How to build a truly good conventional box loudspeaker?

[keyser]

Like all of you hobbyists out there I am very much concerned with the following question:
How to build a truly good conventional box loudspeaker? A lot is thought and said about the subject, but not everything that is thought to matter actually matters, and undoubtedly there are some things that do matter but are still never even thought of.
I hope that in this topic we can get to agreements about what things most certainly matter, which matter less, and which might not matter at all. In most cases I value measurements above subjective descriptions, because everything we hear is measurable and can be explained with what we know about physics.

What do we have? A couple of loudspeaker drivers, a crossover and a box.

Before discussing how to build a good loudspeaker we must first decide on what actually makes a good loudspeaker. I’ll start off with the most obvious parameters, hoping others will add more and/or correct me if I am wrong.
The first parameter would have to be frequency response. Flat within 4 dB’s from 40 – 20.000 hz would be nice, but an even flatter and deeper stretching curve would be even nicer.
The next is distortion. All kinds, but especially higher harmonics and IMD, should be as low as possible. Less than 1,5 % of total distortion from 200-20.000 at normal room listening levels is pretty good. Less still would of course be preferable.
Third, the radiation pattern. It think it should be as uniform as possible, to resemble the radiation pattern of most musical instruments. For the lowest frequencies we see a four pi radiation in free space. At higher frequencies an increasing directionality is unavoidable.

Say you have bought yourself those nice drivers with a flat and broad frequency response, low distortion and a nice radiation pattern. You still need the right crossover to direct them. A good crossover should let the drivers work in their appropriate frequency domains and corrects imperfections in their response and/or impedance.

Those nice drivers and the crossover lastly need a box to put it all in to. A primary function of the box is to prevent the 180 degrees out of phase sound radiated from the back of the loudspeaker drivers to interact with the sound that is radiated into the room. Ideally all of the sound that is radiated inside the box (which is just as much as that which is radiated into the room!) should be transformed into heat. However, a lot of what we hear from our loudspeakers is radiated from the loudspeaker enclosures, instead of the drivers.
Another function of the enclosure is to prevent the back waves from the loudspeaker drivers from getting back to the driver membrane, mixing up with the current signal and thus distorting it.


Summarising: The perfect speaker has a response that covers every frequency the amplifier might offer to it, has no distortion at all and I think has a complete 4 pi radiation pattern in full space and no sound is radiated from the enclosure walls.


Now: How to build a truly good conventional box loudspeaker?
I hope many of you will contribute in this thread. It is my intention to summarize the most important points w’lle find and agree on in this thread and edit them in this first post.

[keyser]

My personal view at this point.

Every good speaker design starts with good drivers. As I said in my post above, we need good frequency response. I’ve heard some very good 2-way speakers, but they never had either the tight, deep bass or the clean midrange of a good 3-way speaker. In my opinion a 2-way is always a compromise in this respect.
The individual on-axis and to some extend also the off-axis response curves should be flat in the region in which they will be used. Small dips are not too critical. Peaks are more audibly objectionable. In the crossover these can usually be corrected for, but it would be better to avoid them altogether.

The drivers should produce as little distortion as possible. I think distortion (once having accomplished a reasonable frequency response) is the very most important parameter in loudspeaker design. Some driver sins can be corrected in the crossover, but not distortion.
Stored energy should also be as little as possible. Our ears are most sensitive to resonances and too slow decay of frequencies in the mid frequencies. Strong resonances in the treble can be very disturbing too, though. For the lower frequencies audible slow decay is usually primarily caused by resonances in the room and not by the woofer itself. Severe driver resonances should always be kept way out of the pass band and sometimes also be notched out in the crossover.

The drivers should in my opinion be crossed as steep as would logically and economically be possible. Low order filters have little or no advantages in my opinion. High-order crossovers enable the designer to use the drivers in their appropriate frequency band. Distortion can be kept much lower and the loudspeaker can be much more dynamic: i.e. tweeters and midranges that don’t have to dig too deep have more headroom in their main operation band. Also, the off-axis response can be much more uniform if say the big woofer doesn’t overlap the tweeters’ region too much.

