Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Active Road Noise Cancellation

This site may earn commission on affiliate links.
With this turned on the other day I noticed a nasty hum, sounded like mic feedback and was actually making a tire noise much worse on certain roads. If I didn't know about this new feature and to try turning it off I probably would have thought there was a serious mechanical issue going on...
Me too. Here’s a video where you can hear it:

 
My car's software finally updated last night and I see I got the ANR feature....as a sound guy I had to look into this further. Took a drive and listened. It can be hard to hear if you dont know what to look for, plus it REALLY depends on the road surface. It is the low frequency tire rumble that primarily gets filtered (pretty much what would be expected).

I've read these threads and its pretty clear a lot of people don't really understand how this works....so I am going to do a quick A/B measurement to get a little data out there that can hopefully make things more clear. This is a just a rough frequency level reading, did it quick and dirty with an iphone and the RTA app instead of my laptop and calibrated mic, and there are a ton of variables, but for the purposes of understanding ANR, its good enough.

Turned off ANR. Drove on a consistently blacktop stretch of highway, turned on autopilot at a constant speed, set the RTA app to do a 3 second average (averages help to smooth out the dB readings). Held the phone mic facing up next to my ear for more than 3 seconds, recorded the result.

Turned on ANR. Repeated measurement.

I did this in a small window of time to minimize any variables from changing road surface, other cars, etc...

I then overlaid the two readings, they were close to identical EXCEPT from 70Hz to 250Hz where we see between 1dB to 4dB of reduction. With a maximum difference occurring around 125Hz (in this scenario at least). As a reference point a reduction of 3dB is equal to half as much acoustic energy.....so that's a lot of cancellation going on in that frequency range.



ANR comparison.jpg




Sound has a wave length that varies by frequency. The lower you go the longer the length of the sound wave.....the wavelength. In a closed space like the cabin of a car you eventually reach a point where the wavelength is longer than the interior space of the car. It wont fit. At this point it can no longer exist as a wave and just becomes pressure, cycling at the frequency of the sound. To a reference points....the 125Hz sound we are talking about above has a wavelength of around 9'. About the longest dimension of the Model S interior. So its definitely getting into that area where its becoming pressure more than a wave.

To cancel out that sound you play the speakers (probably just the subwoofer in this case if I had to guess) out of phase with that pressure creating destructive interference at that frequency and reducing the sound that gets to the ears of the occupants. Practically, you can only cancel a sound in a car once the frequency gets low enough that it becomes pressure, or perhaps you can stretch it to where the wavelength is long enough that the front/rear, right/left passengers are all existing within the same peaks and troughs of the wave. Trying to cancel frequencies above this point gets problematic as what might have destructive interference (cancellation) at the front seats may have constructive interference at the back seats....and you dont want that!

At frequencies above the point where the wavelengths are shorter than the size of the cabin, they can propagate around the car, reflect off hard surfaces, get absorbed by softer surfaces, and interfere with each other constructively and destructively in virtually infinite ways....basically its chaos....and it is not really practical to try to cancel out infinitely variable chaos.

If I had to guess, Tesla will probably refine this some, programming the software to more effectively target some sounds, ignoring others, sussing out mechanical noises like the HVAC or certain specific tires, also while tuning to the auditory perceptions of most people. Plus working out some bugs from what i seen in this thread... Overall I would expect the effect to continue to be subtle though, there are limits to what is possible/practical in this application.


Now I see a lot of people trying to compare this to noise cancelling headphones....you cant do that.
With a noise cancelling headphone the enclosed space and distance from the speaker to your ear is WAY smaller, maybe 1", so you CAN theoretically cancel sounds effectively up to that wave length....a 1" wavelength is equal to about 13kHz for instance....that's in the top octave of human hearing. So noise cancellation in that headphone can exist over a massively larger frequency range, plus WAY less chaos as its just the headphone, the ear, and a couple cubic inches of air space to deal with.
 
Last edited:
Thanks, great post. So if I understand you correctly, the best possible ANC for the driver might imply noise increase for passengers? Do you get any feel for whether the car does things any differently depending on whether or not there is someone in the front passenger seat? The rear seats? It sounds (no pun intended) like the best thing to do when there's a car full of people, or even just one front passenger, would be to just have the ANC stay disengaged. Or can it handle both driver and passenger simultaneously? Deep questions, these.... :)
 
Thanks, great post. So if I understand you correctly, the best possible ANC for the driver might imply noise increase for passengers? Do you get any feel for whether the car does things any differently depending on whether or not there is someone in the front passenger seat? The rear seats? It sounds (no pun intended) like the best thing to do when there's a car full of people, or even just one front passenger, would be to just have the ANC stay disengaged. Or can it handle both driver and passenger simultaneously? Deep questions, these.... :)

Ive only had ANC for one day! Haha. So no, I dont have a feel for it at all, just that toggling it on and off all day as I drove around I did notice it, but it was subtle....and I was looking for it. Thats a big reason why I measured it, need to rule out placebo when an effect is based largely on perception.

