Sony WH-1000XM4 Review (noise cancelling headphone)

[TSB]

When I'm outside I prefer to aware of my surroundings.

Try waiting at the Brussels flixbus station for a few hours, your preference will change

 

[m_g_s_g]

(...)the wireless Sonys and presumably every other wireless headphone on the market sounds fairly bad (...) These bluetooth products (...)

I’m not 100% sure about this Sony model (but I believe so), but my Bose QCIIs and the Plantronics Back Beat Pro 2 (that I also owned and liked and now my wife uses) are not exactly only “a Bluetooth product”. Unlike other Bluetooth headphones, they can work wired, connected through a 3.5mm Jack.

This analog sound signal is optionally digitised with an internal ADC (along with the signal from several can located mics), DSPed for noise cancellation, EQed to each manufacturer‘s house curve, converted to analog again with an internal DAC and finally amplified. When working in this mode, no Bluetooth audio profile codecs (SDC, LDAC, AAC, AptX...) are involved (although a Bluetooth connection can still be active for app control). This is IMO the best (or intended) use case for this kind of NC headphones.

Only if untethered do they work like other Bluetooth products. But then, the NC can be enabled at wish.

If Sony/Bose/Plantronics/others offered an app selectable house curve or PEQ (like Qudelix’s one), this headphones would be (even) more useful.

 

[m_g_s_g]

Also, while you don't *need* noise cancellation in a home environment... it maybe helps more than you think.

Agree! Let’s not forget the moments when the Roomba decides to enter your room , or the furnace to kick in, or the garbage collection truck runs late...

 

[Feelas]

. The onboard DSP and EQ trickery you mention is not going to close the gap with even a $75 wired headphone run off a smartphone hp jack.

Sony ANC HPs are maybe kind ot average, but any amount of DSP & EQ trickery can work wonders if done properly, there's no reason why it wouldn't if THD &FR look correct, so DSP/EQ bashing is just snake oil.

Apart from that, you'd rather NOT run any of these in passive mode; these drivers' the power ratings are low enough to fry them by accident.

Agree! Let’s not forget the moments when the Roomba decides to enter your room , or the furnace to kick in, or the garbage collection truck runs late...

ANC is completely state of the art in quality terms. Why? Because if you manage to manufacture a low-THD driver and/or get it via feedbacking, the ANC itself pushes the noise floor in bass down, thus both preventing clipping (since less noise floor means lower needed bass level to get the same result -> more amp juice for peaks) AND causing unmasking of details which are lost in the noise floor; I think both arguments are absolutely easy to spot and trivial for anyone using properly implemented ANC headphones.

Saying that, it's incredibly sad to see Sony doing such a bad way at tuning.

 

[Berwhale]

Somebody make wireless headphone that has tidal mqa built in and phone is just a remote. It’s like streamer in your head. No more this bluetooth codec mess. Amazing that 4th generation product by Sony still sucks.

Someone is having a go at it...

Streamz (streamzmedia.com)

I can't see any mention of Tidal and I'm not sure i'd want my headphones to advertise that i'm listening to Rihanna

 

[bobbooo]

Do all noise cancelling headphones work this way or do some cancel everything including non-repetitive noises?

No noise-cancelling headphones you buy work this way, including this one. Such misinformation is just based on 'lay intuition' and a lack of understanding of how ANC headphones actually work. As Oratory (an acoustic engineer) explains:

That's a common and oft-repeated belief, but not actually true.

How the noise changes over time (engine vs human voice) is irrelevant to the noise-cancelling process, it's only the frequency range that matters, since ANC is typically a minimum-phase system (meaning the ANC algorithm is able to react virtually in real-time, within less than one period (meaning "in phase") of the signal, which means it's not important how the frequency of the noise changes over time, since the ANC is able to correct for any change in real-time).

In other words:
Noise is being reduced by the same amount of Decibels, regardless of whether it's a human voice or a steady-state drone of similar frequency content.

So why can you still hear voices, while the train engine appears to be all but gone?
Because our brain is incredibly good at picking out voices.