The box should ideally fully isolate the sound radiated inside the box from everything outside it. Complete isolation is so far impossible. However I’m still not really sure about how to accomplish ‘very good’ isolation. This is one of the most challenging parts of loudspeaker DIY. Bracing works well below about 100 hz, but above that frequency the braced but empty MDF enclosure is a complete sieve for the sound. Also, making the bass box too rigid might cause a high Q resonance in the box at higher frequencies. I am not sure that making the box as rigid as possible is always advantageous.
To absorb those higher frequencies that are not killed by the bracing, I think all we can do is fill the box with damping materials. I hear glass fibre wool is a reasonable quality sound absorber. If anyone knows of a better material to convert sound into warmth, please let me know. I hear constrained-layer constructions work really well. I’ve never seen measurements in relation to loudspeaker construction, but I believe it may be very good (google for ‘constrained-layer’ and you’ll find a truckload of information).
The internal construction of the box should minimise direct sound reflections to couple back to the drive-units. B&W works with exponentially tapered tubes in their nautilus designs. For DIY this approach would be very costly and time-consuming. Some other constructions are to accomplish the same goal are easier, like a pyramid shaped construction behind the drivers.



Summarising:
I would suggest using low-distortion drivers with a flat response with high-order crossovers. The box would probably be built using constrained-layer techniques and with many braces. Internal constructions should be made to guide the first reflections away from the drivers. I would also use a lot of dampening materials inside the box.



More later. All your couple of cents are most welcome .
Many points I have still not mentioned. I hope some of you will add a couple.

[Nanook]

Well, ultimately I think the drivers should be capable of covering the frequencies being assigned to each of them.

and lower distortion drivers make sense.

I however do disagree that higher order cross overs should be considered as superior--they do create very nasty phase characteristics--that's why (AFAIK) lower order, lower slope xovers often sound superior. If you really want to do this, go active. If not active, find a tweeter with say an fs of 200Hz, a woofer with rolloff starting at say 6-8kHz and use a 1st order cross over at 1.5k

As far as ultimate enclosures there are so many ..start building. There's the "brute force" approach (heavy), the lightweight minimal stored energy approach (Celestion Aerolam SL 700s come to mind), a happy medium (Baltic Birch), granite enclosures (Eggleston?), etc. or the good materials, good jointery,"regular" box approach--Proacs come to mind.

Good enclosures can be made of pretty much any material. The technique of construction and jointery can have a greater effect.

[454Casull]

1" compression driver on OS waveguide directive down past 900Hz
15" pro woofer, low-passed at ~900Hz (same as the Summa design)
2 18" pro subs, sealed, push-pull (not ideal as far as aesthetics goes), side-firing to reduce height, low-passed at about 150Hz or so, high-pass at ~70-80Hz

If "conventional box" includes horns, then finish that off with a bass horn like the DTS-20.

All in all, should be good for peaks of 110dB outdoors at 4m with relatively low distortion, given sufficient amplification. The room NEEDS bass treatment though.

Generous radii should be specified around the waveguide to reduce diffraction and distortion (according to Dr. Geddes, IIRC).

There's been some discussion lately about reducing enclosure re-radiation... if you've got the resources, just cast a huge box of steel and call it a day.

[singa]

Hi,
I am no expert but someone like you who wants to learn and know about this subject.While the theory and experiences has been published,discussed and realised in practical evaluation/commercial realm ad nauseam by the definition of your topic, the conclusion is that it has failed because the fundamentals or important things are not considered.Really, you say?This may sound like a metaphysical riddle but I assure it is not ponder upon this:

1. What is sound? Just electromagnetic wave?
2. What is Pi or 3.142?
3. What is the golden ratio?
4. What is the transmission medium of sound?
5. What is the impedance of air?
6. Why does sound difract?
7. Why do odd harmonics sound unpleasant?
8. Every thing has a resonance inlcuding humans.
9. What happens when 2 or more frequencies are mixed together.
10. How does the human ear detect or percieve sound?
11. What are tones and it's harmonics?
12. Why do some tones sound pleasant and others are not?
13. There are no perfect speakers,crossovers,damping materials and certainly no conventional boxes.
14. For now everything is a compromise.

