tl;dr We demystify the secret sauce used by Apple to deliver on-par active noise control performance without actually doing so. It’s smart product decisions and psycho acoustics!
If you are looking for an Apple AirPod product review, go elsewhere.
As soon as Apple announced their 3rd generation of wireless earbuds called AirPod Pro complete with integrated active noise reduction we knew, we at Rocket Science had to get a grip on them to test their performance in our lab in Zurich. Quite marvelous what Apple achieved in terms of form factor, as so far active noise control remained reserved for rather bulky over-ear headphones. Apparently they built custom chips besides their H1, one of them, most likely, integrating the active noise control functionality without the need for dedicated (power hungry and large) digital signal processors. See what iFixit uncovered in their teardown.
In order to have a reference for comparison, we gathered some available over-ear headphones. We then performed some basic tests to get a decibel reference for noise attenuation, created by the active noise control algorithms. We isolated the algorithm performance, as we were not interested in how much the foam and plastic would contribute to the attenuation (actually the Apple earbuds do a great job here as they fully block your ear channel). Our main goal was to understand what the tiny chips in those AirPod Pros could really do.
The results were surprising!
Apple AirPod Pro
0 dB(Z) attenuation, 2 dB(A) attenuation
Sennheiser Momentum 3
Note: higher numbers mean more attenuation
Before we go through the chart, first a few words about decibels. The scale is logarithmic, hence an attenuation of three decibels is the equivalent of 50%, where as six decibels is 75% and so forth. It actually reflects the behavior of our human hearing, which does not perceive 50% less noise (in terms of noise energy emitted) as 50% less. You’d most likely think it is just a tad bit less. You’ll need to have reductions of more than 90% (which is more than 10 decibels) for humans to perceive it as an actual, noticeable reduction of half.
Second, decibels come in all sorts of scales. This can be confusing at times. dB(Z) is the ‘virgin’ decibel without any weighting. It does make more sense to talk about dB(A) however, when talking about human perception. This so called A-weighting takes into account, that humans perceive low-frequency noise as being less loud than mid- or high-frequency noise.
Wait, the AirPod Pros algorithms don’t do anything?
Keeping this in mind, the numbers in the above chart start to make a little more sense.
Bose — the original inventor of active noise control headphones — is leading the game. It is quite visible, that they have improved their algorithms significantly for higher frequencies (as far as we see mostly between 500–1’000 Hz) coming from the QuietComfort (old) series, compared to their new Bose 700 release. This improvement is not only visible on charts but is perceived better attenuation quality.
Yet it is no surprise that others such as Sony or Sennheiser have similar performance, as the technology has been around for quite a while now. It is surprising though, that Apple active noise control algorithms seem to ignore low-frequencies entirely. Further, their performance does look far from stellar — which is in complete contradiction to perceived attenuation quality. Wear them and you’ll agree that they work. They for sure have the ‘wow’ effect when switching their active noise control from ‘off’ to ‘on’.
Demystifying the secret sauce
In order to understand what is going on here, let’s dig a bit deeper.
For this, we took a look at the relevant frequency spectrum and plotted the active noise control attenuation performance of each headphone. You can see the Bose 700 (green) reliably reducing noise starting at about 50 Hz up to about 2’300 Hz. Apple on the other hand starts performing at around 500 Hz up to about 2’000 Hz and then even introducing an overshoot above that.
Now, how does Apple do it?
Our theory is that Apple introduced a secret sauce (they even call it magic on their website) with a default feature called ‘transparency mode’. It amplifies frequencies in the frequency spectrum relevant for speech and gives the user a feeling of not being fully isolated. As this feature is on by default Apple introduced their very own reference baseline. When users switch active noise control from ‘off’ (equaling transparent mode) to ‘on’ the perceived attenuation is very large. This can be seen in the chart, visualized by the red line, which compares the ANC performance relative to the transparent mode). It shows a large reduction of up to 20 dB(Z) directly in the sweet spot.
Is Apple cheating?
No. Apple understood how humans hear and took very smart decisions in implementing the technology. It is about the result and not about who gets the best dB attenuation. Interestingly, this may also be one of the reasons Apple does not even talk about these numbers in their marketing (unlike some others). These decisions allowed Apple to build a device with both a very small form factor (as they do not require ludicrous processing power) yet delivering performance perceived to be on-par with others such as Bose.
Have fun with your AirPod Pros!
About the author
Rocket Science is a deep-tech algorithm boutique. The folks at the Swiss startup specialize in active noise and vibration control algorithms. They believe, using noise reduction algorithms outside of headphone applications is [not] rocket science. They work in applications ranging from power plants, industrial machinery, consumer products up to aerospace.
Get in touch with us at www.rocket-science.ch.
Disclaimer: We are not affiliated with any of the companies mentioned. We are aware of the fact that comparing in-ear headphones with over-ear headphones might not be the fair thing to do. Yet, it seemed appropriate for the task at hand. If you have questions about the measurement setup we used or our conclusions, please do get in touch with us. We’d be happy to chat about it but think mentioning all the details in a post like this does not fit the form. We understand it is all about the details! Further, the selection of headphones is random as we used what was lying around. Feel free to send us a pair so we can include it in our tests!
This article has been originally posted on medium.com by Rocket Science on December 2nd 2019.