I've touched the subject in a rather lengthy post about the limits of ANC

Relevant excerpt from that lengthy post he mentions:

Now the question: Can we reach absolute silence as far as our hearing is concerned, a sound pressure level of 0 dB?
Is this possible with ANC headphones?

No, not at the moment. And not in the near future.
And here's a couple of reasons why:

For ANC to work you need microphones recording the noise on the outside of the earcup. This recording is then phase-inverted (every - becomes a + and vice versa) and played back by the drivers of the headphone. Sound that enters the headphone (and subsequently your ear) is cancelled out because it is mixed with its own inverse: Plus and Minus equal zero. All this is done in real-time, which means that as far as the ANC is concerned, there is no difference between speech and the constant drone of an airplane engine. The "anti-sound" is not synthesized, it is simply the real-time recording with its polarity inverted with filters added.

In reality there are a few factors coming into play that make matters more difficult:

1. the noise that enters the earcup is different from the noise recorded by the microphones - because the earcup itself will filter a lot of high frequencies from the noise. This means that the output of the microphones needs to be filtered accordingly, otherwise we would experience amplification instead of cancellation - the opposite of what we want.
2. the noise entering the earcup does not hit your eardrum at the same time as it is recorded by the microphone, because sound takes a finite amount of time to travel from the outside of the earcup to your ear. ~343 meters per second is fast but not infinitely fast. This means that the waveforms of the noise and of the recording will not overlap perfectly. This is no problem at low frequencies but becomes problematic above about 1 kHz, where the two waves will lose coherency.
3. the microphones are not perfect. Since we typically use some of the smallest microphones available they inherently have a relatively high self-noise compared to high-performance microphones like you use in recording studios or measurement setups. The microphones also require signal conditioning (amplification) which involves active electronics that also have some level of self-noise due to their small size. This system-noise is of course different from the outside noise, which means that it will not be part of the cancelling and will be audible. It's possible to reduce it with good circuit design but it's especially noticeable on cheaper ANC headphones, where turning on the ANC mainly introduces hiss.
4. The drivers are not perfect. Inherent distortion (THD, IMD) still exist on drivers, and while it is much less of a problem than it was just a few decades ago, it still limits the performance of ANC since it introduces further uncorrelated noise to the system.
5. Even if we somehow managed to completely make the ear airtight and shut it off from the outside world (e.g. by pouring concrete into the ear canal) there would still be sound travelling along your bones as solid-borne noise. I've actually talked to a manufacturer of hearing protection and hearing aids about this, and they said that by filling your ears with concrete (or a similar substance of high density) you would be able to reduce outside noise by about 43 dB. Hence I assume that's about the maximum that can be done with passive isolation.

Higher-end ANC headphones use not only microphones on the outside of the earcup but also one microphone on the inside of the microphone (in a feedback circuit). The aim is to measure the noise level inside of the earcup. Quite challenging because you need to distinguish noise from the music that you want to be playing.
Plus it introduces another microphone with self-noise to the system (see 3.)

Factors 1 and 2 are why ANC headphones usually don't cancel noise above ~1kHz well and so these frequencies will often still be heard (less auditory masking by bass noise due to the ANC working better down there will also make this higher frequency noise relatively more audible). In fact, as Oratory says, these two factors sometimes do result in the ANC amplifying noise at these mid to high frequencies relative to passive mode, as is the case with these Sony XM4's between ~600 Hz and 3 kHz (Rtings' measurements):

This is relatively poor at these mid frequencies, although these issues are not uncommon among ANC headphones - if you want good noise isolation in the midrange, I'd say passive isolation from custom-molded IEMs would actually be a better choice in this regard.

The previous XM3 model actually does better in the midrange (and just a bit worse in the sub-bass):

As does the XM2 in the midrange (although significantly worse in the bass, particularly upper bass):

In terms of frequency response however, the latter seems the most balanced overall with a predicted preference rating as calculated by AutoEQ of 72/100, slope -0.53 (0 being neutral), the XM3 the worst (rating of just 33, slope -2.09), and the XM4 mediocre (rating 42, slope -1.63):

Unfortunately with these Sony models then it looks like you have to choose between the best overall ANC (XM3) or the best frequency response (XM2), or the relative jack of both trades (XM4).

