The enclosure

I bought the Kevlar drivers and began figuring out how to take measurements so that I could design an enclosure for it. I wanted it to be a floorstander, because Angshu had told me that a standmount is a wasted optimisation anyway: a good standmount needs a stand, which finally eats up as much floor space as a floorstander. There is no real "bookshelf" speaker for high quality sound: only Bose Acoustimass speakers and Korean combo boomboxes can be kept on bookshelves without any worry about losing sound quality; they are too mediocre for their placement to make a difference.

So I knew I wanted an MTM floorstander. I began setting up Speaker Workshop to take impedance measurements. I read all the documents and tutorials and got thoroughly frightened reading about the Wallin Jig. Angshu told me to forget all about it and simply build a Claudio Negro impedance jig. I did, and I happily use it till today without a Wallin Jig in sight. I took all the impedance measurements in free air, got my trusted carpenter, Hiralal Sharma, to build a box for Vas measurements, and then measured the Vas. It may sound like all this took two days, but in fact it took more like two to three months. Things didn't work initially. See my thread on Speaker Workshop forum and on diyaudio about the problems I faced and the path I had to take to get simple impedance measurements to work right. This was August 2005. A lot of this time went in building and understanding the impedance jig, and a lot also went in trying, then rejecting, the added-mass method of Vas measurement. It was only after all this that I got Hiralalji to build the Vas box for me, so that I could use the sealed-box measurement approach. It has a front baffle which doesn't need to be screwed on... I just clamp it to the box using C clamps. I have a thick closed-cell-foam gasket all around the edge for air-tight sealing. This clamped-baffle idea is, as usual, from Angshu. It works beautifully; I'm on my third driver model's Vas measurement by now.

Three of the four drivers had pretty consistent readings, but the last one was a bit out. Its Fs was almost 20% lower than the others. I checked with Angshu and decided to ignore it. Its Vas too was out by almost 20%. When the others gave me Vas readings of about 36L, this one reported 30L. I finally went with a Vas of 35L.

I began looking at various programs to do my box modelling based on the T/S parameters. The box modelling features of Speaker Workshop were okay, but didn't seem too easy to use. I had read Vance Dickason 5/ed, and I was aware of the different alignments for a vented enclosure; SW seemed to support only a small subset of these. I tried WinISD, and found the graphic design of the UI off-putting. I tried Unibox, and loved it.

Unibox is amazing, because it eliminates the idea of vented enclosure alignments altogether. It re-visits the fundamental mathematics of Thiele, Small and others, and calculates the box response from first principles. The alignments like SBB4 and QB3 were shorthands methods designed by the early pioneers to make box design easier in the days when computers were not available by the six-pack. In today's world, vented enclosure alignments seem to be an anachronism. So, I saw this first-principles approach in Unibox, liked the UI, and decided to stick with it.

I tried various box volumes, and discovered that if I select a volume larger than a certain upper limit and set a pretty low box tuning frequency (Fb), then the vented enclosure begins to resemble a large sealed box, but with better extension at the absolute low end (30Hz and below). Instead of a traditional maximally flat SPL curve followed by a clean knee and an 18dB/octave drop below that, I now got a slowly drooping curve, somewhat akin to a sealed enclosure with a Qtc of 0.6, followed by a knee, leading to a sharp drop in the graph. So, I could now create a model which would start dropping SPL from somewhere between 80-100Hz, and drop very gently till about 30-35Hz, after which the knee would come. From what I'd read, this gentle drop seemed the exact converse of room gain, so I thought it would be a great idea to build a box with this sort of unconventional alignment to offset the room gain and get relatively flat, very extended in-room bass. I had some doubts about all this, so I initiated a discussion on diyaudio about it, and got generally encouraging responses.

So, I went with a 60L enclosure volume for my pair of drivers. Now that the details of the volume were finalised, I had to decide on the actual dimensions and construction of the box.

One of the universally held beliefs about speaker enclosure construction is that they are made of MDF, just like apple pie is made, essentially, of apple. Some veterans talk of Baltic birch plywood, but they are a very select minority. My readings about what makes good enclosures, plus inputs from Angshu, convinced me that enclosure deadness was extremely important, perhaps more important than driver characteristics (within limits). Angshu in fact believes that most DIY enclosures radiate so much sound that their total radiation exceeds that from the driver at some frequencies.

