Sarah C. Rich : Can you tell me first a little bit about yourself? I read your bio but would love to hear a bit more about how you got interested in sound and how you connected or focused your academic studies around sound.
Stéphane Pigeon : I learned to play classic piano when I was a kid. As a teenager, I became fascinated by synthesizers, which are capable of creating their own, unique sounds. Musicians can modulate the sound of their acoustic instrument, by varying their playing techniques, but only in a constrained sound space: an acoustic piano will always sound like an acoustic piano. Musicians are bound by the sound crafted by their instrument's manufacturer, and by the long history behind their instrument. If they want another sound, they have to learn to play another instrument. Talented musicians do this.
Synthesizers give you access to numerous sounds, through a single user interface : a piano keyboard. And this was the interface I learned to play with — what good luck! Synthesizers opened my ears to new sonic worlds. Despite my 10 years of piano education, I was not particularly talented. With synthesizers, I could design both sounds and music. And little by little, I naturally compensated for my lack of playing skills, by programming complex sounds. Today, I realize I've just achieved the ultimate shift: some of my sounds are so complex that they will only need one note to create something pleasant to the ear that can last for minutes, the length of a song — just one single note!
In parallel, my scholarly education has been oriented toward engineering. I followed in my father's footsteps. Little by little, I realized that there is a lot of mathematics hidden behind music and sounds. Not the pseudo-science thing (this could be an interview on its own) but the real nature of sounds alone, physically. Most of the time, sounds are artists' playgrounds. The engineer's mind will develop an understanding of what an artist often naturally feels. But with that understanding comes the ability to manipulate things at their core.
With synthesizers, I discovered a passion for sounds. This passion put me in contact with Roland Corp., one of the most famous electronic musical instrument manufacturers, when I was a teenager. I started designing sounds for them, in parallel with my engineering studies. This gave me a common thread to my studies: naturally, I tried to orient things I was learning with things I was doing for Roland. The closest I could get, was to become an electrical engineer. From there, I then specialized in signal processing, and then forked into image processing (there was no sound processing available at my university). Image Processing was a good choice though. An image is a two-dimensional signal. Sound, is only one dimensional. I found out that everything that I learned with images could be transposed to sound… only much easier!
SCR : When I set out to write this story, the question I had was, essentially, “How is white noise designed?” What I expected, knowing very little, was that there were engineers mixing all kinds of different noises into one indistinguishable sonic blur, which washed out other ambient sounds. I even tend to imagine, when sitting in my baby's room as he goes to sleep with white noise in the background, that I can pick out familiar sounds from that blur — bells and car horns and water and voices. Then I read in Hillel Schwartz's book Making Noise that in fact white noise is “noise at any decibel level, to any ear, through any volume of space, at any time. It would be considered noise because, by definition, it was random and carried no information; it could be heard only as itself. "White noise," this would be called, by an inaccurate analogy to the optical white noise that is all spectra of light… White noise is patternless sound, 'featureless and unpredictable at all frequencies...'”. Having learned that, I thought, does this mean white noise is, in a sense, impossible to "design?" Is it basically one unvarying thing that results from the science of audio frequencies, as opposed to engineered noise? Can you help me understand which interpretation is correct?
SP : I take the opportunity of this question, and the fact that you are referring to your baby, to write about a frustration of mine that grew little by little as I worked with sound. As you will see, it relates to color, and babies, and will give me an opportunity to introduce the concept of 'frequency.' We will need to further discuss this interesting topic.
‘Frequency' measures the repetitiveness of a phenomenon. It is expressed in Hertz [Hz], which is equivalent to 1/s — which translates to "per-second". Something that repeats itself every second has a fundamental frequency of 1 Hz.
Light is an electromagnetic wave, vibrating at a given frequency. The frequency of light is something that we experience everyday: we can see that frequency and we call it color ! The lowest visible frequency is 400,000,000,000,000 Hz (400 THz) and is called red: below, it becomes infra-red, something that we cannot see anymore but feel as warmth. The highest visible frequency is violet (800 Thz) — above, is ultra-violet, which is invisible to us.
Sound is a pressure wave, which can be described by a frequency too. Audible frequencies are from 20Hz (bass) to 20,000Hz (treble). The order of magnitude is totally different, but it is of no importance. Below you have infra-sounds and above ultra-sounds. Note how with light, we used infra-red and ultra-violet but with sound, we only use infra-sound and ultra-sound.
