Directivity Experiment: Wall Splasher 2-way

[BenB]

I've been curious about the impact of lateral (side-wall) reflections from speakers. Some people treat them, others don't. I was pretty sure I preferred them. I had a hunch that the ipsilateral (same side) reflections were beneficial for adding breadth to the soundstage. After seeing the measurements of the C-note on this site, I developed a plan to use the same waveguide tweeter in a speaker that would have variable directivity.

In order to accomplish this, I would put a tweeter on the front, but have a second tweeter on the side facing away from the listener. The side tweeter would be optional, and could be turned on, or off. The front and side would also both have mid-woofers. I have used the RS125-4 in previous designs and liked it, and had several laying around. They don't need a lot of volume, so putting two in a small bookshelf was possible. I ported it to get some extra bass out of them, but they still only reach down to 60 Hz or so.

I attempted to keep the front and side drivers as close as possible, by placing them close to the corner. This required me to flip the placement of the side drivers, in order to make room for the woofer magnets. That's why the front woofer is on the bottom, but the side woofer is on the top.

I used a bi-amp speaker terminal. The bottom terminal is wired to the front drivers and the side woofer. The side woofer has to be on all the time, as it basically fills in the baffle step. At low frequencies you're listening to both woofers, and you can't just turn one off without changing the direct sound spectrum. The top terminal is wired the the side tweeter, allowing me to turn it on or off at will. The side tweeter doesn't impact the direct sound much at all, particularly on the design axis, which is about 15 or 20 degrees inside of the line normal to the face of the speaker.

Of course, this is a flawed design that I put together just to get a hint of what the impact would be, and see if there was promise to the idea. The crossover is at about 2500 Hz. That means the side woofer is "splashing" the ipsilateral side wall with frequencies up to 2500 Hz all the time. That's most musical content. I don't find the tweeter to sound particularly high end, but I needed 4 of them for this experiment, so the price was right. The bass extension is limited.

Having said that, it was very interesting to be able to turn that side tweeter on and off at will. I found that I did prefer it on. I also found that the speaker sounded much better than it should given all the flaws and compromises. On certain songs, the effect of blasting that outside wall with energy is pretty awesome. It really fills the room, but in a way that still maintains a lot of coherence and detail if compared to something like the bose 901.

So now I've been contemplating what an idealized version of this speaker concept would look like. I'll post about that, and some measurements in a follow-up.

 

[abdo123]

Can you show us in-room measurements of 2 tweeters on, 1 (main) tweeter on, and 2 tweeters on but the side one is opposite phase?

raising the volume has always been my method of choice to 'excite the room' with controlled directivity speakers, while also maintaining the higher ratio of direct sound / reflected sound that controlled directivity designs have.

 

[BenB]

What do you mean by in-room measurement? Gated to prevent reflections, or not?

 

[BenB]

Here are some quasi-anechoic measurements on the listening axis. I should probably mention that the side tweeter plays about 3 dB lower than the main tweeter. That gives it a little less opportunity to interfere on axis. Ideally, the drivers that are aimed off axis would be amplified separately, and the user would be able to continually adjust their output.

Blue shows the response with just the front tweeter playing.
Red shows the response with the side tweeter also playing.
Green shows the response with the side tweeter wired out of phase.
Note the scale is +10/-20 dB.


If someone were listening on this axis in an anechoic chamber, I think they'd be hard pressed to know if the 2nd tweeter was on or not. The same cannot be said when room reflections are present...

 

[BenB]

I tried to do what would constitute an in-room measurement. I simply asked Holm Impulse to perform "amplitude smoothing" of the total response with 1/12 octave smoothing. It looks a bit funny to me, but this is what I got. I had the mic about 7 or 8 feet back, with the speaker about 4.5 feet from the left wall, and a little over 5 feet from the front wall.

I included the impulse response measurements, because those clearly show the difference in the sidewall reflection. The sidewall reflection comes in with a delay of just over 6 ms. The stuff around 8 ms might be related to the front wall. I'd guess the stuff just before 4 ms is the ceiling, and the stuff just after 2 ms is the floor. I haven't done the math to confirm these. I need to go eat dinner soon.

The first image shows the total response with just the front tweeter in Blue, and with both tweeters in Red. The side tweeter adds about 1.5 dB of extra output in the 3-8 kHz range. It also adds significantly to the transient nature of the sidewall reflections in the impulse response.


The second image compares both tweeters wired with normal polarity (Red) to both tweeters with negative polarity on the side tweeter (Green). The frequency response is slightly smoother with normal polarity. In the impulse response, you can see that the most transient aspects of the sidewall and front wall reflections have opposite polarity.


