Sonos Move on and off-axis measurements (an interesting case study)

[napilopez]

Edit: Proper spinorama in this post.

So I got around to measuring the Sonos Move. I've previously argued that audiophiles don't give Sonos enough credit, considering the very few measurements I've seen have been solid, they have some legit design chops behind their speakers, and their target audience generally isn't too picky.

The Sonos Move is an interesting case in that it's the company's first portable, Bluetooth speaker. To improve performance for its specific use case, it features an unconventional design. Specifically, the tweeter fires downward, sandwiched into an odd waveguide that pushes sound out through a small horizontal slit that wraps around most of the speaker's circumference. It looks like this:

(Credit: IGN)

And opened up (credit, me):


This has two theoretical benefits. One, it significantly limits potential damage to the tweeter, as its not directly exposed to the elements. The speaker is IP56 rated and is built to withstand all sort of mayhem. Even water gets in through the side, the waveguide directs it downward. Two, it leads to wide dispersion out to 90 degrees and beyond. So how does it actually perform?

(N.B: All measurements were performed in Bluetooth mode, as the Move does not have a line-in/aux jack and I don't have a Sonos connect to route wired sources to the speaker. Though as far as I can tell these measurements should be fairly representative of performance over Wi-Fi, I can't confirm that.

It's also worth noting Sonos has a room EQ feature called TruePlay which adjusts the speaker's response based on its positioning. Previously, this could only be done by waving your phone around for calibration, but now the speaker can do without user input. But TruePlay does not work in Bluetooth mode, so I wasn't able to measure it. I also disabled the speaker's 'Loudness' feature, which is basically Fletcher-Munson compensation.)

Well, on-axis, things aren't off to a great start. This is a quasi-anechoic measurement - you can ignore the dip below 200 Hz, as that's an artifact of the method.

Eesh. Nicely linear midrange, but that is some wonky treble. Though the dips are narrow, their prominence would make it seem like the treble is overall balanced too low. This, I assume, is caused by the odd waveguide and subsequent reflections caused by pushing the tweeters sound through a narrow slit.

Maybe things would be better within the listening window? This is an average of the sound out to 30 degrees horizontally and 10 degrees vertically. I also spliced the nearfield bass response.

Not much better, and this didn't seem to represent what I was hearing.

Luckily, we know that you need comprehensive anechoic data to really assess a speaker. I don't have an anechoic chamber, but doing what I can within my apartment and my limited knowledge, we get a better picture of what Sonos was going for. Here's a full horizontal set of measurements out to 90 degrees, the listening window response, as well as an early-reflections curve that approximates the angles used in a proper spinorama..

It all starts to make some sense once we have enough data for an early-reflections curve, which probably best approximates what you'd actually hear in a room for a speaker like this. Though still messy, you can see how the messy treble balances out significantly when multiple angles are taken into consideration, and how the upper midrange and treble are basically at the same level no matter how far off-axis you are. It's very well possible this graph would smooth out further if I measured in 10-degree intervals a la Harman rather than at 15 degrees because I'm lazy.

In practice, this makes the speaker quasi-omnidirectional. With most speakers, the treble dulls as you move off-axis. But the Move maintains energy in the top active all the way into the top active. Of course, the direct sound is still important, especially if you're listening from up close. But considering these speakers don't have a line-in and stereo doesn't even work in Bluetooth mode (you need to use them in Wi-Fi mode for that), I doubt anyone's planning on using these on their desktop. There's a broad midrange dip far off-axis, but I wouldn't consider it a major problem as it does not show up on the early reflections curve.

Also note the bass extension. Measured at around 80dB from 1m, they have impressively linear bass down to 40-50 Hz, and drop off precipitously after that. So not much in the way of sub-bass, but it should be nicely balanced for most anything acoustic.

Of course, this is an active dsp speaker, so how much bass you get will vary depending on your listening levels. Unfortunately, I have not been able to push the speakers to their limit for fear of getting stabbed by neighbors in the middle of the night after driving them insane with one too many REW sweeps. Anecdotally, the speaker seems to get quite loud before bass cuts out. But I'll get to testing sooner or later.