[omni]

Keyser, Glass fiber wool may be a reasonable quality sound absorber, but I wouldn't use it... There is one important reason not to use it. Fiber glass is melted spun glass, similar to cotton candy, in its production. Those tiny glass fibers break apart and float around in the cabinet with all the air movement inside the cabinet. These fibers can ultimately make their way into the voice coil assembly, then poof.............distortion, and inevitable destruction of the driver. Singa, I hear ya man.................Lets not forget about the ART in speaker building...... I too, have been venturing into another design after several years of some earlier attempts, and am finding great joy along the way, but frankly have grown tired of all the analysis paralysis I have tortured myself with. I have chosen to adopt good acoustic principles, seek counsel from members on this forum, whom I can relate to, and attempt to blend these experiences into a design I hope to be pleased with. Many compromises along the way have been encountered, both acoustically and financially. I think this is the thrill, for me, along with hopes that the end result will meet my expectations............... I may even be pleasantly suprised.........If not ?........ Tweak..................Respectfully.........Omni

[keyser]

Thanks for your replies.
How do you build a good box?
I've built my current mid/high speaker boxes which measure 43 * 24 * 36 (in cm) out of 1" thick (25mm) MDF and used one horizontal and one vertical cross brace. The baffle and the back are both 2" thick. The box is completely filled with sheep wool.
However, at first the box strongly resonated around 300 hz. It took me quite a while to find out what caused this resonance.
To flush-mount the woofer, I had routed the rebate too tightly, so the woofer frame got tightly stuck between the wood around it. When playing a tone around 300 hz, the frame must have been moving back and forth, while actually stuck between the wood. This must have caused the buzzing, for once I had mad the rebate just a bit wider, the problem was over.

However, it is remarkable how much sound still passes through the walls of this enclosure! I initially thought it was possible to bring its level down as much as for it to become practically inaudible, but no. At low frequencies the box doesn't seem to resonance much, but in the (most critical!) mid band it can be felt vibrating, even at moderate sound levels.

The constrained-layer technique I mentioned before is said to be a very good way of minimising the sound that comes through the enclosure walls. Has any of you ever tried it?

[69stingray]

Singa- are you an major in psychoacoustics ? Great questions.

Flat within 4 dB’s from 40 – 20.000 Hz would be nice,

I can only hear from 40 - 14 kHz, so I kind of agree. Not sure where the 4 dB threshold comes from....no experience to talk to.

Bigger question....what is the goal of this loudspeaker? Is it simply a series of transducers designed to accurately reproduce the input signal? If so, that may not be a "nice" sound loudspeaker. I bet if the “ultimate” loudspeaker was designed, it would not be up to par. Rule flat response curve (@ listener’s position) with zero distortion – what would that sound like? If you replayed a recording of an “un-amplified” concert, would it sound the same as in person?

Interesting article:

What Should I Look for in a Loudspeaker - An Engineer's Perspective

[mashaffer]

I doubt that you will get much agreement on what makes a good conventional box or even whether a conventional box can be good.

For me the criteria are a little different from yours. The entire frequency range needs to be there but an arbitrary anechoic response specification is not necessarily a useful thing given that we are going to be listening in a real room. My frequency requirement is 16Hz-15KHz with the extremes no more than about 6dB down from the midrange(an additional 8va lower for home theater). but this is total system response including the subwoofer which is almost certainly necessary.

We want to avoid crossovers in the 65Hz to 4KHz range if possible and any crossovers should be low order if possible. Idealy I would like to avoid any crossover components on the midrange or at least have those componants spec'ed at least one 8va beyond the natural rolloff of the driver (or driver baffle system). Naturally the midrange will have to be large enough to produce this range of frequencies with adequate headroom and moderate excursion.

In some applications we can compromise the midrange breadth for the sake of controlled directivity and extra efficiency where that is important (e.g. outdoor, large room, or rooms with undesireable reverberation characteristics). In such applications the midrange horn could be limited to say the 200Hz to 1.5KHz range for example.

The drivers should be of rather high effeciency to maximize dynamics and reduce compression and heating of the voice coils. The midrange driver/amplifier/enclosure system should provide a smooth and even response without the addition of filtering components. Tweeters should have adequate directional control for the room in which they will be used and they should be mounted so that the acoustic center is in line with that of the midrange.

If the woofer uses a tuned enclosure it should be tuned low enough to push phase effects as low as possible. Also, we don't want flat anechoic response to a cuttoff and then sharp drop from there. Far better to take advantage of room gain and have a gradual rolloff leading up to the cutoff frequency. However, for outdoor use the flat up to cutoff approach would be appropriate

If a box is used it should be large with a wide baffle and properly braced to damp box colorations. This minimizes box interactions with the driver and minimizes the baffle step problem. The woofer should have its own seperate enclosed space so that its backwave does not effect the midrange. A rear facing BSC woofer is preferable to a filter circuit. In addition to the size of the box, stuffing should be used to reduce the amount of backwave that is reflected back out through the drivers.

There; does that sufficiently stir up the pot?

mike