Audio equalization with band noise
Here, I’ll be describing a simple yet powerful method for audio equalization that I found on some audiophile forum a few years back, and never could find again. It’s something that can be done at the software level, at the meager cost of configuring a 10-band equalizer for your system, which may or may not have one built in already.
If you live in a world where you have access to affordable audio gear that strives for a linear response at every link of the chain, which easily produces a frequency response that sounds even to your ears at every volume level in every room possible, then you can stop reading to saunter back into that magical place. I myself don’t, and most of the population does not live in that world.
Recorded sound usually takes many steps until it starts wobbling the air near your ears. In the case of digital audio, it has to be converted to analog and then amplified in multiple steps until it can drive some kind of speaker, to finally let the waves propagate (and interfere, and reflect) across a room to then combine at the point you’re standing with your ear’s response curve. Each of those points introduces it’s own kind of distortion to a response curve. The rabbit hole goes pretty deep here, some people even try to use power supplies that try to eliminate any noise that may be in their grid.
Traditionally, the solution was to spend many years and eye-watering sums combining and configuring gear until you got something that sounded good to you. Most of us cannot engage in this for understandable reasons. Instead, I think it’s worthwhile to spend some time tweaking your software to get the best sound out of whatever hardware you happen to have. The tone/timbre/coloring that your speakers have, or any weird detailing is intrinsic to them, and we won’t be focusing on that here. This method relies on the assumption that the volume of their response curve at every point plays an important part in how they sound, and that trying to equalize the heard sound pressure level of the whole spectrum will help your ear parse whatever you’re hearing, and will lead to a clearer and plain better sound.
Compared to measuring a reponse curve in a room with a microphone of a well-known response curve, then trying to negate that with a generated equalization curve, this method is dead simple, yet fairly holistic for what it is. The idea is to turn the volume up until the bands reach a comfortable volume for listening, then trying to tweak 10 bands with small offsets to try to compensate for wide hills and valleys in your hardware (and in how you hear it). It’s simplish and coarse, but it also directly targets how well you hear your audio.
Download the sample file(s), and play them back through your hardware at listening level. The fastest way I’ve found is to just use the file that plays all 10 bands, each band for one second. I wish I knew enough about audio to tell how this band noise was generated, but I guess it’s not impossible to figure it out if you want to generate band noise for more or fewer bands, or if just want to make your own to be sure. Kudos to whoever figured out how to generate these bands, though.
Listen to any bands you suspect to be higher or lower than the others, and make small adjustments to your offsets until they start feeling equal. This sounds ridiculous, but that’s all this method is. Just try to get each band equal enough that you “feel” a similar “pressure” on your ears.
Over the years, the many systems I’ve tried this on showed that some things really cannot and should not be covered by this method. When adjusting my offsets, I try to keep the following in mind:
- try to stay within a 3-4 dB range (a +/- 1.5-2dB each way in other words), it should usually be enough, otherwise you may have worse problems than slight inequalities in your response
- be careful with adjusting the high frequencies, because response curves tend to go a bit wild around there… turning those frequencies up may just end up amplifying some small, sharp peaks into an uncomfortable range
- leave the bass frequencies for last, and do not try to calibrate them purely based on this method; always listen to your favorite tracks to make sure the bass hits comfortably compared to all the other bands (this mostly applies to the 31 and 62 Hz bands, and a bit to the 125 Hz, which is usually the first you hear well on smallish speakers)
The irony is that on the other end of the dough spectrum, this method can bring little to no benefit on a system that is selectively assembled so that the final sound ends up pristine in every way. However, merely choosing the right components can bring you pretty close to a clean, pleasing tone: I’ve listened to these bands on a mid-end system once, the components of which were painstakingly collected to work together as well as possible, and found myself wanting to make no adjustments.
The above method helped me many times to push whatever (often low-end) hardware I had lying around (everything from the cheapest earphones and headphones all the way to integrated amplifiers and separate DACs) to sound the best it feasibly can, and I hope it will help you to do that just as easily and cheaply. I happen to think that if we’re stuck with whatever hardware we have, pushing it to still be the best it can be almost always makes a world of a difference, and allows us to extend its life, to keep it out of landfills.