When it calibrates it says "optimized for front seat passengers" (or something like that). So its implying that it is focusing on that, but if you were to get outside the parameters I explain above, you would start to increase noise for non optimized locations in the cabin.

To answer your specific question, to optimize for one location, generally it is at the detriment to everywhere else. So it is definitely a matter of compromises. BUT below the point where sound waves become pressure in the cabin all the locations become basically equal so as the frequency goes down, the cancelling signal becomes effective in a larger and larger zone of the car, as the frequency goes up, the effective location becomes smaller, and eventually will create the opposite effect outside the cancellation area. Due to the shape of the car its fairly safe to assume the sound at front seats is roughly the same regardless of side, so the cancellation should be similar (except that the subwoofer is on one side of the car...BUT since the frequencies it plays are low....it doesn't matter because the wavelength is so long! Making any sense yet?)

So probably what is happening is that for the lower frequencies wavelengths that can still fit in the car, the waves are there but the cancelation audio is timed to do its maximum cancelation (overlaying a trough on a peak, and vice versa) at the front seats. The rear seats still get some benefit so long as the both the front and rear seats are still within the same peaks and troughs of the sound waves most of the time. When it gets to the point where the distance from the peak to the trough is the same as the distance from the front seat ears to the back seat ears, well at that point it would be quiet in the front, and louder at the rear. And thats not really what you want in any circumstance so you dont try to cancel sound at that high of frequencies.

I get what you are saying since the seats have occupancy sensors you could make it dynamic based on where the passengers are but for the minor benefit it seems like a lot of investment for not a lot of return. There is a lot of other sound deadening going on anyway (note in my readings above how steep the roll off is as the frequencies go up. Thats traditional sound deadening measures, plus cars just dont really make that much noise higher up.

So with the previous posts in this thread where it seems that ANC made things worse, my assumption would be that there was a bug, something out of phase in the processing, or the delay not correctly setting, so that the generated sound that is supposed to cancel noise does not line up with the ambient noise, so it does not cancel, and creates a racket instead.

And for any future reader sound engineers, physicists, and people who know a lot more about this than I do. Please dont bash me too hard. I know that it gets a lot more complicated than this... but Im trying to post the concepts that this is based on so that the discussion can be a little more informative going forward.
 
Last edited:
When it calibrates it says "optimized for front seat passengers" (or something like that). So its implying that it is focusing on that, but if you were to get outside the parameters I explain above, you would start to increase noise for non optimized locations in the cabin.
After calibration completed my car now says "optimized for all seats" or something to that affect.
 
My car's software finally updated last night and I see I got the ANR feature....as a sound guy I had to look into this further. Took a drive and listened. It can be hard to hear if you dont know what to look for, plus it REALLY depends on the road surface. It is the low frequency tire rumble that primarily gets filtered (pretty much what would be expected).

I've read these threads and its pretty clear a lot of people don't really understand how this works....so I am going to do a quick A/B measurement to get a little data out there that can hopefully make things more clear. This is a just a rough frequency level reading, did it quick and dirty with an iphone and the RTA app instead of my laptop and calibrated mic, and there are a ton of variables, but for the purposes of understanding ANR, its good enough.

Turned off ANR. Drove on a consistently blacktop stretch of highway, turned on autopilot at a constant speed, set the RTA app to do a 3 second average (averages help to smooth out the dB readings). Held the phone mic facing up next to my ear for more than 3 seconds, recorded the result.

Turned on ANR. Repeated measurement.

I did this in a small window of time to minimize any variables from changing road surface, other cars, etc...

I then overlaid the two readings, they were close to identical EXCEPT from 70Hz to 250Hz where we see between 1dB to 4dB of reduction. With a maximum difference occurring around 125Hz (in this scenario at least). As a reference point a reduction of 3dB is equal to half as much acoustic energy.....so that's a lot of cancellation going on in that frequency range.



View attachment 784946



Sound has a wave length that varies by frequency. The lower you go the longer the length of the sound wave.....the wavelength. In a closed space like the cabin of a car you eventually reach a point where the wavelength is longer than the interior space of the car. It wont fit. At this point it can no longer exist as a wave and just becomes pressure, cycling at the frequency of the sound. To a reference points....the 125Hz sound we are talking about above has a wavelength of around 9'. About the longest dimension of the Model S interior. So its definitely getting into that area where its becoming pressure more than a wave.