So why do people continue to perpetuate the mis-truth that ANC doesn't work for non-repetetive noise then? Partly it's just blind regurgitation of oft-repeated audiophool 'wisdom' without critically thinking about the issue, as is the case with many other audio myths. And partly I suspect it's the common fallacy of conflating correlation with causation, without understanding the actual casual processes involved in the technology. Because our ears are most sensitive around the frequencies ANC stops working well, and environmental noises at these frequencies tend to be less repetitive (e.g. speech, which as Oratory said our brains are even more attentive to), compared to noise at lower frequencies (e.g. aircraft drone), some people incorrectly draw a causal link between 'non-repetetive noise' and 'ANC not working', when in fact the actual causal factor influencing the latter is 'high-frequency noise', and the co-incidence with the former is merely one of correlation. I suspect an additional factor at play here is auditory sensory gating, in which the brain has been measured as lowering its attention to repeated identical auditory signals (interestingly a function which is impaired in schizophrenic patients). Basically, we automatically filter out to a degree repetitive environmental noise, and so we are naturally relatively more attentive to non-repetetive noise in our environment, which makes perfect sense evolutionarily, but as with many of our atavistic traits, doesn't gel too well with modern, first-world problems such as wanting to listen to music through headphones in peace in a noisy coffee shop full of plates and cutlery clattering, chatter, laughter and crying babies. Thanks evolution!

 

[JohnBooty]

No noise-cancelling headphones you buy work [by canceling "repetitive" noises], including this one.

Thank heavens you posted that. I saw folks mentioning the "repetitive noises" thing a few times on this thread and it went totally against my (loose) understanding of how ANC works. I thought perhaps I'd misunderstood it all along.

 

[ExUnoPlura]

I own XM4s and just received (finally) AirPods Maxes two days ago (space gray). I've been using the XM4s with the following iOS app equalizer settings:

These settings improve sound quality but I prefer the AirPods Max when doing A/B testing with a range of sources. The subjective take-away is better balance in the Apple product. ANC is fine between either (no A/B testing yet). Weight/clamping force is OK. Max case is stupid, blah blah. Spatial audio is pretty good and, most importantly, I found that XM4 multiple Bluetooth source switching was simply unreliable. Maxes are great.

 

[Koeitje]

No noise-cancelling headphones you buy work this way, including this one. Such misinformation is just based on 'lay intuition' and a lack of understanding of how ANC headphones actually work. As Oratory (an acoustic engineer) explains:


Relevant excerpt from that lengthy post he mentions:


Factors 1 and 2 are why ANC headphones usually don't cancel noise above ~1kHz well and so these frequencies will often still be heard (less auditory masking by bass noise due to the ANC working better down there will also make this higher frequency noise relatively more audible). In fact, as Oratory says, these two factors sometimes do result in the ANC amplifying noise at these mid to high frequencies relative to passive mode, as is the case with these Sony XM4's between ~600 Hz and 3 kHz (Rtings' measurements):

View attachment 108943

This is relatively poor at these mid frequencies, although these issues are not uncommon among ANC headphones - if you want good noise isolation in the midrange, I'd say passive isolation from custom-molded IEMs would actually be a better choice in this regard.

The previous XM3 model actually does better in the midrange (and just a bit worse in the sub-bass):

View attachment 108994

As does the XM2 in the midrange (although significantly worse in the bass, particularly upper bass):

View attachment 108995

In terms of frequency response however, the latter seems the most balanced overall with a predicted preference rating as calculated by AutoEQ of 72/100, slope -0.53 (0 being neutral), the XM3 the worst (rating of just 33, slope -2.09), and the XM4 mediocre (rating 42, slope -1.63):

View attachment 108997

Unfortunately with these Sony models then it looks like you have to choose between the best overall ANC (XM3) or the best frequency response (XM2), or the relative jack of both trades (XM4).

