Updated 7/5/2023
Most of the designs on these pages first took shape around 18 years ago, when I was making analog crossover boards with ExpressPCB. I had latched onto the idea of making a loudspeaker testbed that would allow comparing drivers under controlled conditions. I had blocked out a design for a 6-way loudspeaker in which the drivers could be switched with others, while using steep filters and volume compensation to ensure a “fair” comparison. This would be a “listening test” to help myself and others understand what matters most in loudspeaker design.
What started out as a rather ambitious 6-way analog crossover with volume controls and relay switching ended up as a series of DSP boards with constantly evolving software for control. Most of those designs are documented in these pages. It’s been a long ride with a lot of experimentation and learning, and it’s time to reflect on what’s next and how this journey will end.
Desktop Speakers are “easy”
There are many DSP-based designs on these pages for desktop listening, using a PC music player or else Wi-Fi or Bluetooth streaming sources. DIY’ers spend a lot of time looking at a computer screen, so this has been an important area to focus attention for this website.
For a desktop speaker, we usually are not concerned as much about stereo imaging, wide dynamic range and full 20Hz-20KHz response. The primary issues in designing for the desktop are clarity and accuracy, with low distortion and minimal coloration. It turns out that these goals can almost always be met with high quality drivers that are properly band-limited with active crossovers, powered by low-distortion amplifiers. There are many 3-way active crossovers on these pages that can be used to build excellent desktop speakers that can be very satisfying for daily listening.
Since technology constantly moves on, there is still a lot of room for innovative hobby work for this “genre” of speaker design, as more and more low-cost system-on-a-chip (SoC) hardware becomes available. Most Bluetooth chips and many class-D amplifier chips have basic DSP circuitry that a clever designer can program to make high quality desktop speakers with minimal circuitry. So, there will still be DIY community interest for new desktop speaker designs. But for this website, we’ve reached the end of this type of design.
Designs for the Listening Room or TV
The “serious” loudspeaker that goes in the listening room or doubles as the TV speaker, has largely been ignored in these pages, but that wasn’t intentional. The goal all along was to show that very high-quality multi-way loudspeaker systems with Wi-Fi streaming could be made rather easily using high-order crossovers. However, the pages got bogged down in a few too many variations of the ADAU1701 and didn’t transition to the ADAU1466 early enough to address more complex designs.
But another reason for not detailing some larger 4 or 5-way designs is that the conventional multi-way speaker box doesn’t seem to be the most satisfying solution for large rooms. The conventional approach for large “Hi-Fi” speakers is to use multiple drivers mounted on a large box, with a crossover to constrain the output of each driver to frequencies below which the driver becomes directional. Even though the drivers themselves are (mostly) non-directional, the reflections from the baffle edges introduce phase delays that make it easier for listeners to “locate” the source of the sound. Also, the drivers are not co-located, which creates phase information that the ear uses to locate the source. As a result, most conventional speaker boxes tend not to “disappear”, but instead, can be localized and recognized as the source of the sound.
The large multi-way “speaker box” designs can sound very impressive, with low distortion, high dynamic range and high accuracy. However, they tend to “stick out” and don’t feel integrated with the room. The bottom line is that conventional speaker boxes are not satisfying to some designers who expect better imaging or greater ambiance and spaciousness.
Working the Room
It’s a common assumption that the ideal loudspeaker would be a vibrating point source with enough output to fill a room. However, this ideal is impossible to achieve in practice, and instead we are forced into many compromises that impair the loudspeaker’s ability to create a solid, credible stereo image Some researchers have argued that recorded music is not necessarily best reproduced using “compromised” point sources, but that instead we must address the role of room reflections and manage the beamwidths to achieve the best sound for a given type of room and recording. So, what are the key issues, and what options should we consider in a quest to design a “better” loudspeaker?
CBT Line Arrays
The most common technique to address improved imaging is the CBT, or constant beamwidth transducer. This technology is used in waveguides, but it is also a feature of the curved line array. There is one “electronically curved” line array described on these pages, but a second-generation version will be developed that will allow greater control of the curvature, driver shading and other parameters to better evaluate this type of CBT. There is still a lot to learn from type of speaker, and this new design will be a fun way to explore the CBT line array.
A characteristic of the line array that I find most impressive is that the sound falls off as 1/r rather than 1/r^2. That is, the sound pressure at twice the distance from the line array is 1/2, whereas the conventional speaker would be 1/4. This radiation pattern is very effective at filling large rooms, and creating a stereo image that remains constant as the listener moves about the room. And it is one of the keys to making the speaker itself seem to “disappear”.
The CBT line array requires either a curved baffle or else electronic curvature, and both options are difficult to implement. The approach taken here is to use many small amps with many channels of DSP (usually 24 separate channels) to provide electronic curvature and shading for a multi-way line array. The second-generation version that is being worked will still be a prototype, but it will be much closer to a commercial product. It will be nice to see a complete solution for multi-way CBT line arrays that a DIY’er could use for a wide range of drivers.
Dipoles
The most common technique for increasing the “spaciousness” or ambiance of a loudspeaker is to manage the reflected sound from the driver by using a dipole radiation pattern rather than the monopole radiation from a conventional loudspeaker. The dipole speaker is typically associated with the speaker designs made popular by Siegfried Linkwitz. It seems imperative to have at least one dipole design on this web site.
As Earl Geddes points out in a notable thread on diyAudio.com, the goals of better imaging and improved spaciousness “are mostly incompatible… What enhances one degrades the other. One has to decide on what they want from their system.” However, the experimental speaker described in the article “The Return of the Marthas” has characteristics of the CBT plus the reflected sound properties of the dipole speaker. This speaker should prove to be an interesting design to evaluate, as it will allow real-time switching between dipole and monopole modes, with all the frequency compensation required for each mode.
Other Solutions to address Room Issues
There are many other technologies employed to improve imaging, improve “spaciousness” and improve speaker integration with the listening space. However, life is short, and we can’t explore all the interesting technology challenges and opportunities that keep emerging. Once the second-generation line array is complete and the write-up on the Marthas is complete, our speaker-building journey will end.
Calling it quits
At some point in your life, you need to start wrapping up old projects so that your kids won’t be cursing you as they clean up your piles of parts after you pass away. It feels like that point has been reached or is coming soon here at Audiodevelopers. This site has been active for 16 years, but we are coming to the end of this long journey. New designs and additional code development will need to come from other sources. There will still be some additional projects documented here, but they will be ones that only require a small footprint, such as writing or cleaning up code that can be passed on to a new generation of experimenters.