To cancel out that sound you play the speakers (probably just the subwoofer in this case if I had to guess) out of phase with that pressure creating destructive interference at that frequency and reducing the sound that gets to the ears of the occupants. Practically, you can only cancel a sound in a car once the frequency gets low enough that it becomes pressure, or perhaps you can stretch it to where the wavelength is long enough that the front/rear, right/left passengers are all existing within the same peaks and troughs of the wave. Trying to cancel frequencies above this point gets problematic as what might have destructive interference (cancellation) at the front seats may have constructive interference at the back seats....and you dont want that!

At frequencies above the point where the wavelengths are shorter than the size of the cabin, they can propagate around the car, reflect off hard surfaces, get absorbed by softer surfaces, and interfere with each other constructively and destructively in virtually infinite ways....basically its chaos....and it is not really practical to try to cancel out infinitely variable chaos.

If I had to guess, Tesla will probably refine this some, programming the software to more effectively target some sounds, ignoring others, sussing out mechanical noises like the HVAC or certain specific tires, also while tuning to the auditory perceptions of most people. Plus working out some bugs from what i seen in this thread... Overall I would expect the effect to continue to be subtle though, there are limits to what is possible/practical in this application.


Now I see a lot of people trying to compare this to noise cancelling headphones....you cant do that.
With a noise cancelling headphone the enclosed space and distance from the speaker to your ear is WAY smaller, maybe 1", so you CAN theoretically cancel sounds effectively up to that wave length....a 1" wavelength is equal to about 13kHz for instance....that's in the top octave of human hearing. So noise cancellation in that headphone can exist over a massively larger frequency range, plus WAY less chaos as its just the headphone, the ear, and a couple cubic inches of air space to deal with.
Thanks for doing this. I was gonna do the exact same thing once I get my ANC update.

One thing I wanted to do is put the Mic near the lower door speakers and the subwoofer to measure what is coming out.

I had ANC in my Jeep Summit (much quieter than any Tesla) but there was no way to shut it off. There wasn’t even an easy fuse to pull. What was interesting was the ANC was directly done in the AMP. Not the preamp. Which made sense.

The reason I mention that is. Does Volume setting matter? And Must Music be playing for ANC to be active? That’s what I wanted to measure. I was gonna try playing Silence. And measure at my head, lower door and at the sub. At various volumes. On rough road and on bridge seams. That’s a ton of work though. I often end up with less than satisfying results like your RTA.

Also I want to point out that BOTH the foam in the tires and some ANC systems are not focusing on low frequency noise. Because it’s so complex and not regular. But they are focusing on how the car transmits “thumps” (drumming the tires) when the tires hit seams in the road. So I was gonna find some bridge or road that has some good “thump thump” in them and measure that too. The roar you get from tires tends to be a mix of higher frequencies.

Oh one other thing the Jeep supposedly did cover was a low level exhaust noise from their V8 Hemi. Especially when it switched to eco mode. Using half the cylinders. It made an uncomfortable sound. I didn’t have the V8. It was a controlled frequency they could cancel.
 
Last edited:
After calibration completed my car now says "optimized for all seats" or something to that affect.

Ah, ok. Sounds like they very well might calibrate it differently depending on if there are rear passengers. I havent played with the setting while people are in the back yet. Will have to look at that next time someone is back there.
 
Last edited:
Oh one other thing the Jeep supposedly did cover was a low level exhaust noise from their V8 Hemi. Especially when it switched to eco mode. Using half the cylinders. It made an uncomfortable sound. I didn’t have the V8. It was a controlled frequency they could cancel.

Yeah, I have a Honda Ridgeline that has ANC as well, its primary purpose seems to be to counteract the odd 3 cylinder thump that happens when the 6 cylinder does cylinder deactivation. Its pretty transparent in use. Sounds like a similar implementation to your Jeep.
 
I have no perfect ear or anything but my experience seems to suggest turning on ANC amplifies the lower frequency - this is somewhat noticeable especially when I go over a bump. It almost seems like they are amping up the lower frequency to dull out any road noises? In any case, it makes my ear slightly uncomfortable and I like it better with it off.

Also, it's weird that ANC feature didn't get turn on right after I get the update (I was actively looking for it), but by 2nd day it somehow showed up.
 
  • Like
Reactions: WilliamG
I have ANC but it works really bad
It plays very noticeable sound when its on, and sometimes it produces boom sound that is very scary
I have disabled it and hope this will be improved

Cadillac ANC feels better and more mature
 
I have ANC but it works really bad
It plays very noticeable sound when its on, and sometimes it produces boom sound that is very scary
I have disabled it and hope this will be improved

Cadillac ANC feels better and more mature

Yes, it's really bad. I had this problem on my 2021, and it's the same on my new 2022 S. The sound it puts out is flat-out broken, and noisy. The car is so much quieter when the feature is toggled OFF.
 
  • Like
Reactions: Jabbahop