So why do people continue to perpetuate the mis-truth that ANC doesn't work for non-repetetive noise then? Partly it's just blind regurgitation of oft-repeated audiophool 'wisdom' without critically thinking about the issue, as is the case with many other audio myths. And partly I suspect it's the common fallacy of conflating correlation with causation, without understanding the actual casual processes involved in the technology. Because our ears are most sensitive around the frequencies ANC stops working well, and environmental noises at these frequencies tend to be less repetitive (e.g. speech, which as Oratory said our brains are even more attentive to), compared to noise at lower frequencies (e.g. aircraft drone), some people incorrectly draw a causal link between 'non-repetetive noise' and 'ANC not working', when in fact the actual causal factor influencing the latter is 'high-frequency noise', and the co-incidence with the former is merely one of correlation. I suspect an additional factor at play here is auditory sensory gating, in which the brain has been measured as lowering its attention to repeated identical auditory signals (interestingly a function which is impaired in schizophrenic patients). Basically, we automatically filter out to a degree repetitive environmental noise, and so we are naturally relatively more attentive to non-repetetive noise in our environment, which makes perfect sense evolutionarily, but as with many of our atavistic traits, doesn't gel too well with modern, first-world problems such as wanting to listen to music through headphones in peace in a noisy coffee shop full of plates and cutlery clattering, chatter, laughter and crying babies. Thanks evolution!

You can equalize the curve, but you can't equalize in more ANC.

 

[starfly]

View attachment 108825

Haha, yeah well the imperial system is great for rough measurements. You can use your feet for, well, feet. You can use the first section of your index finger for inches. Great for doing quick ball-park measurements. But for precision, the metric system is just better, and also easier to work with when needing to do conversions.

And well, I've always just hated needing to do measurements where it's 3 3/16", of 5 1/4'. 180mm or 18cm or 1.8dm or 0.18m is just much easier

 

[danweast]

To sum it up the wireless Sonys and presumably every other wireless headphone on the market sounds fairly bad compared to any budget priced wired headphone. The onboard DSP and EQ trickery you mention is not going to close the gap with even a $75 wired headphone run off a smartphone hp jack. I have a pair of wireless Sonys and I had to acclimate myself to their mediocrity, distortion and unnatural sound signature in order to reap the benefits of wireless convenience. These bluetooth products are for people that want to listen to music while taking a dump or making an omelet in the kitchen.

I think this is a Sony issue, not an industry-wide issue. Sure, Bluetooth codecs are still far FAR from lossless with all the data they throw out and compress, but anecdotally I've listened to quite a few wireless headphones/TWS earbuds and each generation it's more and more impressive from the companies that clearly put in the effort.

I just got the recently released Samsung Galaxy buds pro, $200, no LDAC or aptX to be seen. To my untrained ear, they sound tonally fantastic and I have very little to complain about. Crinacle's measurements show them as a variation of the Harman curve, which of course makes sense.

Spoiler: Crinacle Galaxy Buds Pro measurement

This is of course extremely smoothed, but it shows that the DSP is at work on this product--and glad it is. The same story applies for the Airpods line of earbuds and headphones, I haven't tried the Max yet but the TWS sound very good.

So then if DSP is what allows these mainstream products to sound so good, then why is Sony's DSP-corrected headphones looking like this?

Spoiler: ASR XM4 FR Measurement

This isn't an "anecdote of being a wireless headphone" issue, this could have been corrected in the same chip that they fixed some of the egregious deficiencies with no internal amp at all. I've listened to the XM3s and agree with many of the subjective opinions--they sound bloated, boomy and they're fatiguing for me.

I'm a hard advocate for bringing headphones to tonal correctness via EQ, but these products that both include DSP correction and use it, and come out with this response, make an added EQ on top feel like putting a bandaid on a botched surgery. I now own a pair of Samsung earbuds that are out-of-box correct for cheaper, and I can't stop listening to them even with their minor flaws.

 

[wolf_walker]

With the XM3, if you start EQ'ing with the included app you drop back to a lesser bluetooth codec rather than LDAC. I use wavelet on android to avoid this, they have a pretty decent preset for the XM3, and then up the bass a hair. I enjoy em while doing stuff, wireless and all that. As delivered, even turned on, they were a boomy muffley mess, my cheap old Sennheiser 4.40 Bt were (not great but) better in most ways to my absolutely flawed ears. The preset in wavelet cleaned them up a ton though, very pleasant to listen to now. And they were $200 shipped new, any more and I'd be more annoyed.