Listening notes: Graphs aside, they are probably the best portable, battery-powered bluetooth speaker I've heard, which one would hope considering they're also the most expensive battery-powered bluetooth speaker I've heard, at $400. It's not a high bar to cross, but still, it sounds good. I can't notice any of that treble wonkiness unless I'm listening from up close, paying close attention, and am playing the mellifluous sound of sine tones.

The Move's most notable aspect is how good it sounds no matter where you're standing. Not quite as good as the truly omnidirectional Homepod in this respect, but better than most speakers. In a stereo pair, its soundstage is noticeably wide and enveloping.

The bass is pretty remarkable. It's not a Devialet Phantom Reactor, but it also doesn't cost $1,200 per speaker. You notice the lack of sub-bass if your music has a lot of it, but you're unlikely to notice in much other program material since the bass is so linear until the drop.

Better yet, you can take that sound to the beach, the park, a party, the rainforest, or wherever you see fit. They're not the pinnacle of hi-fi, but it's nice to see Sonos seems to have taken care in balancing sound quality with durability.

Summary: If you don't use the Move as studio monitors or expect the pinnacle of hi-fi, they should sound pretty great for portable speakers, with good timbral balance. It's a very nice change of pace from Bluetooth speakers that either sound tinny for lack of bass, or bloated from trying too hard.

 

[JJB70]

Excellent review!! A lot of people think I am just being wilfully contrarian when I say that there has been and continues to be some serious R&D in the wireless speaker and soundbar segments and that company's like Sonos are making some impressive gear. I honestly think this segment is a better pointer to the future of audio than the audiophile niche. At one time most speakers in this segment were one trick ponies in providing a lot of noise from a small box if you weren't too bothered what it sounded like but now there are some genuinely very good speakers from companies like Sonos, Devialet, Apple and even Bose.

 

[Thomas_A]

Would be interesting to compare that to the IKEA Eneby 20 & 30. According to measurements by the Swedish Audio-Technical Society the Eneby 30 is impressive. Others have measured the portable Eneby 20:

file.php

 

[FrantzM]

Great review and one that furthers the need for speakers testing here, at ASR.

We are at a point where the better electronics are transparent , no adverbs. No "essentially", no "virtually". a $9 DAC surpass our hearing abilities. ASR perhaps alone in the audiophile ecosystem is showing the way. Transducers are the next step toward the evolution in High Fidelity reproduction.

While many may find this to blasphemy, this SONOS and other inexpensive speakers such as the JBL LSR 305 and 308 sound better than many audiophiles darlings costing 10 times as much. Not only do the SONOS sound better than some of those expensive things, they sound good in a variety of rooms.
We need to test transducers. ASR-style.

 

[kaka89]

Could you explain how to interpret these graphs?

Although I have basic understanding of what these metrics is referring to, I often found different speakers have quite similar frequency response. A cheap LS50's graph looks somewhat as flat as S400 to me.

 

[Darwin]

Interesting although I don't fully understand the measurements either.

 

[napilopez]

Could you explain how to interpret these graphs?

Although I have basic understanding of what these metrics is referring to, I often found different speakers have quite similar frequency response. A cheap LS50's graph looks somewhat as flat as S400 to me.

So interpreting speaker measurements is a bit of an art - and it's certainly easier when you are actually performing the measurements yourself and listening to the speakers. Audioholics has a great write up that's probably better than anything I can come up with.

