Table of Contents
The sound of higher directivity speakers
Reading this article from Sausalito Audio got me thinking about what it is that I like about horn speakers for hifi. Interpreting "Spinorama" Charts
In that article they point out that narrower dispersion speakers are generally not preferred, with their sound being described as more colored and less open sounding than speakers with wider dispersion. This has been my impression as well. But I have learned to hear past the coloration effect of narrow directivity and enjoy the greater sense of clarity it offers. I noticed many years ago that my initial impression of a good wide dispersion speaker was always that it sounded very open and neutral. After listening for a while the effect would start to morph into a confusing sense of something like white noise being added to the signal. It's not that I could actually hear white noise in the background. It just gave me a feeling like there was something hazy about the sound, and within about 20 minutes this would become fatiguing. I would describe it as a sort of pastel effect on the sound, or perhaps a sense of compressed dynamics. Higher directivity speakers like horns do this much less but at the expense of sounding less open and more colored. The thing is I can adapt to that coloration, and the less open effect can be mitigated with wider stereo speaker placement. So coloration bothers me less than a high ratio of reflected to direct sound in the long run.
Horn Loading is About Reducing Exit Shock
Horns are known to increase the efficiency of speaker drivers. It is also known that for lower frequencies the horn has to be bigger, both in its length and the size of its mouth to properly “load” the longer wavelengths of lower frequencies. So how is it that the efficiency is increased? What is being lost when a direct radiator tries to make sound without a horn or waveguide? The answer, I believe, is that the horn drastically reduces exit shock, or diffraction, as the wave travels forward into the air. From this I am implying that direct radiators are extremely diffractive devices when they are producing wavelengths significantly larger than the cone diameter.
Years ago when I was trying to design horns from scratch, me and my friend were pondering how important the length of the horn is relative to the mouth size? What happens if you have a really long exponential horn with a small driver at the throat vs another horn with the same size mouth and same flare rate, but you cut off a bunch of the throat and put a bigger driver further up the horn? You end up with a shorter horn, same size mouth, and a bigger driver. What if you just get a big driver the size of the horn mouth and throw the horn away? I will argue that you get the same bass response!
Now there is another issue that goes on with horns that use compression drivers. Compression drivers allow a better impedance match to the air so the speaker's voice coil can “feel” the air pressure building up in front of the cone and actually push into that rather than mostly just working against the mass of the cone and voice coil. Compression loading actually changes the electrical impedance of the assembly significantly.
But the horn part, what does it do? how long does it have to be? How big does the mouth have to be?
I did an experiment recently using 1“ PVC pipe as a “horn.” This is a zero flare horn, with a 1” throat and a 1“ mouth, chopped off flat. No roundover. I made it 1 foot long and put a JBL 2426 H compression driver on one end and took some measurements.Having zero flare means this horn has a serious exit shock. But there was enough length to load 1000 Hz, holding pressure in the pipe against the driver. So did I see any efficiency gains over just measuring the compression driver without any horn attached? Sadly, no. It just added the difficulty of the pipe resonating as a result of the exit shock and it's own 1 foot length. I wanted to try this because I've found that I can successfully suppress the resonance with a pre-conditioning of the signal. That makes it sound more natural, but doesn't help with the loss of efficiency.
Comparing the output of the 1 foot PVC pipe to a good waveguide showed a dramatic loss of efficiency, both on and off axis in the lower frequencies especially. The PVC pipe basically didn't help at all.
So in summary, just getting pressure built up in front of the driver with a pipe long enough to maintain pressure down to 500 Hz doesn't really do much good if you just release it without a gradual transition to a larger opening. The opening has to be big enough to reduce the exit shock, and the transition has to be gradual enough to do the same.This is why horns have to have some length. The real target is the mouth size, and the length is whatever it takes to gradually get there. If exit shock is to be reduced it follows that some directionality will be added. The PT waveguide in this experiment is only half as long as the 1” diameter PVC pipe, but it supports the lower frequencies dramatically better. The difference in output level was extrememly obvious as soon as the signal sweep played.
A further observation from this is that any diffractive event is lossy. Ron Sauro told me that, so I know it's true. I just didn't realize how incredibly lossy it can be. A highly diffractive event is highly lossy. So any time a driver is small compared to the wavelengths it is producing it MUST be very lossy. Even if the driver weighed next to nothing and the electrical efficiency was close to 100 percent, I doubt you'd get to 10 pecent actual efficiency at converting electrical power to sound power when the driver is significantly smaller than the wavelength it's trying to procude. Unless, of course, it is assisted with a horn or waveguide.