So I decided that the choice of material alone would not fix the deadness problem; this needed much more effort. And if this sort of care was taken, then the choice of material would become secondary. So, to prove my theory, I decided that I would not use MDF for constructing the Asawari. I'd use good quality ply. And I'd use techniques of damping and vibration reduction fanatically, so that I'd get an enclosure deader than most DIY speakers (or commercial mid-fi speakers) built using MDF.

I decided to use 18-20mm marine plywood for the enclosure. This is less expensive than good MDF, and is available at all timber dealers in India. Marine plywood is better than commercial ply because it has little or no pores or airgaps --- it's simply better constructed. If you find A-grade commercial plywood, that'll do too, I think, and will be about 20% cheaper than marine ply. And I designed an extensive structure of bracing, to emulate the matrix bracing that I had read about. Angshu had told me about matrix bracing --- he said that B&W speakers are braced this way. He had used it for one of this small floorstanders and had got very good results. I also remember reading somewhere on Seigfried Linkwitz' site that if bracing is applied in the form of a grid with grid sizes of about four inches, then the resonant frequency of the panel will be raised to single-digit kiloHertz levels. I decided to aim for something like that.

I used the same plywood for the bracing as for the outer walls. The matrix I designed resulted in rectangular patches of unbraced wall surface not more than about 8"x6" between braces. And there were vertical and horizontal braces, all reinforcing each other. I used two sheets of ply for the front and rear baffles, and three sheets of ply for the bottom, just to allow a deep bottom plate to drive long screws in if needed. I used a single sheet of ply for the top surface, but fixed a 12mm thick sheet of float glass on it with Araldite, to make it a composite sheet. I have tried to draw the horizontal and vertical braces below. The horizontal braces have two cross-beams, and the vertical braces are rectangles, with a central rectangular cutout. Some of the smaller braces, behind the tweeter, are in the shape of fat capital-H, shown in the drawing as H-brace. These were too short to be made in rectangular shapes, hence the H.

Another thing I decided to do is keep a separate crossover chamber at the rear of the enclosure. Since this was my first pair of speakers, I wanted the full freedom to remove the covering panel for this chamber and pull out the crossover, without affecting the air-tight sealing of the acoustic chamber. On hindsight, I think this was a good decision. In the internal diagram, you can see this chamber, about 16cm deep and more than a foot tall.

I decided to play it safe with the volume calculation. So I built a spreadsheet with the volumes of each item, including the cone size and magnet size of each driver, the volume of each brace, and the volume of the corner columns (pure timber) which would eat up the internal space. Subtracting them from the gross internal volume, I arrived at the net internal volume. And I tweaked the depth and height to get the net internal volume as close to 60L as I could. I tweaked the height and depth in steps of +/- one centimetre, and arrived as close to my target as I could. I finally reached 60.7L. I froze the dimensions at that point.

The front baffle has rounded vertical edges, with a radius of 1" (2.5cm). I built this by using wooden beams of 3"x2" cross section, running the full height of the baffle on both sides. These beams were shaped as per the corner detail shown below, and one edge was rounded. In the diagram you can see how I am using two sheets of thick ply plus one sheet of 3mm ply plus one sheet of 4mm veneer for the front baffle, and one sheet of thick ply plus one sheet of 4mm veneer for the side walls. The extra addition of the sheet of 3mm ply on the front baffle is for recessing. The tweeter needs to be flush mounted, and I decided that it would be easier to do the recessing by cutting the smaller mounting hole in the thick ply sheets, and the wider flange-fit hole in the 3+4mm of ply+veneer. That way, I got about 7mm of recessing, which was sufficient for the gasket seal plus the thickness of the tweeter flange. I thought this was a smart way to do flush-mounting and was patting myself on the back, but I later saw that Roman had documented it on his Website.

My carpenter Hiralalji began work based on the drawings I gave him. He made the basic box quite quickly with the side walls, top and bottom, and then began working on the "partitions" as he called them -- he meant the bracing. These braces took a lot of time because he had to cut each one precisely to fit tightly inside the shell, and also had to cut out the central portion to make a large hole in the middle of each brace. All in all, he must have worked, with a helper, for a month, to complete the pair of enclosures and cover them with veneer sheet.