But here's a very lengthy 'summary' of how I go about making and interpreting measurements:

• First off, to best assess a speaker's sound, you need to capture its sound without the influence of room reflections. I don't have an anechoic chamber, so I use a gated measurement method that gets pretty close. Software like REW makes it easy to see when the first reflections 'hit' a microphone - so I simply cut off the data right before that happens. The speaker is mounted on a stand placed on a lazy susan turntable as far from every wall as possible. I measure at 15-degree intervals, then place the speaker on its side to repeat the measurements for vertical data. It's all a makeshift solution but one that's proven surprisingly reliable when compared to measurements made by others with more resources. I'm still learning, but I'm lucky that my job lets me try enough speakers to learn by trial and error in addition to books and online literature.
• The consequence to gated measurements is the data becomes less accurate at lower frequencies, depending on how soon the first reflection hits. In my case, the data becomes unreliable below around 200hz, so I take a nearfield bass measurement instead. Nearfield measurements create an artificial bump in the bass, but this easily rectifiable with software. I then splice the nearfield measurement to the gated farfield measurement where the curves line up.
• In general, you want a speaker's response to be flat on-axis and change smoothly to its sides. We hear a combination of the direct sound reaching our ears first and reflected sound caused by a speaker radiating in all directions. Both affect our perception of timbre and tonality. In general, the closer you are to a speaker and the bigger your room, the more direct sound you hear relative to reflected sound. The further you listen and smaller your room, the more you hear the reflected sound.
• In practice, especially for living room listening, the on-axis sound may not represent the direct sound. People usually don't sit perfectly still and centered. Some designers even make their speakers measure flatter slightly off-axis, since most people don't toe in their speakers all the way. Hence the "listening window" measurement, which averages measurements within a 30-degree horizontal and 10-degree vertical window. It's a better indicator of the direct sound. Nearfield monitors may be an exception, as they're usually designed with stationary on axis listening in mind.
• Another reason the listening window is important is that sometimes the on-axis curve shows dips or peaks that aren't actually very audible (common with non-KEF coaxial speakers, for instance); the listening window measurement also helps smooth out these insignificant discontinuities.
• Flat listening window and smooth off-axis is good, but how manufacturers go about "smooth off-axis" varies dramatically. It's truly where there's a lot of room for preference and different use cases. It's where modern designers probably spend most of their efforts.
• In particular, there's some debate between narrow and wide directivity/dispersion designs. Narrow directivity speakers seek to minimize the intensity of early reflections and tend to have a more focused, precise soundstage. Wide directivity speakers tend to increase lateral reflections and create a more expansive but diffuse soundstage with a wider sweetspot.
• I once read someone describe narrow directivity speakers as transporting you to the recording venue, while wide directivity speakers invite the musicians to perform in your own home. I agree with that sentiment, and it makes some logical sense. The wider the directivity, the more reflected sound you hear, and thus the more of your room's influence you hear.
• There is some research that suggests people like more lateral reflections for recreational listening, and hence tend to prefer wider directivity speakers. I'm in this camp.
• On the other hand, other research suggests lateral reflections are only beneficial if the room is big enough and therefore if the reflections are delayed enough. Otherwise it can potentially muddy up the sound.
• Along the same veins, there is research to suggest audio professionals prefer fewer reflections when mixing, and more when mastering or listening recreationally. Again, whatever may be "ideal," it seems there is room for preference here. Sometimes it's simply a matter of how the track was mixed.
• As for how this looks in graphs, the Neumann KH80 is perhaps the pinnacle of speakers I've measured and is on the narrow directivity side of the spectrum. Super flat on-axis and listening window, and the sound decreases in level as you go further off axis and higher in frequency.
  