For reference, these measurements were taken sort of in the corner of a fairly large basement (with low ceilings). The response in the 40 Hz range is real. I was using sweeps and was quite surprised to hear so much output, thanks to the room reinforcement. Too bad about the dip at 60 Hz.

 

[Duke]

I had a hunch that the ipsilateral (same side) reflections were beneficial for adding breadth to the soundstage.

... it was very interesting to be able to turn that side tweeter on and off at will. I found that I did prefer it on. I also found that the speaker sounded much better than it should given all the flaws and compromises...

So now I've been contemplating what an idealized version of this speaker concept would look like.

Welcome to the weird and wonderful world of polydirectional loudspeakers! Credit to the late great Richard Shahinian for coining the term.

Polydirectionals have not been studied and modelled and measured and refined nearly to the extent that front-firing loudspeakers have, so there is much more great undiscovered country out there waiting for you. With polydirectionals you can manipulate the reflections independent of the direct sound, mitigate the baffle step, and even reduce the floor-bounce dip if you so choose. The tradeoff are greater complexity with less available guidance, spreading your driver budget more thinly, and perhaps some speaker placement constraints.

I hope you are aware that at least one noted researcher chose a polydirectional design after years of loudspeaker evaluations at the National Research Council facility in Canada. You may have heard of him - his name is Floyd Toole, and his speakers were the bipolar Mirage M1's. And while Floyd no longer uses polydirectional speakers, he still augments the in-room reflections of his conventional speakers, now by means of upmixing to dedicated surround-channel speakers. If you have the Third Edition of his book, I highly recommend section 7.4.6, "Observations of an Audio Enthusiast - Toole (2016)".

Toole recently posted that good multi-directional speakers are a "viable alternative to forward-firing designs."

You have discovered a new world which, imo, those who cannot see beyond measurements and into the realm of applied psychoacoustics are not aware of. You will be forced to trust your ears because conventional measurements are unlikely to tell enough of the story. For instance the difference in spatial and timbral quality you perceive is not adequately conveyed by comparing the Side Tweeter On versus Side Tweeter Off curves, is it? I encourage you to not be dissuaded by armchair skeptics, should they weigh in. See your ideas through to completion; give them the best chance of success that you reasonably can.

I wish you the very best as you move towards the idealized version of your speaker concept.

 

[BenB]

Thanks Duke. I read the section in Toole's book that you referenced. Interestingly, I see some similarity between my in-room measurement of the wallsplasher, and Toole's in-room measurement of the M1 (Figure 7.20, labeled "steady-state room curve at prime location. No Spatial averaging."). Of course the M1 had a lot better bass performance.

I see Toole uses the term multidirectional, while you've chosen polydirectional. Any reason?

I do think it's advantageous to preferentially reinforce the ipsilateral sidewall reflections, as opposed to making something truly omni. The front-wall reflections may add a (false) sense of depth, but they don't decrease inter-aural cross-correlation, which has been linked with our sense of ASW and envelopment. On the other hand, you don't want people listening from a spot where they can see (and thus hear) both sets of drivers (because they will interfere with each other poorly if there are different path lengths), so from that perspective it makes sense to make them diametrically opposed.

I did an experiment where I swapped my wall-splashers (left to right, right to left) so that they preferentially reinforced the contralateral wall rather than the ipsilateral wall. I perceived this as a detriment rather than a benefit. Again, it's nice to have this kind of unusual speaker to enable such experiments.

I believe in order to do this right, the speaker should probably be at least a 3-way design, and the mid and tweeter should probably have constant directivity. The mid should probably reach a bit below the baffle step. Since the speaker will be omni below the baffle step, the idealized version probably has a sizeable baffle, or else uses some audio trickery to maintain directivity low in frequency, like the D&D 8C. I have no real way of knowing how low is low enough.

It's easy to see a path to maintain well controlled nearly constant directivity down to 1000 Hz (big improvement over 2500 Hz) if I were to utilize something like a KEF Uni-Q coaxial on each side. I haven't seen anything like them available for DIY, however. The DIY coaxials don't measure anything like the Uni-Q, showing interference problems and inconsistent dispersion. It seems to me that the Uni-Q actually manages to maintain constant directivity to a lower frequency than a waveguide of the same size (compare the diameter of their mid to the mouth of a waveguide with constant directivity to 1000 Hz).