Note: The nature of sound is totally different from light. In pseudo-science literature, you will find people who believe that light will turn into sound as its frequency decreases. This is not true. An orange will never turn into an apple, even if it has the same color. It is just not ripe yet ;-)
Here comes the frustration mentioned earlier. We gave names to the frequencies of light, but we forgot to associate names with sound frequencies. Why???
Light frequencies – the colors – are some of the very first things that your baby will learn. One of the first baby books he will receive as a present, will be the one with those colorful figurines, with the name of the color printed in bold letters: blue, red, green. As he grows up, he will learn myriad color names, and will be able to distinguish them... very precisely: this is not blue exactly, but turquoise, or cyan, or indigo… Later, when he will look at a color again, he will find the words to describe it precisely, and will be able to picture it in words to a friend.
Not with sound. The vocabulary doesn't exist yet. Something that has no words associated with, has no chance of existing culturally. Sound 'colors' are not in our vocabulary, and therefore, are not in our mind. Actually, we have three words: bass, treble… and medium for everything that is in between. How poor is this? We should invent new words for sound colors. Not even reuse the light color names, because this would lead to misinterpretation! Sound has not much in common with light. Such a vocabulary remains to be invented, and taught to our children.
That being said, it is convenient to associate colors with sound frequencies, because the color is indeed related to frequency, only for light. Again, these phenomena are totally different, so there is no "scientific" way to map color between light and sound — although much pseudo-scientific literature does. It is just convenient to use these color names, as the vocabulary for sound hasn't been invented yet.
At this stage, you may say that I am wrong: musicians developed a vocabulary to describe sounds, with musical notes. They have a system that works indeed… to describe music, but not sound. To convince yourself, click the following link: myNoise's White Noise then move any slider, then ask a musician to describe that sound. He won't be able to, even if he knows all the notes and has perfect pitch. The sound of these sliders, each need a name.
Now, I will answer your questions, precisely. Your understanding of white noise is exact. Few precisions.
White has a very precise scientific meaning, when it comes to a spectrum: it means that all frequencies are present, in equal proportion. It is indeed an analogy with light that turns white, as soon as you mix all the frequencies (colors) together in equal proportions. So, a signal (e.g. sound) that has all the frequencies present – from OHz to infinity - with the same levels, is described as white.
Because humans only hear between 20Hz and 20kHz, for us, all frequencies only means 20Hz to 20kHz. So, for humans, white noise and a noise that contains only the audible frequencies, from 20Hz to 20kHz, in equal proportions, sound the same. So, both are called white noise. This is a shortcut.
One can prove mathematically that only a purely random process is able to generate white noise. For the mathematician, white noise is a signal made of uncorrelated samples, exactly like numbers output by a random number generator. So, to produce white noise artificially, take the best random number generator you can get, draw numbers and save a series of them into a sound file. There you have it! There shouldn't be any loop, because then you break the randomness and add a repetitiveness equal to the loop size. One can show that this intruduces imperfections in the lower range of the 'white' spectrum.
You will it find interesting that there are random processes in the nature. Take a heavy rain, as heard from a distance. Each drop is part of a huge random process — you will never be able to determine precisely how many drops are crashing to the ground at a specific moment of time! If you look at this rain spectrum, it will be a kind of white noise. I invite you to take a break, and listen to myNoise's White Rain now, and read its description.
SCR : Can you tell me a bit about the other colors of noise and how the manipulation of the levels of those colors alters the quality of the noise?
SP : Our hearing is not linear. We do not sense the frequencies the same. As you understand this, you now understand that white (flat) is only an engineering concept: white noise is nice to make measurements, but has not that much sense for our hearing. Why? Basically, our hearing works with octaves (the musical term) or 'factors 2' (the engineer view) in the frequency range. We say that our hearing is logarithmic.
For the ear, there is as much sonic space between 20Hz and 40Hz (the first octave of our hearing) as between 10,000 and 20,000Hz (the last octave). So, 20Hz (=40-20) at the bottom of the spectrum, counts for as much as 10,000Hz in the top (=20,000-10,000). Yes, indeed. Any factor of two will sound similarly wide in frequencies.
So, starting from 20Hz, there are 10 octaves in our hearing: 20-40, 40-80, 80-160, 160-320, … 10,000-20,000. This is exactly why there are 10 sliders in myNoise generators!
Now you may start to understand why white noise is not natural and sounds too bright for our ears — too hissy, too high a frequencies. If you want to hear something flat, you need the octaves to produce the same amount of energy: there must be as much energy between 20 and 40Hz, as 40 and 80Hz, as… as 10,000 and 20,000Hz. This means it is not flat at all! Frequency must have amplitude decreasing as the frequency increases, to offer flat sounding octaves. This is the so-called Pink Noise. It has been called pink, because the light would turn pinkish, is we applied the same principle to light.