• Many speakers with large, deep, well-implemented waveguides show very similar radiation patterns to the neumann. You see very similar off axis response trends on the Buchardt S400, KEF R3, and KEF LS50W, which also use large, deep waveguides (in the case of the Kefs, the midrange cone acts as a waveguide for the tweeter).
• Wide directivity designs, by contrast, tend to try to keep more energy higher into the treble. These speakers typically have shallower waveguides or none at all. See Devialet Phantom Reactor and Focal Chora.
• Then you have wacky responses tuned 'by ear' or with little regard to off axis radiation. These speakers tend to have an unstable soundstage with characteristics that change as soon as you move a bit off-axis, and they are more finicky about positioning. I haven't measured many of those luckily, but the B&W Formation Duo is in this camp. Shame - it's a good speaker otherwise.
• The horizontal radiation is most important, especially for spatial cues, but the vertical response can affect timbre too. An advantage of coaxial speakers like KEFs is that they largely maintain this smooth radiation pattern in the vertical plane too.
• The Early Reflections curve averages several horizontal and vertical angles and gives a good estimate of the combined sound we hear in a typical home listening situation. The Sonos Move is the first time I'm including an early reflections curve in my measurements, but I'll do so more in the future.
• Sound Power, which I do not measure right now because it requires some extra math I haven't quite figured out how to do efficiently yet, depicts how much energy the speaker radiates in all directions.
• Wide directivity speakers show an early reflections curves that is not much lower in level than the listening window curve.
• Most speakers have some degree of a dip around the crossover frequency in the early reflections and sound power curves here because of dips in the vertical response - unavoidable with vertically aligned drivers.
• If a bump is present in both on axis and off-axis curves, it's a resonance and can be easily EQ'd out.
• If a bump or dip is only present at a few angles, it's a directivity issue and cannot be fixed with EQ. If there's a dip in the on-axis that doesn't show up off axis, for instance, fixing the dip on axis will introduce a bump off axis.
• In-room, non-anechoic measurements are not worth very much unless averaged from several positions. When looking at an in-room measurement, the frequency response should tilt down by roughly 10 dB from 20hz to 20Khz. Similarly, you'll see a downward tilt in the early reflections curve and sound power curve. if these lines were perfectly flat, the speaker would sound too bright.
• For a typical room, you can predict an averaged in-room measurement with a weighted average of 12 percent listening window, 44 percent early reflections, and 44 percent sound power.
• Lastly, don't forget about bass. It makes up about 30 percent of preference ratings in blind tests. The only thing to keep in mind with active speakers is that their bass output is usually tied to different volume levels. While a passive speaker will almost always produce the same amount of bass at different levels, active speakers will often cut down on the bass amount as you play music more loudly.

For a speaker like the Sonos Move, which won't often be listened to in the nearfield and which will rarely be heard directly on-axis, it makes sense to optimize the early reflections curve and sound power (which I didn't measure but usually looks like the early reflections curve but with a steeper slope). Note how though the treble is wonky on-axis, it has a similar overall energy level in the treble all the way out to 90 degrees, meaning you should get a similar timbral balance from most listening positions.

As for your question about the LS50 and Buchardt S400; you're right in that they measure somewhat similarly, but the devil is in the details. The LS50 has a dip in the midrange on axis, while the S400 is flatter. Both have similar radiation patterns, but the LS50 has smoother vertical radiation by virtue of its coaxial design. They each have their strengths and flaws, but a big part of the Buchardt's advantage is in the bass, which extends far lower than the LS50.

Hopefully that gives you something to work with. As always, I recommend Dr Floyd Toole's book for basically everything you need to know about understanding speaker measurements, but you can watch this lecture of his for a great summary.

 

[kaka89]

Thank you a lot for the detailed explanation.

How is the listening window calculated? By looking at the graph, they are not a simple average of off-axis responses.

 

[napilopez]

Thank you a lot for the detailed explanation.

How is the listening window calculated? By looking at the graph, they are not a simple average of off-axis responses.

Different sources use different systems, but it's all the same basic idea.

The Spinorama/Harman/CEA-2034A Listening Window is a simple average of 9 curves: the 0/±10/±20/±30 horizontal and ±10 vertical curves. I believe they came to these values after studying the typical seating positions and angling of speakers. You can see how Harman calculates its the curves here.

The NRC/Soundstage Network uses five curves: 0/±15 horizontal and ±15 vertical. This gives more of a bias to the vertical measurement, which is why you typically see a bigger dip in the crossover region in Soundstage network's listening windows measurements.

Stereophile/JA I believe does seven angles, 0/±5/±10/±15 degrees horizontal and no vertical.

My listening window is a mishmash of all the above: Seven curves, 0/±15/±30 horizontal and ±10 vertical. This came to be out of convenience and some convoluted reasoning.

