What You Need To Know About Computing Audio

- Dec 30, 2020-

Let's talk about EQ, which we love and hear. EQ is equalizer, which is originally realized through digital signal processing. For example, the usual multi-band EQ belongs to the second-order Bi-Quad filter, which is a kind of IIR filter. Digital signal processing, or DSP, is originally calculated and processed by the corresponding chip. So all equipment and software with EQ can be said to be "computing audio".


Not to mention other more advanced DSP processing, just talk about EQ, which is widely used in Bluetooth headsets, Bluetooth speakers, active speakers, home theaters, cinemas, car audio, line arrays, sound columns, subwoofers, mobile phones, laptops, Tablet PC, TV and other products.


And even for earphones, except for a few cases like AKG K371BT/361BT, almost most Bluetooth earphones have EQ processing. The final sound and frequency response curve of the Bluetooth headset has multiple EQ corrections and adjustments, and it is usually PEQ (Parameter EQ). As long as you just spend dozens of dollars on Taobao to buy a Qualcomm CSR Bluetooth module, you can roughly understand what functions the current Bluetooth chip contains. (Current Bluetooth chips usually also include more DSP functions such as limiter, ASW extension, etc.)


In other words, basically all Bluetooth products are "computed audio".


Of course, if it is some other products, it may even be much more complicated than Apple Sony. For example, the DSP of some home theater power amplifiers of Denon has 4 ADI chips. Of course, these DSPs also handle the video part.


Let's take a look at what Apple has calculated.


It is nothing more than active noise reduction, transparent mode, adaptive wear and spatial audio.


Oh, yes, active noise reduction also belongs to DSP. Because, in layman's terms, active noise reduction is to pick up noise AD samples, calculate an inverted sound through DSP and then play it through DA. The process of DSP processing noise can be regarded as computing audio. In this way, all ANC headphones should be considered "calculated audio."


As for adaptive leakage compensation, from the current results, it seems that it only works at low frequencies, and the effect of high frequencies is still questionable. And adaptive EQ is a function that at least I think is more gimmick than practical, superfluous, or a marketing function. It should be that anyone who has an understanding of the entire audio production and playback link, or the "audio circle", will understand that any EQ for music types is amateur and has no practical meaning.


If you think about the "audio cycle" carefully. You will understand that this kind of "tuning" is not helpful in improving the chances of users buying quality products. ——Dr. Floyd Toole



One of the audio industry claims I hate the most is to include music genre EQ settings or audio companies that design/sell for specific music genres. As this article suggests, the basic logic here is that different music genres require different speakers/headphones for playback. What is even more absurd is that a single EQ will magically repair or enhance the sound of all recordings of that type in some way. This is total bullshit (before you understand the science)! !

When I saw such a product, it immediately sent a message: the person who designed it does not understand audio, and it is likely that the marketer will win the audio engineer/scientist battle.

If there is any specific EQ suitable for a genre of music, it has already been added during the music production mix. As long as relatively accurate/neutral speakers are used in the process of recording/making music (this has become the norm), the recordings played through accurate headphones or speakers will be retained.


Adding more EQ will backfire and will only add more confusion to the audio loop.

So let's stop this wrong approach!

-Dr. Sean Olive


Even some more advanced algorithms can identify based on the content of the musical instruments included in the music, not just the style of the music, how does the algorithm recognize the creator’s intention when mixing, or the monitoring equipment used by the creator? And if the speakers used by music producers are relatively naturally balanced, then as long as the headphones in the playback link are also naturally balanced, do you need to do more adaptive equalization? Similar and more advanced functions, I have seen people implement AI and machine learning before, but the manufacturers gave up because they thought it was meaningless.


On the other hand, even the basic EQ and frequency response curve are not perfect. DSP is just a platform, which can optimize the sound or degrade the sound. The specific quality depends on the algorithm and specific debugging parameters carried by the DSP. The effect evaluation should be based on the actual results. Rather than as long as there is a chip, the sound will be as good as magic, so almost all Bluetooth headsets should sound good, but in fact it is far from the case.


For speaker systems, DSP is nothing new. Whether it is Dirac Live calibration or miniDSP mentioned some time ago, whether it is Soundbar or DSP pre-stage, it is all digital signal processing.