Speaker Directivity and Room EQ
How do you EQ an exponential horn? That's what this entry is about, although it could apply to any speaker or speaker driver that exhibits increasing directivity at higher frequencies.
Back around 2017 I completed a pair of large corner horn speakers. These incorporated an exponential midrange horn with an exponential tweeter horn mounted coaxially. The reason for this choice was a goal of maximal efficiency and minimal speaker membrane motion, which was taken to be the ultimate cause of poor audio quality. The mid-horn was intended to operate from 200 Hz up to 2000 Hz, with the tweeter horn operating from 2000 Hz on up. This was a loud and powerful combination that could produce enrapturing sound at times. But it could also annoy the ears, and I assumed a number of causes, not the least being it's poor off axis response. The mid horn would beam at 2 kHz, and then the tweeter above that would have much wider dispersion before again beaming up above 4 kHz or so.
After much frustration with this horn assembly I decided around 2020 to retire the coaxial tweeter horn and employ a large constant directivity waveguide for the tweeter instead. To match directivity to the mid horn required a much lower crossover for the tweeter - about 600 Hz, which left the mid-horn running a single octave from 300 to 600 Hz. This solved much of the tonal irritation I was perceiving but not without some loss of ultimate impact and especially efficiency. Getting the tweeter down to 600 Hz took a lot of EQ. This setup also ruined the look of the speaker because placing the big waveguide above the mid-horn was mechanically awkward.
Recently a thought occurred to me about how to target the room curve for a speaker. I decided to measure the output from the mid horn at the listening position with a long IR window to see its native cumulative in-room response. To my surprise, this turned out to be a remarkably straight line response from about 300 Hz up to 6000 Hz with a roughly 3 dB / octave slope. That's a lot of slope for a room curve, but I decided to play some music with just the mid operating by itself full range, and to my surprise it sounded remarkably natural and quite easy on the ear. That was encouraging, so I decided to match the tweeter and woofers with crossovers at 300 and 6000 Hz, their levels adjusted to match the midrange nicely so that the rather steep 3 dB/ octave slope was extended at the listening position. This was bass heavy, but not as bad as I expected. I decided to EQ the bass response below 300 flat, so zero slope below 300 Hz and 3 dB/octave above that. Now I had some remarkably nice sound, and the big constant directivity tweeter horn wasn't really doing much anymore.
Encouraged and inspired further, I decided to re-activate the coaxial tweeter horn. After listening and making some adjustments I settled on a crossover to the tweeter at 3500 Hz. Some room EQ that's always required between 100 and 400 Hz was then applied by ear, and then I heard some great sound from this crazy mid-horn / coaxial tweeter arrangement that I had thought was a write off! So what's going on here?
I decided to look again at the nearfield measurement of the mid-horn to see what the curve looked like. It is also sloped downward with rising frequency, but at a much greater 6 dB/ octave. What I used to do with this horn was EQ it flat nearfield and then try to adjust from there. But what does that do at the listening position? It gives a 3 dB/ octave rise at the listening position. Now that isn't likely to sound good. I've never seen a room curve recommendation with a rising slope in any frequency band. Obviously EQing that horn flat nearfield was not a good approach.
The frequency response slope decreases with distance on an exponential horn, which means we can't eq the exponential horn flat up close. It needs to have a slope that won't be reversed at the listening position. But why does the slope decrease with distance? This is a simple fall-off rate issue. Sound that is more directional falls off at a slower rate. So as you move away from the horn, the less directional lower frequencies fall off faster than the higher frequencies. But isn't total energy in the room still the same? Yes, to some degree. The bass spreads out over the room more quickly and eventually bounces back to the listening position, adding to the cumulative total. Some of the reflected energy gets absorbed before it makes it back to the listening position, so in my case at least, the direct sound fall-off rate is overwhelming the later reflections addition to the cumulative sound level at the listening position. The early reflections are largely redirected away from the listening position by the corner placement of the speakers.
In summary, I've found a way to return my speakers to their original design configuration and have them sound very pleasing by NOT equalizing the output of the mid-horn flat at nearfield. I will continue to listen and perhaps tweak the equalization of the mid horn to result in a more gentle slope at the listening position and find out if that's better or worse. I've got a lot of questions remaining. Why doesn't it perceptually sound unnaturally dull when listening off-axis, or in another room? The upper frequencies being so narrow should be adding less total energy to the room. Somehow I'm not perceiving it as at all unnaturally “dark” sounding anywhere in the room or in adjacent rooms. It's really nice and easy on the ears. It also can get a lot louder because the EQ requirements are greatly minimized, and the tweeter isn't being asked to go so low.