Basically, when I was first coming up with how I wanted to capture and present measurements, I decided on doing NRC-style graphs with curves for various angles because I write for a mainstream tech site and I think that's easiest to explain. I also like being able to see the specific response at different angles - it helps with positioning and toe-in, especially in the nearfield - and I thought 15-degree intervals (as opposed to the 10 or 5 degree intervals used in spinoramas) were enough to tell me about directivity without making graphs look cluttered.

It would also speed up the process a bit; audio isn't my primary coverage as a reporter, so I can't spend all my time measuring speakers.

But then I decided to add 5 and 10 degree data points for vertical measurements because I like how Stereophile gives advice on speaker/ear height, and some speakers do show major changes with small vertical changes. So I took the curves I had and made the closest approximation to the Spinorama listening window I could. Because I have two fewer horizontal data points, my listening window curve does give vertical measurements slightly more weight than in a real spinorama would, but it's all really 'close enough.'

That said, I am considering just moving to 10 degree intervals in the future though so that I can make proper curves more compliant with the spinorama/CEA-2034A standards instead of approximations. Most of the time spent measuring speakers goes into the setup and presentation of the data rather than the actual measuring anyway, so a few more data points wouldn't take up that much more time. I'm more concerned with it making graphs look too cluttered. That Sonos Move off axis data is already messy enough; imagine it at 10-degree intervals!

Lastly, note that the listening window can also vary among sources if they are using a different reference axis; I've sometimes seen the NRC and stereophile measure speakers at different heights, for instance. Usually, everyone just does whatever the manufacturer specifies - usually at tweeter level or between the midwoofer and tweeter. But sometimes the manufacturer doesn't specify, so we take our best guess.

 

[sfdoddsy]

A while back I measured The Sonos Play One against the Apple Homepod, Yamaha Musiccast, JBL 104 and some lowish end KEFs for a bedroom system.

The Sonos came of f surprisingly well. Considerably better than the Homepod (massive bass boost).

I now have a bunch scattered around the house and they do their job very well.

 

[jhaider]

Most of the time spent measuring speakers goes into the setup and presentation of the data rather than the actual measuring anyway,

Ain't that the truth!

 

[napilopez]

Ain't that the truth!

Someome who understands . Yep, the measuring is basically just press a button, turn and repeat!

Balancing a multi-thousand 60 LB speaker on a stand at 7 feet high though and then turning it on its side and somehow balancing it safely for vertical measurements? That's a bit more of a pain.

 

[napilopez]

As promised, here is how Sonos treats bass at different SPL levels.

Notes: This is measuring a single speaker from the center of my living room. Distance is 130cm or about 51 inches. So don't directly compare to any similar measurements I've made where I was playing stereo speakers near the front wall and measuring from the listening position 10 feet away.

Measurements are 1/6 smoothed, but not averaged. Lowest measurement is at 50 percent volume, and then up by 10 volume point increments until you reach 100 percent at the top.

A few interesting things to note.

Bass compression kicks in very sharply at just under 84dB to protect the driver. If I'd measured this at 1m like an intelligent person, it would probably stop right at around 85 or 86dB. Compression is limited to the area under 100Hz, unlike some speakers that use a more gradual transition.

Interestingly, Sonos is applying some protection to the tweeter too, as it also cuts out at about 5-6KHz. Haven't seen that before, but given its use as a Bluetooth speaker, people probably don't care too much about treble performance when you're just pumping out the jams, although 5KHz is lower to start limiting than I'd like or have thought necessary.

On the other hand, this in-room measurement also reveals how just how well the messy anechoic treble balances IRL, and why its messyness is virtually inaudible in regular listening. It's a pretty ideal in-room performance, which seems to follow the 9-10dB tilt we want from 20hz to 20KHz within its SPL comfort zone (taking into account the nearfield measurements, and my room modes).