That being said, pink noise doesn't sound even flat to the human ear, because there are other issues with the ear, such as an increased sensitivity of the ear around 1k to 3kHz. As a mother, you may be interested to know that this is the typical frequency range of your baby crying. It seems that during evolution, it made sense to develop an improved hearing in the frequency range of crying babies, or – the opposite – have babies who cry at the frequencies where our hearing is the most sensitive. I think you can easily guess why the individuals who developed that ability had more chance to survive.
So, pink noise is just an approximation, to get a spectrally flat sounding noise.
The ultimate color is grey, which is the only noise that is spectrally flat to the human ear, once everything has been taken into account. There is no generic grey noise, because each human is different. Grey noise must take your particular ears into account. myNoise.net is the only place that is able to produce such a perceptually flat noise. As you just learned, to do so, I have to measure your hearing. This is done online, through the so-called calibration process:
Little by little, you can understand why sound is an engineering thing too (see intro).
SCR : I read somewhere that people refer often to "white noise" but many prefer "pink noise," which made me wonder whether our preferences for noise quality may have changed as your soundscape has changed. For example, in a place where a soundscape tends to be more industrial (cars, construction, etc.), do we end up favoring "white noise" that carries frequencies that are either closer or farther from industrial highs and lows?
SP : You now understand why pink noise makes more sense for our hearing, and why grey is even better. However, for many applications we should NOT care about 'white', 'pink' or 'grey'; for example, sound masking. If you use (white, pink, brown) noise to mask a disturbing sound in your room (your neighbors) or tinnitus, what you really do care about, is if the noise effectively masks the nuisance - no matter what the color. I'd say, give me any color, as long as it is the one that masks the nuisance best!
Most of the white noise machines – online and available as an app – are so un-innovative. They just let you chose between class settings such as White/Pink/Brown, and argue about which one is better. What we really need is full control over the spectrum, so we can shape the noise exactly as we want, as to increase its efficiency for the task we want the noise to fulfill. This depends on the nature of the sound you want to block, but also on your hearing sensitivity, the speakers, and the audio system. Yes, the optimal noise machine has to take everything into account! This is not marketing. It's evidence. But few have found it yet, or don't know how to tackle the problem.
Let's take an example. Consider that
- The noise that disturbs you though your wall has a predominant peak around 300Hz
- The headphones you are using have a peak around 1kHz
- Due to ageing, your hearing is not good from 4kHz and above
Such a situation makes it clear that a white/pink/brown spectrum will not solve the problem optimally. You need something that takes everything into account: boosting the noise around 300Hz to make it better mask the nuisance; reduce a bit around 1kHz to compensate for your headphones' peak; and boost frequencies higher than 4kHz, to compensate for your hearing loss. Now you understand the power behind the 10 sliders offered by myNoise.net Either you move these sliders experimentally until you are satisfied, or you use the calibration process, and load your calibration data. This is the recommended way, as it is easy to perform, takes less than a minute, and is optimal. Afterward this data can be used toward any of other calibrated generators.
For example, go to the rain noise page, then load your calibration data (Personal > Your Hearing Curve) assuming you have already performed the calibration before, and suddenly you will hear the rain as you use to, when you were a child, and your ears were still in perfect shape. It is a unique experience and mood changer for many who suffer from a hearing loss.
SCR : Our baby is 11 months old and we have sampled a handful of noise-generating apps to play while he sleeps. The one I prefer is "waterfall" as it, to my ears, is the closest to the one simply called "white noise". However, I'm absolutely sure that I can hear the looping of the track when I sit and listen to it long enough. I could almost reproduce the embedded sound I hear on the piano. Would this be a distinguishing feature of "engineered" sound as opposed to white noise? As I understand it, white noise by definition has no loops, while almost every other manufactured sound you can buy in the app store must loop. True?
SP : myNoise's waterfall noise is calibrated. This means that I did a lot of heavy processing offline, as to be sure to generate a pink spectrum when all the 10 sliders are aligned. This is the base line. From there, you can move the sliders and produce any color you want. If you look at the slider presets, you will find the other colors. Just click them and see the sliders moving accordingly. This all means: myNoise's Waterfall noise is a heavily engineered sound generator, but I defy you to hear any loop. And to have an ever-changing sound, just hit the Animate! button. Then, for sure, there is nothing that repeats itself! This is something that any loop-based generator will never be able to achieve, and makes all the classic noise machine apps obsolete.
