Why build your own speakers?
Most commercial speakers are built to a strict budget. Considering that the manufacturer gets maybe only 20 percent of the final retail price, and from that he must pay all his costs and still make a profit, it becomes clear that there is a limit to the quality of parts and construction that can be incorporated into his speakers. It is possible to obtain much better performance from self built speakers, especially when spending only a couple of hundred pounds.

Problems with commercial systems
Cheap drivers - many systems use cheap far eastern/Indian built drivers. These can be good value, but there is often a huge variance in parameters (and hence performance) between supposedly the same drivers. It is not uncommon for the tweeter in £300 speakers to cost the manufacturer only two dollars.
Restricted cabinet volume - usually the volume of the cabinet is less than it should be, meaning the bass response is not as good as could be obtained from the driver. The manufacturer does this to make small enclosures that are cheaper to transport and easier for a listener to accommodate.
Poor quality electronics - cheap crossover components and basic internal wiring do not give the best sound quality. Manufacturers often release special editions of sucessful models with nothing more than improved electronics and a different badge.
Undersize port - bass reflex ports are non-linear, and cause compression and distortion even at low levels. The smaller they are, the worse the effect, and commercial designs tend to use small ports to save money.
Acoustic wadding - cheap speakers may skimp on or completely eliminate this essential element.
Cheap cabinet - many speakers have thin cabinets, which resonate easily, and colour the sound. These are cheaper for the manufacturer to build and transport.
Restricted development time - the manufacturer can only devote a certain amount of development time to a particular speaker, to ensure that its sales will cover the development cost. This can often result in imperfect speakers being released onto the market.
Some companies design and build their own drivers, claiming their research and development is the best there is. This is the hi-fi vouge, and in years gone by produced the best drivers on the market. In reality the company is forced to use the new driver across its whole range for a couple of years, in order to repay the development costs. This is obviously not ideal, in that the driver will not be suited to lots of different enclosure volumes and loadings, and that the company is stuck with a poor product if the driver turns out to be imperfect.
Other companies buy units off the shelf from major driver manufacturers. They can recieve a lot of flak for this, but it's misguided. Nowadays it's a better option as it allows the manufacturer to choose ideal drivers for their range, and the driver manufacturer will be able to cover his development costs much more easily (due to his greater specialisation and production volume).

Advantages with commercial speakers
The big advantage with commercial speakers is that the manufacturer has the experience to create a good design, and can afford the test equipment and facilities to properly evaluate, modify and optimise his designs. This is simply not possible for the average home constructor. You also get a chance to listen to the speaker before you spend your money, and decide if you like it or not!

When is it worth building your own speakers?
Essentially when you want good performance but you can't afford to spend lots of money. You may also enjoy the actual building stage, or you may look forward to having something individual at the end. There are no hard and fast rules, but assuming you have the tools, construction skill and time required, cost is a good place to start - how much have you got to spend? If it's no more than a few pounds then forget it - go round the boot fair and pick up something secondhand. If you've got a grand burning a hole in your pocket, then is it worth risking it on something you can't listen to before buying? Probably not. But if you've got maybe 30-200 pounds to spend, then it's definately worth doing.
There are exceptions of course, in special cases. My first rear surround speakers cost me 10 pounds to build - they had to be small enough to be carried in a rucksack along with lots of other junk. My original centre speaker was made from the only magnetically shielded driver I had, and cost me nothing to build as I already had all the parts lying around.

What design do you choose?
I'm not going to even try to tell you how to design loudspeakers; get hold of a book! Vance Dickason's "The Loudspeaker Design Cookbook" is one of the best, and it avoids getting heavily mathematical (although you'll still need a calculator). It's in its fifth edition now, ISBN 1-882580-10-9. I am however going to list what I see as the reasons for choosing a particular design.
Bass loading
Sealed or Infinite Baffle enclosures are the simplest to build, and to design, as they are nothing more than an airtight box. Although they do not produce the greatest level (loudest) bass from a particular cabinet/driver size, they have the most shallow rolloff (12dB/octave) and best transient response. IMHO they produce the best sounding bass.
Ported or Bass Reflex enclosures are more complicated to design. They have a steeper rolloff (24dB/octave) but give the smallest cabinet size and loudest bass for a particular driver size. When well designed they can sound very good, however bass is always slightly blurred as their transient response is imperfect.
Transmission line (and the similar quarter wave pipe) enclosures are less popular. Not much design information exists for these, however there have been several sucessful commercial designs. They have as good a transient response as a sealed cabinet, and possibly an even shallower rolloff, but they are not very efficient and generally produce a very large cabinet.
Bandpass enclosures by definition only have a narrow response bandwidth. This makes them only suitable as subwoofers. Although they generally form very small cabinets, their transient response and efficiency is poor. They are usually used in systems where small cabinet size is more important than sound quality. Not recommended.
Number of ways
I am from the school of thought that the less complicated something is, the less there is to go wrong. In audio this is often true, in that if something is left out of the signal chain then so is its distortion! However in speaker design, if you want to cover the "full audio range" (20Hz to 20kHz) and achieve a smooth power response in a typical listening room, it is usually necessary to use at least a two way system. Crossovers introduce complications and expense however, so if you can put up with a restricted bandwidth then a good quality single way driver could be the way to go. I feel a smooth power response across the full frequency range is of primary importance, so a three way system is necessary. If you can live without lots of deep bass, then two ways is sufficient.
Crossover order
As before, the less complicated, the better. However the least complicated crossover (1st order) does not give a steep enough rolloff to e.g. prevent overloading of the tweeter, so a second order crossover is necessary. Anything more complicated adds too much distortion to the signal chain, and will ring audibly.

Speakers I have built
My main stereo speakers are based around Audax drivers, the TW025M3 tweeter and HT170F0. I intended to put construction plans and measurements of them here, but both drivers have now been discontinued! Audax are replacing the woofer with a plastic chassis version, and I intend to try the new TW025A12 tweeter at some point.

Bargain Basement Audio Tips
A few tips for improving the performance of your audio system for not much money. Some need a soldering iron, but all have been tried and found to work by yours truly...

---