Update December 22, 2025
I tried to adjust the room slope at the listening position to 0.5 dB/octave. This sounds too bright. I've also tried to model the fall-off rates for narrow vs wide dispersion drivers and the results are not showing a consistent brightening. This may be a specific problem with my room and the specific characteristics of the speaker I'm using. Fall-off rate is not intrinsically about directivity. It's more about the curvature of the wave front. It's possible to have a narrower pattern that is still made by a curved wave front, which means it will have the same fall off rate as a wider dispersion pattern that has the same wave front curvature at any given distance. As we move away from a small plane source we see that it's wave front curvature becomes very close to the same as the curvature of a point source at the same location. It's only up very close that this fall off rate differs a lot. This may be why 2 meters is considered sufficient for on-axis measurements. The anechoic on-axis response shouldn't change much with distance after that for most any “normal” speaker. My simulations did suggest that room acoustics can have some surprising effects, with diffusion being typically more effective on high frequencies than lower, this could mean diffusive elements can beam more high frequency energy back to the listener. If the walls are bass leaky, this could result in a brighter than expected response at distance, although the same would hold true with a speaker that was not beamy up top. I'm starting to suspect that the reason my speakers need a fairly steep room curve is due to the fact that for the anechoic response to be flat on axis at the listening position, the off axis energy in the lower frequencies is actually quite a bit higher. My problem may be that I don't have any way to get the anechoic on-axis response at 2 meters because of room reflections, so I'm really shooting in the dark and getting confused by the reflections creeping in. I had better luck by starting with an no-EQ result and then working backwards from there. I suspect now that if I could figure out how to get the on-axis anechoic response for the speaker what I like the sound of would actually look pretty flat.
Update January 2, 2026
It turns out that I can't seem to get a sound I really like that makes much sense in terms of ancehoic on axis response being prefectly flat for each driver. I'm not sure why, but it always seems best if I don't EQ the speaker drivers, but only set the crossovers and adjust their levels by ear. The end result in this room is a fairly steep room curve as noted earlier, but with some specific frequency cut backs as a room treatment.
Update January 20, 2026
Somehow I now am EQing the drivers flat fairly nearfield and I am liking the sound. I should never be a speaker reviewer. I don't know if it's my perception that changed or what but now I'm liking the sound with the coaxial tweeter EQ'd flat on-axis measured at 2 meters and crossed over all the way down at 600 Hz to the mid-horn, which is also EQ'd flat from it's 600 Hz crossover down to about 130 Hz. I don't understand what changed, but now at least this makes some sense. It actually sounded prettty good with a large overlap between the mid and tweeter, covering from about 1000 Hz to 5000 Hz working together. It was a little crunchy sounding at times but very present and lively and not strident or overly sibilent. Not perfectly neutral but very easy on the ears and worked well for me with everything I listened to. With the 600 Hz crossover I avoid dips in the off-axis response. Instead the off-axis starts sloping down at about 700 Hz or so in a straight line. This also sounds good. Not quite as punchy and intense but seemingly smoother, less colored.
Use of Absorptive Materials in Loudspeakers
1. The Case of Loading a Dome Midrange (and maybe some woofers too) into a Large Waveguide.
I have been recently thinking about and experimenting a little with this project. One problem I immediately encountered is that the acoustic center of a dome is up toward the center of the dome. If the dome is placed in a waveguide so that the top of the dome is protruding inside the waveguide to some degree, as is typically seen in many studio monitors with waveguides, the sound radiating from the center of the dome will cause reflections off the sides of the waveguide at higher frequencies, which destroys the desired constant directivity that the waveguide is meant to provide. This means that the dome must be moved back so that the sound waves that enter the throat are all perpendicular to the sides of the waveguide. But how do studio monitors get away with it and still show smooth dispersion? The answer is somewhat complex, but generally they use very shallow waveguides, and sometimes some complex shaped items in front of the tweeter to better facilitate spherical wave entry into the waveguide. They may also use other tricks with the precise shape of the dome.
The reason the dome acts more as a point source at it's center is that the dome moves like a piston in the air, and the surface of the dome at it's center is most perpendicular to the piston motion, so the center produces most of the sound. If we move this point source back behind the throat opening of the waveguide it can produce smooth dispersion from the waveguide without creating internal reflections. However, sound energy not directed into the throat will escape to the sides. If the sides are enclosed to create a back chamber, the higher frequencies will bounce around the chamber and enter the waveguide late, causing poor frequency response and resonances. To solve this, absorptive material can be added to the chamber. The compromise that's been made is efficiency loss due to absorption in exchange for a more constant directivity output.