It may be possible to rig-up a system that's closer to ideal than what I have using a couple of KEF atmos modules, crossed to a nice woofer on a small baffle at around 1000 Hz.

https://us.kef.com/home-theater/dolby-atmos-home-theater/q50a-dolby-atmos-speaker.html

In fact, if someone has something like an R3 with the atmos add-on, they could probably point the add-on at the sidewall, feed it with a version of the stereo signal high-pass filtered at 1000 Hz, and see how they like it. It's a cheap, easy test for someone who has a speaker system like that, but it would be a large investment for those of us who do not.

If I worked at KEF, I'd probably grab a bunch of Uni-Q drivers and some nice woofers and build something DIY just to try this out.

But using parts available to the DIY market, I can't say I see the best way forward. I'm typically more of an array guy than a waveguide proponent, so perhaps someone can educate me on achieving constant directivity, with a 6 dB beamwidth somewhere between 90 and 120 degrees (I've certainly considered building triangular speakers with drivers on 2 out of 3 sides), and reaching as low as possible.

 

[Duke]

I see Toole uses the term multidirectional, while you've chosen polydirectional. Any reason?

Habit. I encountered Richard Shahinian's writings long before I encountered Floyd Toole's (rather sparse) comments on the subject, so the original term in my head was "polydirectional". I think it and "multidirectional" are equivalent.

... you don't want people listening from a spot where they can see (and thus hear) both sets of drivers (because they will interfere with each other poorly if there are different path lengths), so from that perspective it makes sense to make them diametrically opposed.

Agreed. I think you made good choices in this regard... in my experience, having only one array (per channel) contributing to the direct sound results in better clarity.

It's easy to see a path to maintain well controlled nearly constant directivity down to 1000 Hz (big improvement over 2500 Hz) if I were to utilize something like a KEF Uni-Q coaxial on each side.

If you can get good pattern control down to 1 kHz ballpark, imo that would be excellent. Your idea of adapting the KEF Atmos module seems promising to me, assuming its behavior is good of course.

But using parts available to the DIY market, I can't say I see the best way forward. I'm typically more of an array guy than a waveguide proponent, so perhaps someone can educate me on achieving constant directivity, with a 6 dB beamwidth somewhere between 90 and 120 degrees (I've certainly considered building triangular speakers with drivers on 2 out of 3 sides), and reaching as low as possible.

Wish I could help with a recommended driver combination, but I'm pretty much a waveguide guy. I have never tried to meet a beamwidth target using direct radiator drivers. My guess is that you may have to eyeball the published off-axis data for a bunch of different drivers, and/or make your own educated guesses, until some combination stands out from the rest.

While you probably want to optimize the response of the front-firing array for best first-arrival sound, the spectral balance of the contribution from your side-firing array will tend to be the summed response across its effective radiation pattern angle, rather than its response along one particular axis. In other words, the frequency response "goal posts" for the two arrays may not necessarily be identical. As if what you are contemplating wasn't complicated enough already!

 

[Newman]

I see Toole uses the term multidirectional, while you've (Duke) chosen polydirectional. Any reason?

Poly means many, and is effectively the same as omnidirectional.

Multi means more than one, and is the right term for what you are doing.

Remember, your listening tests involved switching the loudness up every time you turned on the splashers. This alone will strongly bias the listener to prefer it on. The splashing effect would have to be pretty subjectively bad before it would cancel out the strong positive bias of the extra loudness.

So, try your audition again with some quick/auto way of compensating the loudness, so the total loudness is the same when splashers on as when off. Your initial findings might be reversed.

Although, having said that, DIY confirmation bias is pretty huge and you will be struggling to control it.

cheers, and good luck

 

[HammerSandwich]

Another creative build, @BenB.

My first thought is that disabling the side tweeter will emulate adding sidewall absorbers that are too thin. Rather than balanced reflections, you get less HF. Beats me if that explains your preference, but it's at least consistent with the conventional wisdom.

Since this speaker is a bit like AR's old magic speaker, I'm curious about the effect of delaying the reflections. This would require an active version, so it's probably easier to experiment with several bookshelf speakers. If I ever have 4 identical speakers sitting around, I'll have to give that a try. FWIW, I'm in a smaller room lately, and much prefer killing the near wall bounce.

 

[BenB]

Wish I could help with a recommended driver combination, but I'm pretty much a waveguide guy.

Even though I have achieved directivity targets with arrays in the past, I understand that this application will likely necessitate waveguides instead. Unfortunately, I don't know nearly as much about waveguides. So I'd certainly be very interested to hear your recommendations and learn from your insight and experience.

 

[BenB]

Another creative build, @BenB.

