbled sounding) literally because the amplifier is able to move the large motor of the speaker more exactly (especially when it comes to reversing he direction of movement) because the voltage output by the MOSTET output stage is mostly dropped across the impedance of the speakers voice coil of 8 ohms, rather that burnt in the internal resistance of the amplifier.  So an amplifier with a damping factor of 600 at 8 ohms has an output impedance of 8/600 = 13.3 milliohms!  Compare that to a high end power amp like the Technics SE-A5 that has a damping factor of 140 and you can see how well suited this amp is for a subwoofer application.  The single board of the amp is directly mounted to the large heat sink and this layout is well suited to allow the back of the heat sink to stick through a slot in the back of the subwoofer cabinet.  Also the K4010 Velleman design uses TL072 operational amplifiers on its front end and so has a high quality -18 volt & +18 volt supply line for them.  As these are the same devices that will be used to make the active filter, this will make it easy to power the  filter from the main amp board.  Not only that, but there is some space for a small extra board to be mounted onto the heat sink along side the main amp board, so this will be a perfect location to mount the active filter.  The K4010 is in fact half of a K4020 without the case so the construction is the same as shown in the K4020 manual.  At the time, it cost 134.99 British pounds (about $202).

The Active Low Pass Filter (Rev A)

As you might expect if you have read this far, the active filter I built was based on the REL Stentor design.  Originally in 1994, the version that I built was pretty much an exact copy of the REL but made on a piece of grid style copper clad breadboard that was cut to be 3.9 X 2.8 inches, which fitted into the available space on the K4010 Velleman amp.  The lines shown in black were the connected solder pads on the back of the board and the red lines are where the wire straps that were on the front.  Rather than using a pair of TL072 op amps, I used a single TL074 that had four matched devices in a single chip and left one of them unused.  I connected up two rotary two pole switches like the REL and connected them to the board with screened multicore cable.  Once the system was working, by trial and error I found that I needed a crossover frequency of just 17Hz where R1 = R2 = 548K and C1 = 24n4 and C2 = 12n2.  Of course, the REL Stentor version did not support such a low crossover frequency, and so I had to add an extra 10n and 20n capacitor (two 10n in parallel) afterwards.  To make the lowest position on the switch make C1 = 24n2 and C2 = 12n2, I setup the connections to the switch so that the necessary capacitors (2n2 & 4n4) were connected in parallel with the 10n and 20n capacitor.  This now makes the whole chart look like the one shown and causes a bit of a gap in between 22Hz and 30Hz which makes no difference in my case, but these issues are resolved in my more recent Rev B version.  You could easily arrange the capacitor values to give any size steps you may need.  I also added a couple of 9.1 volt zener diodes (shown as BZX 95C 9V1 above) and 270 ohm resistors to form a simple regulator for the TL072 power supply at 9.1 volts.  In fact I have since found that these are not necessary and the TL072 can be connected directly to the +18 & -18 volt lines on the K4010 Velleman.  So just put a wire strap in that place and omit the 270 ohm resistors.  The diagram here should be sufficient for you to build your own version if you wish.  Let me know if you have any questions!  At the time, this board cost almost nothing to make!

The Cabinet

 For the cabinet I followed the same general layout as the REL, that is a down firing woofer with the bass reflex port in the baffle with the driver and the amplifier mounted through the rear of the cabinet.  As I needed some information about how to work out the porting, I purchased a book called "Eminence High Power Loudspeaker Enclosure Design and Construction" (ISBN 0 9518252 1 6).  This contained all I needed to know and only cost 9.95 British pounds from Maplin.  This book may or may not be in print any longer, but as you can see from the cover, it has some interesting designs.  It has various mathematical information and graphs and charts that enable you to work out the diameter, length and number of bass reflex ports needed in the cabinet for a given box volume and a required resonant frequency.  There are also various alignment charts for SBB4, QB3 and C4 alignments that enable you to work out what frequency to tune the cabinet to based on the drivers Thiele & Small parameters.  This is all done with rulers and pencils in an old fashion slide rule style!  In the end, I just decided to tune the cabinet to 33.5Hz to match the resonant frequency of the driver and this resulted in a box as shown in the table.  These days there is various software that can work all this out for you very accurately like Eminence Designer.  In fact I now know from using Eminence Designer that my box is actually tuned to 32Hz, rather than 33.5Hz, but the charts in the book were pretty close!  I built the cabinet from heavy & dense 3/4 inch MDF.  As far as I can see, the kind of MDF you get from Lowes or Home Depot in the U.S.A. is nothing as dense at the stuff you can get in England, which is more like HDF and is much heavier.  All of the sides except the baffle were double layered making the sides 1 1/2 inches thick and the whole thing weighs a ton!  I also put a 4x2 brace across the baffle in between the speaker and the ports and another under the top of the cabinet in the same location.  Then in between those two braces I added another 4x2 to make an H shaped brace.  In other words, it's built like a "Brick Shit House" as they say in England!  Then I covered all the outside edges with an L style cross section molding mitered at each corner.  Then I fitted four big heavy feet that I got from an old dresser.  Next the whole thing was painted with a furniture grade black satin Teflon paint and I must say, it looks great!  Where the molding goes around the top, I put an extra strip of flat molding under the corner molding so it is raised more so that there is sufficient depth for the glass top to sit, although now ten years later, I still haven't got around to getting that glass top cut!  Where the amplifier heat sink fits through a slot I cut in the back, It is secured with small plastic cube blocks intended for fitting shelves and then the small gap around the heat sink is sealed with black silicone sealer to prevent any leaks.  In fact I did test the subwoofer before sealing that gap and you wouldn't believe the howling sound that air makes as it goes in and out of the slot!  This shows that there is good reason why the reflex ports must be of a certain minimum diameter to maintain port air velocity at a sub sonic level!  All of the wood, moldings and paint came to just under 100 British pounds (about $150).