SCR : I don't know how much you've studied the effects of sound/noise on our psychology, but I'll admit I've wondered at times whether hearing one of these artificial noise generators for hours at a time during sleep could have an effect on my baby's brain function. If he subconsciously registers that subtle looping of sound, will it impact the way he processes or thinks? Do you know anything about this or have any opinions?
SP : Hearing looping points is something that I cannot stand. It drives me crazy. I never would be able to fall asleep with a simple (short) loop-based mp3 player: my mind would anxiously wait till the loop cycles, and say there it is and then wait for the next cycle to complete. However, I don't think a baby has the cognitive processes mature enough to be able to care about that yet. But, if you have a choice, I'd say: you would prefer to use a sound where the looping is inaudible, over sound that obviously suffers from a horrible loop pattern.
I found out that some people have analytic hearing, and some other don't. Some are bothered by the tiniest loop hint; others won't hear an obvious and clumsy loop.
I can imagine that if you maintain the practice of playing clumsy loops to your baby, his brain will soon learn not to be bothered by that loop (I hope so). Then, he will probably be the one with less analytical hearing as an adult. To me it makes sense, although it is only a supposition.
SCR : Related to that, do you know whether artificial sounds can be engineered "better" or "worse" in the sense that the loop could in theory be made totally imperceptible and it would have a closer resemblance to white noise? Sometimes I imagine if I could just livestream Niagara Falls then I'd have a loop-free waterfall in my ear, whereas there's no way to play a recorded Niagara Falls that doesn't eventually repeat its pattern.
SP : The loop can in theory and practice be made totally imperceptible, especially with a sound like a waterfall. If you hear a loop in such a sound, it really is the sign of poor engineering skills.
Yes, there is a way to record Niagara Falls that will never repeat itself, almost. The waterfall noise on myNoise.net comes from a real waterfall recording, and will repeat in exactly 188.027.101 years and 193 days (I just did the computation, based on the parameters of my sound engine, and this particular noise). Would that be OK? I think so ;-)
SCR : Any other thoughts you have on the design of real and artificial white noise? Is there even such a distinction?
SP : White noise is a theoretic concept. In order to implement it right, you need infinite time and frequencies. For example, a sound that only exists for a finite duration is not white — it will miss some of the extra-low frequencies. But once you understand human hearing, and the applications you want to address, you soon realize that you don't need the noise to be white.
White, or Pink or Brown, are very simple settings that have been used because people never realized how to provide the right noise. Custom color is the answer. It's a noise whose color compensates for your personal hearing thresholds, and the flaws of the speakers or headphones you are using, and takes into account the characteristics of the sound you want to mask.
We are still living in the ice-age of noise machines, still stuck in these white, pink, and brown presets. These are quite efficient for their simplicity, but do not stand up to the efficiency of custom-shaped noises.
Higher efficiency, means the same masking only at lower sound levels. The masking properties of white noise are no miracle: you are playing a sound above the nuisance in order to mask it. This means that you end up with a total sound pressure that is higher than the initial nuisance, but more tolerable for the brain, as you now are hearing a static sound instead of a nuisance. Static, after a while, disappears from your conscious perception — how wonderful our brain is. Yet, the acoustic level is there, higher than the nuisance in first place. You are bombarding your ears with a lot of frequencies, and some are useless in your particular case, and others are louder than necessary. Custom-shaped – when designed right – is able to reach the highest efficiency, the optimum spectrum. This means less sound pressure, for the same result. In practice, this is less fatigue for your head.
Just a last word regarding my philosophy related to my Internet sites:
I learned so much from the Internet, and I still do every day. It is amazing to realize how many high quality resources are available, free, just a few keywords and clicks away (thank you Google). We do live in an exciting era of shared and easily-accessible knowledge. However, one easily forgets that those resources do not come about spontaneously, and are the result of passionate and knowledgeable people, giving a lot of their free time to the community. When I realized this, the time had come for me to pay back...
myNoise is available for free, and so are the noise machines. Actually, you get almost everything for free - definitely all the essentials to treat serious problems such as tinnitus, ADHD and panic attacks - and the rest, what I consider the topping, is just there to thank people who help support the site, as a bonus. Without a bonus at all, the support rate drops drastically; this is human nature. I need this support to gradually switch from my regular day job to maintaining myNoise.net and keep developing new technologies. I have many ideas; I just lack the time to implement them as long as I have to keep my regular day job.
New generators are added almost every week. They are available for patrons first, but turn public after a while. That is to say that this website is definitely not commercial but the work of a crazy passionate guy, trying to help people the best he can… and whose satisfaction comes from providing innovative solutions to everyday problems.