How FM Radio Works: A History and Exploration of Frequency Modulation

I recently saw a video on the VWestlife channel here on YouTube talking about the MPX filter functionality on cassette decks equipped with Dolby Noise Reduction circuitry. I'll link the video in the description and I encourage you to go watch it, but for now I'll tell you this feature has something to do with FM radio. Since as a broadcaster I knew immediately what MPX was, that video gave me the idea to talk more about FM and what broadcasters do with it. Broadcasting is complicated and for you to understand some of the things we do, you need to understand why we do it first. The history of the development of FM is a compelling story, so I wanted to tell part of it, then we'll ease into the fun technical stuff. For background, any sort of periodic wave has three basic properties. Peak-to-peak amplitude, which is the distance between the high and low peaks of the wave. Frequency, which is how many times the wave cycles per second. And phase, which is a figure representing how far through the cycle a wave is at a given instant. Phase is a more confusing property and somewhat outside the scope of this video, so we'll leave it alone for now. The term modulation means changing one or more of those wave properties in a defined fashion to convey information. In radio broadcasting, that information has traditionally been an analog audio signal. Commercial radio broadcasting in the United States got its real start in 1920, when the first scheduled broadcast transmissions were started within weeks of each other by WaterNow KDKA in Pittsburgh and WWJ in Detroit. Back then, all radio broadcasting was accomplished using Amplitude Modulation, or AM, which varies the amplitude or power of a radio wave, sometimes called the carrier, at a given frequency in proportion to an input signal, like the fluctuating voltage representing a sound wave. You can think of it like looking carefully at the changing brightness of a light bulb and writing down how bright it is. AM is a very simple technology, but it has its downfalls. Imagine you're trying to look at the brightness of this light and record how bright it is, but someone has come in and started taking pictures with the flash on. All the extra light flashes make it hard to get a good read on the brightness of the bulb at that moment, so some of your data might be wrong. Those flashes of light can be compared to real lightning. Not only does lightning create a bright flash of light, the static electricity discharge also generates a lot of radio frequency interference. If you try to listen to an AM station during a thunderstorm, your radio will pick up this RF interference and interpret it as static crashes every time lightning strikes nearby. That's not desirable when you want to have an accurate, clear reproduction of the program material, and lightning isn't the only thing that causes RF interference. In 1928, Major Edwin Howard Armstrong, who had developed the superheterodyne circuit 10 years earlier during World War I, started researching a way to reduce or eliminate the vulnerability to RF interference in radio communications. John Carson of AT&T had concluded earlier in the decade that there was not a viable solution, but Armstrong wasn't convinced because Carson had based his findings entirely on mathematical analysis rather than experimentation. Carson had explored frequency modulation, or FM, in his work to eliminate static. FM works by changing the instantaneous frequency of the carrier wave in response to the input signal. Instead of increasing and decreasing transmit power, FM transmissions increase and decrease the speed of the wave oscillation, thus changing the frequency while keeping the amplitude at a constant value. A lack of an input signal would keep the carrier at its nominal assigned frequency instead of swinging slightly above and below it. Conventional thought at the time hypothesized that since FM transmissions have a constant amplitude, that alone would be effective at reducing a radio transmission's susceptibility to interference. To go back to the light bulb analogy, this would be like moving the light bulb from side to side and writing down where it is in relation to its resting place instead of how bright the light is. John Carson's 1922 report to the Institute of Radio Engineers argued that FM, despite having a constant amplitude signal, really was not any better than AM at rejecting common forms of radio frequency interference. But Armstrong noticed a flaw. Carson had kept the maximum amount of deviation, or how far the carrier's instantaneous frequency was allowed to deviate from its assigned resting frequency, was too low compared to the range of audio frequency signals used to modulate it. Carson had invented single sideband modulation, a type of AM, in 1915 and mistakenly assumed that like traditional double sideband AM, transmitting sound frequencies of, for example, up to 2.5 kHz required occupying an RF signal bandwidth of an amount equal to that on either side of the carrier frequency. I'll explain the mechanics behind this later in the video. Carson, in developing single sideband modulation, had reduced that by half for AM. He limited the deviation in his FM experiments to the highest sound frequency he expected to modulate, resulting in a total signal bandwidth