Porous or fibrous absorptive material will tend to become less effective at lower frequencies for any given thickness. So high frequencies are more readily absorbed. This can be a problem sometimes, helpful in other situations. In this situation the high frequency effectiveness works as an advantage. As the frequency goes down the waves will get longer and eventually not be able to bounce around in the chamber, but will instead emerge mostly in-phase with the direct sound from the speaker dome, complementing it and increasing output from the waveguide. This is helpful because EQ can be used to flatten the response, and the driver will now be more efficient and produce reduced distortion, and possibly allow it to be crossed over at lower frequencies.
Another advantage to having a chamber with absorptive material behind the waveguide throat is that it provides an opportunity to load more low frequency drivers into the chamber from the sides. This is a version of Danley's Multi Entry Horn concept. It's potentially better in that it doesn't require holes to be made in the sides of the waveguide to accommodate the low frequency drivers, which can be a source of diffraction of higher frequencies inside the waveguide. Note that nobody ever puts holes along the sides of their waveguides if they don't have to.
My first experiment with this idea involved the use of a conical horn with a 60 degree flare, a 1“ throat, and about 18” at the mouth. The result of loading the driver into this horn with an absorptive back chamber was a reduction in on axis treble energy at 13 k Hz, but no reduction in off axis energy at that frequency. This meant that 30 degrees off axis now had the same energy as on axis. At 400 Hz there was a 6 dB gain in out put both on and off axis. This meant that the on axis and 30 degree off axis frequency response were now the same, so constant directivity was successfully achieved albeit with some loss of total energy output in the higher frequencies, and a gain in output in the lower frequencies. There was some roughness in on-axis response in the middle frequencies due to mouth reflections from the conical horn, which does not have an exit flare. It may be possible to reduce these by using absorption around the mouth of the horn, which could be smaller and easier to construct than an adequately sized exit flare.
Web Based MATT Tool
The Musical Articulation Test Tones (MATT) is a modulated sweep signal developed by Art Noxon at Acoustic Sciences Corporation to test room acoustic performance.
The signal can be played through a high fidelity sound system to reveal issues with a room's acoustics. This signal can be listened to from the listening position to reveal audible issues, and it can also be recorded and then analyzed with special software to find out what frequency zones are most problematic and get an overall assessment of a room's acoustical performance in the bass and lower midrange frequencies. To get an analysis has so far required the user to send a recording of the MATT playing in their room back to Acoustic Sciences Corporation, where a specialized script is then run to analyze the signal and generate a report.
A long term goal has been to make this test easier for end users to execute and analyze. One difficulty has been making it easy to know what frequency a listener is hearing when they hear problems. Another has been to allow an end user to get an automated computer analysis of the results that is graphically explicit and clear, and allows them to get more insight into what the results mean. Ideally an end user could look at a graph of the results that shows score and graphed response at various frequencies, and then allow them to play it back again in its entirety, or select problematic areas and play just those, and allow a quick comparison of what the problem area sounds like compared to the reference signal. This listening would be done using headphones because they don't suffer from room acoustic issues, so the differences can be clearly heard.
I've been working on a project to develop a web browser based MATT analyzer tool that will allow end users to load a recording they've made and have it automatically graphed and evaluated. The initial learning curve has been around getting access to the data stream using web audio APIs available in Javascript. One issue was getting RMS averaging levels that were over short enough duration to show the detail that the MATT analysis requires. This has been solved adequately now. To make this analyzer more useful will require user interface improvements for playback, along with options to record directly from a device such as a phone or tablet, or PC with a microphone attached. This will allow an end user to use their phone, tablet, or PC to send the test signal to their audio system and then record the results and get the analysis all in one step from a single device.
To do list:
- Allow recording the MATT directly from a phone or other web device.
- Enable touch screen support and auto detect various devices screen size and orientation to automatically optimize layout appropriately.
- Allow saving recordings as web browser cookies
- Allow deleting recordings
- Allow re-naming recordings and adding notes
- Improve playback and selection controls
- Change layout so that there's only one graph showing at a time, with an option to show overlays of multiple measurements on the same graph
- create an auto loop playback function that will play a short selection repeatedly, alternating between the recorded result and the MATT reference file itself
- auto detect home computer or cell phone and adjust interfaces to allow playing and recording the MATT simultaneously on a laptop or desktop computer, but hide that option on mobile devices that don't support simultaneous playback and recording.
update January 30, 2026
The MATT web app is now able to record the MATT and analyze the result using an attached microphone on a computer or mobile device. Unfortunately it looks like Android and iOS do not support sending sound out over Bluetooth while recording at the same time. This is disappointing because it prevents a very convenient option for playing and recording and analyzing the MATT all at once with just a cell phone. On the bright side, it's possible to use a laptop or desktop computer to do a simultaneous play and record of the MATT.