I suspect we're going to see a lot more headphones that are slapped together to look good or fit well or be cheap or whatever and then EQ'd via DSP all to heck to make them sound worth a crap. It makes sense for mass produced consumer stuff. And in fairness it works fairly well. I would assume the likes of Sony and Apple have the budget to have or have done some extensive testing and tuning. But those guys aren't going to be aiming for what we would necessarily.

I have a Fiio bta30 fed from my D10S via toslink so I can use my XM3's with the LDAC codec and a bit better range, works well. Lag is ok for watching movies the little I tested them, too much for games though. Makes them more useful being PC-able though, and I can EQ them via peace like anything else.
And the bta30 is a slick little box, not perfect, but slick.

 

[Jimster480]

Wow, its surprising to see that Sony is selling something so deeply flawed....

 

[jmaz87]

Seeing this review is definitely bitter sweet... After being relatively happy with QC35's and QC35II's for several years I decided to get the XM4's at launch in part due to the usb-c and better battery but mostly because ANC is even better esp after a controversial firmware update before the bose 700's launched resulting in measurably worse ANC in the 35's...

They sound so different from my QC35's that it began my recent journey into HIFI so i suppose i should thank both for sounding so blah...

 

[ririt]

I own a XM3 which I found very useful when traveling and waiting for hours in various airports around the globe. Following the nice review of the XM4 from Amir, I used APO+PEACE to eq it when connected to my PC. I am really convinced by the improvement on its sound quality when EQed using the Oratory 1990 parameters. Can anyone tell me if this is an convenient IOS app to get the same kind of EQ when the XM3 is connected to my iphone?

 

[sweetchaos]

Can anyone tell me if this is an convenient IOS app to get the same kind of EQ when the XM3 is connected to my iphone?

Only 1 iPhone app supports PEQ, as far as I know....
Neutron Music Player (paid app)
https://apps.apple.com/us/app/neutron-music-player/id766858884

 

[Piotr]

I don't think anyone buying these headphones intends to use them via cable. They are Bluetooth headphones and that's how 99% of people are using them. The convenience is tremendous and worth sacrificing a bit of quality (if it's only a bit). No DACs, no headamps, no cables, just Bluetooth, and freedom.

Therefore, having a review focusing on a cable connection is pointless. This is not all about. The cordless, dac-less, headamp-less setup is the future, and that's obvious. It's not on the same level yet but it will be; it's just a matter of time.

 

[velasfloyd]

I don't think anyone buying these headphones intends to use them via cable. They are Bluetooth headphones and that's how 99% of people are using them. The convenience is tremendous and worth sacrificing a bit of quality (if it's only a bit). No DACs, no headamps, no cables, just Bluetooth, and freedom.

Therefore, having a review focusing on a cable connection is pointless. This is not all about. The cordless, dac-less, headamp-less setup is the future, and that's obvious. It's not on the same level yet but it will be; it's just a matter of time.

I use this mostly with cable, and I will not think 99% use it with cable: I have no data to think that way.

I like to use it with cable because I don't like to have a bluetooth receiver on my head. When the cable is connected, bluetooth is off and I feel on peace. I bought it because the ANC. So when I was in a flight, for example, I used the ANC and the cable. May be there is a lot of people who use it this way as well. Plus, it is a good use case as well when you want to save battery from both the headphone and the dap/phone.

 

[amirm]

If I owned this headphone for $300, I would want to use it at home wired all the time and not worry about charging it. Rechargeable batteries have X number of cycles. You don't want to use that up when you are using it at home. This is why I focused on creating a good EQ for them using wired connection.

 

[FrantzM]

I own a XM3 which I found very useful when traveling and waiting for hours in various airports around the globe. Following the nice review of the XM4 from Amir, I used APO+PEACE to eq it when connected to my PC. I am really convinced by the improvement on its sound quality when EQed using the Oratory 1990 parameters. Can anyone tell me if this is an convenient IOS app to get the same kind of EQ when the XM3 is connected to my iphone?

Me too. Please