This was an interesting project for him because he saw a router for the first time. (This applies to me too in fact.) I bought an inexpensive PowerMaxx router from Lohar Chawl for him to use. He spent half a day just inspecting it, fitting a bit into it, and trying to make test cuts. After that, he gave it his seal of approval and said it was a lovely way to cut circles. The router cost me about Rs.1800 (about USD 40.00) and each router bit cost about Rs.150 (USD 3.00). These bits keep breaking after every few days of heavy use.

The enclosures finally got done sometime in December, and then I got them polished using a melamine polish. This is a thin waterproof treatment, where the last melamine coat is sprayed on using a spray-paint setup. Hiralalji charged me Rs.8000 (about USD 160) as just labour charges for the carpentry work, and the polishing chap took Rs.2500 (about USD 50) for the labour and material of the polishing. In addition, the cost of plywood, veneer sheets, and adhesives exceeds Rs.14,000 (almost USD 300).

I then added Dacron lining on all the internal surfaces. I smeared Fevicol with my fingers all over the internal walls and on all surfaces of all braces. I then took a pillow filled with Reliance Recron -- that's their name for Dacron, I believe -- cut it open, and pulled out the stuffing from it. I slapped on about one to one-and-a-half inches of this silky-smooth white fibre on all these internal surfaces; it seems to hold well with Fevicol. In case you are wondering whether this Recron is different from normal cotton-wool, lay your mind at rest: it is. It feels totally different. Recron feels sort of slippery and synthetic like polyester, and is lighter, silkier, and fluffier than even good cotton. The audible effect of this difference is of course something I can't comment on. Why did I use Recron instead of ordinary wool? Because people on the forums keep talking about using Dacron, and no one talks about using cotton. This could simply be because they get Dacron more easily than good cotton in most Western countries, I don't know. :)

Once this was done, I fixed the drivers on the enclosure using closed-cell foam sheets as gaskets, wired them up internally, brought the leads out into the crossover chamber, and began thinking of SPL measurements in preparation for crossover design. I used multi-strand copper wire of the kind electricians use for wiring up all the equipment in medium-sized offices. My electrical cable for internal speaker wiring had an effective cross-section area of 6 sq-mm, which translates roughly to as much copper as solid-core 9-10 AWG wire, I think. This sort of single-conductor multi-strand copper wire cost me about Rs.40 (about USD 1.00) per metre; I purchased it from my neighbourhood electrical supply shop.

Below are some photos of the construction process. Each is linked to a larger version of the same photo. These photos show the enclosure with top, bottom and side walls, and the braces partly in place. The vertical beams on the two sides of the front baffle are clearly visible, though the front baffle itself has not been fixed yet. Can you see the "H" shaped cross-braces behind the place where the tweeter will fit?

In the following set, there is a well-lit view of the boxes under construction, followed by a shot of the boxes with the front baffles in place and the vertical edges rounded. In the open-box shot, you can see the triple-thickness bottom, single-sheet top, and the way the vertical beams for the front baffle have been shaped with grooves to fit the double-thickness front baffle. The grooving which Hiralalji used for the front baffle is a bit different from what I had drawn in the corner detail above. He decided not to keep the two sheets of the same width; the inner sheet is two inches narrower than the outer one. And in the next shot, you can clearly see the final appearance emerging from the woodwork.

In the next set, you can see the router I bought from Lohar Chawl, and the pile of interchangeable ports I decided to make. I made three pairs of ports, with lengths ranging from 11cm to 17cm. Each port is fixed to a square sheet of plywood, and I can swap ports by removing four screws. Since these ports fire downwards, they are invisible and not finely finished or veneered. I really love the idea of a removable port, and I'll probably use it on all future bass reflex constructions. I also got two square pieces of ply cut of exactly the same size as the rest, to allow me to convert the box to a sealed enclosure if I wanted to.

And here is the finished pair in the living room, with the old Wharfedales behind them. The Asawaris look well finished, but have a fairly boxy shape, due to almost two feet of front-to-back depth. The two inches of gap between the enclosure and the base board are clearly visible: this gap is open on the front and sides, and closed at the back. The 12mm glass sheet as the top plate looks quite smart, I thought, and also makes for a wear-resistant and usable table top.