My first thought is that disabling the side tweeter will emulate adding sidewall absorbers that are too thin. Rather than balanced reflections, you get less HF. Beats me if that explains your preference, but it's at least consistent with the conventional wisdom.

Since this speaker is a bit like AR's old magic speaker, I'm curious about the effect of delaying the reflections. This would require an active version, so it's probably easier to experiment with several bookshelf speakers. If I ever have 4 identical speakers sitting around, I'll have to give that a try. FWIW, I'm in a smaller room lately, and much prefer killing the near wall bounce.

I've been meaning to re-run my A/B experiment with an added EQ to maintain the same listening position spectral levels when the side tweeters are on and off. Unfortunately, we replaced all the windows in my house, and the new windows require custom mounting for every blind... so all my spare time is devoted to that at the moment. I'll get to it eventually... may be after halloween, which is sort of a big deal for my family and will also take up my spare time.

 

[BenB]

I just re-did the sighted wall-splasher A/B test. Thanks to recommendations from this forum, I used equalization to achieve a more similar loudness and spectral balance at the listening position. This required me to boost the output from about 3-10 kHz by approximately 2 dB when the side tweeter was off.
First I tried A/B testing just the application of the EQ (side tweeter always off). It was easy to tell when the EQ kicked in.
Next I performed A/B testing with the following 2 conditions:
1) Side tweeter off, equalization on
2) Side tweeter on, equalization off

This made it much more difficult to discern the differences, or identify a preference. With certain material I could tell that the front part of my room was a little more engaged when the side tweeter was on, but it was subtle. This only happened during complex symphonic music, with a wide soundstage, and only during certain parts of the piece. I had my daughter apply her young ears to the task, and using a studio recording for source material, she could barely identify any difference.

The difference was so subtle that if both speakers were offered commercially, I would simply buy the one that was less expensive (which would be the simpler one, without wall-splashing).

This doesn't mean that a more idealized version of the speaker, with adjustable directivity below 2.5 kHz would lead to similarly subtle differences. Also, I'll probably want to follow-up with a mono test, since this was done in stereo.

I do think this result implies that dispersion impacts preferred voicing. Without the side-tweeter, I prefer boosting the on-axis high frequency energy a bit. With the side tweeter active, I don't need that boost. Of course, this is one man's sighted test experience (and his disinterested daughter), so take it for what it's worth.

 

[Shefffield]

It's easy to see a path to maintain well controlled nearly constant directivity down to 1000 Hz (big improvement over 2500 Hz) if I were to utilize something like a KEF Uni-Q coaxial on each side. I haven't seen anything like them available for DIY, however. The DIY coaxials don't measure anything like the Uni-Q, showing interference problems and inconsistent dispersion.

You might want to take a look at the Omnes Audio CX 3.1 coaxial driver. It's tuned to be a good high-mid driver down to approximately 300..400 Hz. Also inexpensive enough to order a bunch for testing. Despite its low price, all the reviews I read so far praise it.

Loudspeaker shop | Omnes Audio CX 3.1 | DIY

Omnes Audio CX 3.1 - Test conclusion Hobby HiFi Heft 1/2017: Mid-high coax with aluminum and titanium membrane. With the CX 3.1, Omnes Audio now delivers an even better midrange coax at a still amazin

www.oaudio.de

I plan to purchase a pair some time next year..

 

[Duke]

Poly means many, and is effectively the same as omnidirectional.

Multi means more than one, and is the right term for what you are doing.

I think you are right, and have switched to using the term "multidirectional", though I may occasionally lapse.

Remember, your listening tests involved switching the loudness up every time you turned on the splashers. This alone will strongly bias the listener to prefer it on. The splashing effect would have to be pretty subjectively bad before it would cancel out the strong positive bias of the extra loudness.

The calculated SPL difference the splashers make is less than two-tenths of a decibel at the listening position. With high resolution measurement equipment that could probably be adjusted for, but not fast enough without something specialized which I don't have and see no reason to purchase.

The audible difference with and without is not so much in the realm of "sound quality"; it is much more in the realm of "spatial quality", and I don't think 2/10ths of a dB of SPL change is going to make a difference in perceived spatial quality, at least not a large one.

That being said, I acknowledge that the slightly higher SPL with the splasher engaged results in a not-quite-level playing field.

Although, having said that, DIY confirmation bias is pretty huge and you will be struggling to control it.

All testing is blind, which is easily done by driving the splashers with separate amplifier channels which can be switched on and off by the person administering the test.

And if the splasher is turned up TOO high, sound quality (specifically clarity) is subtly but audibly degraded, so it's not like the slightly higher SPL is universally seductive.