Good stuff. Again, it's a speaker made for a specific purpose and use case. It seems to have been acoustically designed to optimize its performance as a wide-dispersion Bluetooth speaker, rather than just ignoring sound quality. Just don't play too loud if you want great sound

 

[napilopez]

I made a spin for these based off incomplete data (15 degree intervals, only three rear curves, off axis data below 200Hz interpolated by VituixCAD). It's not exactly the easiest spin to look at it, but still gives us an idea of what Sonos was going for. The woofer's response, at least, is very smooth.

To make more sense of the curves, here are the lines with 1/6th octave smoothing. Considering on-axis listening is not a concern for this speaker and that the tweeter waveguide is almost omnidirectional, note how well the predicted in-room response and power response average out:

I get the impression that if I'd measured this on tighter intervals, say 5 or 10 degrees instead of 15, the curves would have averaged out even more. This is visible in the previously shared in-room responses at 1.3m, but also in particular with the total horizontal reflections curve, (ER curve minus vertical data). Even with just 15 degree intervals and at 1/24 smoothing, the curve balances out quite nicely.

Again, an interesting approach at creating a timbrally balanced, almost omnidirectional speaker while maintaining durability. I love seeing these sort of unique designs, and they don't always come from traditional hifi companies.

 

[blodsbror]

I would love to see how the Sonos Play 5 Gen 2 measures. Especially to see how the bass output goes when the volume rises. Have you seen any measurements around ?

 

[napilopez]

I would love to see how the Sonos Play 5 Gen 2 measures. Especially to see how the bass output goes when the volume rises. Have you seen any measurements around ?

Unfortunately not. I saw one in room measurements that seem flattish but nothing close to extensive.

 

[txbdan]

This is interesting. I have a couple Sonos Ones floating around the house and in a stereo pair I always thought they sounded pretty good. In my experience, TruePlay helps quite a bit. DSP magic is indeed magic. Active DSP-corrected speakers are probably our future.

 

[david moran]

I am coming to this a little late, but one of the main Sonos design engineers (Michael Chamness) worked through the 1980s at dbx (me too), where he ran the Soundfield Imaging loudspeaker program (after its inventor, Mark Davis, left for Dolby). Paying close attention to horizontal radiation pattern became second nature to all of us, actually first nature, come to think of it, following in the wake of prior radpat work by Allison and Villchur --- all of this at the same time as if not preceding the similar findings by the Canadian researchers. While the dbx SF work focused on a particular broadband-oval hor radpat to achieve a particular effect, it is not news to anyone who knows Chamness's demonstrated prowess and experience that Sonos designs sound as impressive and natural as they do.

(Davis's AESJ paper on his R&D work including comprehensive hor radpat measurements of the entire line, in other words polar plots of Chamness's products, is attached, just fyi.)

 

[jonfitch]

Thanks for the review. I picked up a Sonos Move and a B&W Formation Flex for a casual bedroom setup. My opinion is tonally the Sonos is extremely pleasant while the B&W has the standard zingy sound you get from B&W speakers (it has the 600 series tweeter supposedly). That said the bass region sounds extremely tubby and bloated on the Sonos. It's got a smaller woofer than the Formation Flex but feels like it's playing at least 10-15hz lower in room.

I find both of these kind of unbearable to listen to anything higher than ~70db listening level. With the Sonos, the boomy and tubby sounding bass just dominates the sound, and with the B&W I can't stand the rising treble. But if you put a gun to my head I would rather listen to the Sonos.

That said I'm surprised just how wide the Sonos dispersion is. While it's no Homepod with 8 tweeters, with it's a single downward firing tweeter but feels like its multiple tweeters in that it sounds almost omnidirectional, and the directivity graph really helps show how this is happening with the dispersion widening below 2.5K. The B&Ws are very directional, which in some ways is a good thing, because otherwise the treble response would be a real problem, but it does make it sound very small. IMO the Sonos in mono sounds bigger image-wise than a stereo pair of the Flex, which is interesting.

 

[Vict0r]

I have no experience with the Move, but I have a few of those Sonos Play speakers around the office at work, and they sound rather pleasant. The bass especially is impressive for such a small enclosure. They go much louder than you'd expect too and they fill out a room nicely. Pretty decent stuff.