Response Analysis

Now let's have a look at the overall response of the system.  The graph shown in red here is produced by Eminence Designer from the Thiele & Small parameters of the original Big Cat driver that I got from the 1994 Maplin catalogue, and is zoomed in on the region from 5Hz to 2000Kz as we are not interested in anything beyond that for a subwoofer.  This response curve is based on a mathematical calculation of the driver parameters combined with the actual box and port design.  As you can see, the design gives a soothe Butterworth curve with a corner frequency of about 55Hz and then drops smoothly down to 10Hz, so this is the response you could expect if you just connected the speaker to an amplifier without a filter.  Next, I plotted the green line that shows the response curve of the Active Filter which as I said above, I ended up setting to 17Hz.  So as you would expect, the corner frequency is 17Hz where the level is -3dB and after that the level decreases a further -6dB (second-order) for every octave being -39dB at 400Hz at the limit of this graph.  Next, I plotted the blue line that shows the aggregate effect of the other two lines which is what we will get when the active filter and the speaker are in series.  As you can see, what you get is in fact a band pass effect with a Butterworth response at each end and corner frequencies of about 35Hz and 80Hz with the pass band being flat at -18dB.  This is quite a surprise to people who think they have their subwoofer set to a very low crossover frequency and that that frequency alone dictates the response of the entire system!  Clearly this is not so.  By adjusting the volume control to give the gain stage of the filter something approaching 18dB of gain brings the flat pass band pass portion up to the level of the other speakers in the system and this is close to the optimum setting although the bass amp in the subwoofer will be working much harder that the amps in the rest of the system.  So what kind of passive crossover has a gain of 18dB?  None of course!  And this is why active subwoofer systems can take bass where there has never been any before.  It makes no difference how big and powerful your main speakers are, it just isn't possible to compete with an active subwoofer!  So overall, what we have here is a subwoofer system that has an upper corner frequency of 80Hz with a second order Butterworth response falling off at 6dB per octave.  It would be possible to increase the maximum music power handling of the system by employing a high pass shelf filter in the region below 8Hz or so and this is definitely what the REL Stentor could do with.  However in the case of my design, the speaker is so big and powerful compared with everything else, that I have never noted any issues with the speaker cone popping although this is clearly a possibility if the system was to be used in a large venue, but for domestic use, it's fine as it is.  The Cone Displacement graph above shows the actual amount of cone movement, and as you can see, this rises sharply at frequencies below the resonant frequency of the driver.  Another point is that the existing design is also capable of reproducing deep subsonic pressure waves especially when standing in a room with wooden floors.  This is the sound that you "feel" when you see those big explosions on THX films or listen to music like R Kelly and this ability would be reduced by employing a high pass shelf filter.

The Sound and Current Usage

No words can describe how blown away I was when I first powered up my new homemade subwoofer!  Although it took a couple of days of messing around with the filter settings, once it was set, it sounded simply fantastic!  The bass was smooth and tight like nothing I had ever heard and it went way down to the center of the Earth!  At the same time, music that did not have much or any deep base was totally unaffected by the subwoofers presence and there were no negative coloration effects at all that I could note!  It blended perfectly with the Velleman K4000 & Quad ESL-63 speakers that I use as my main front stereo system.  I quickly decided to test it with the Peter Gabriel Real World tour concert and found that at last I had a system that could reproduce the proper sound of the concert without blowing up!!  THX movies where of course great fun with bundles of deep base that were previously unnoticed.  An interesting point to note is that the Quad ESL-63 speakers are not exactly noted for their bass, yet I had to have the crossover frequency of the active filter on the subwoofer set to a very low 17Hz to balance the system, and I found myself setting the REL Stentor in the same general way.  This indicates to me that other people are either using very small front speakers or have their subwoofers set at too high a frequency and are overlapping a lot with their main system.  I never ever imagined that the sound could be so good and that I would be so pleased with it as I only intended for it to be an experimental prototype.  I thought that I would be driven to continue experimenting and messing about with it and that it would take at least months if not years to get it right!  Apparently not, and the same design with the original paint (and still no glass top) is still working in my current system at Chester River today!  This is probably due in part to using absolute top quality components and paying particular attention to design details and build quality and I had great fun doing it!  Still most people who don't know how speakers really work (and even those that do!) are intrigued about how good it is and find it difficult to grasp that I actually did design and build this thing myself!  Seeing as it all cost little more than 400 British pounds in 1994 (about $600) including the electronics, speaker, wood and everything and the REL Stentor cost 1595 (about $2392), it just goes to show how much meat you actually get in a commercial product with my version costing only 1/4 the amount and being infinitely better!  Over the years there have been many times the subwoofer has accidentally been left switched on for days at a time and it has been subjected to some extreme temperatures while in storage on some occasions.  Even so, it has never ever failed to operate perfectly and I'm sure it will run long and loud for many years to come!  More recently while setting up what I intended to be a THX system at Chester River, I discovered what I consider to be a fundamental flaw in the THX specification (at lease for stereo music use), and due to this I had to build a new active filter in order to properly integrate the subwoofer with my new Marantz AV-9000 preamp and this is described here.

Pictures

The internal pictures show the newer Rev B version of my active low pass filter & mixer that is described here.