of twice the highest audio frequency expected to be reproduced. If this sounds familiar to you, the Nyquist-Shannon Sampling Theorem, serving as the basis for all digitally encoded sound, has a similar constraint, which has been covered in depth by Alec over at Technology Connections. However, Carson's research didn't prove entirely useless because it showed there was a lower limit to the amount of FM deviation needed to reproduce sounds of a given frequency. This limited deviation resulted in the light bulb not moving very far from the line it rests at, meaning that in an environment with a lot of noise and bright flashes, you might not be able to tell how its position has changed amid all the commotion. Several researchers in the years following Carson's 1922 paper further explored the concept and universally concurred with it, once again believing the key to reducing static was keeping the total RF signal bandwidth as narrow as possible to keep the signal-to-noise ratio high. These dudes were stuck on AM thinking, and they needed to have their horizons expanded a bit. In 1928, Armstrong began a five-year journey to create a completely new broadcasting standard to finally make the static-less radio a reality. FM still makes use of a superheterodyne receiver, but instead of a simple product detector sensitive to changes in amplitude, an FM receiver then passes the received signal through an amplifying stage, utilizing a limiter that isolates frequency changes while reducing the effects of amplitude changes, then a signal demodulator sensitive to differences in frequency called a discriminator that yields the original modulating signal. From that point, the demodulated signal is then ready to be amplified and sent to headphones or loudspeakers, just like an AM radio receiver. Armstrong quietly received patents in 1933 on his improved FM system and offered them to RCA thanks to an agreement giving RCA the first shot at any of his inventions. RCA's president, David Sarnoff, was surprised at the drastic changes required and turned down Armstrong's offer. In fact, RCA had focused so much on developing television systems at the time, believing they would replace radio outright, they chose not to invest in radio at all and later denied Armstrong access to their labs and equipment. As a result, he sold much of the RCA stock he received in exchange for the superheterodyne circuit and proceeded to self-finance further experiments in partnership with other radio companies like General Electric, Zenith, and Philco. In 1936, Armstrong published his own paper in the same Institute of Radio Engineers journal where Carson and others had insisted FM wasn't the way to go. In the 1936 paper, Armstrong meticulously laid out his new wide-band FM broadcasting system and demonstrated the experiments. He proposed a system allowing carrier deviation of 75kHz for a total signal bandwidth of 150kHz and detailed his tests of using this system to transmit audio frequencies up to 13kHz with much higher quality and resistance to interference than AM. As it turned out, increasing the maximum carrier deviation not only gives you more resolution in demodulating the signal, making the resulting audio clearer and quieter, it also makes the relative difference in instantaneous frequency larger for a given input signal, which makes it easier to follow the swing of the carrier wave in the presence of interference. This was an earth-shattering occasion in broadcasting. While the new wide-band FM system was incompatible with existing broadcasts and much more complex, it opened the door for the quality of entertainment broadcasts to increase dramatically. Later that year, Armstrong presented his wide-band FM broadcasting system to the newly formed Federal Communications Commission. A press report of this event quoted engineers saying this was one of the most important radio developments since the first earphone crystal sets were introduced. Armstrong also detailed some curious behaviors of FM in his paper. In AM broadcasting, since the radio receiver is simply detecting the amplitude of any radio waves at a given frequency, being in the range of multiple AM stations transmitting on the same channel will cause a receiver to demodulate all of them simultaneously. This interference would be audible if a distant station had as little as 5% of the signal strength of a stronger station. This required that AM stations on the same frequency be very far apart from one another and use directional antenna systems, and many of them had to stop broadcasting at night to avoid ionospheric, or skywave, radio propagation causing interference with higher classes of stations. In later years, some of these so-called day-timer stations were allowed to continue broadcasting at night as long as they reduced their transmitter power to that of a literal light bulb or less. The highest classes of AM stations even today are called clear channel stations and enjoy the privilege of being the only station broadcasting on their assigned frequency and the authority to broadcast a non-directional signal at maximum power 24 hours a day, with a few exceptions. As an aside, these characteristics unfavorable for broadcasting are why aircraft radio communications specifically use AM. However, with FM, Armstrong described an effect where a distant station on the same channel needed to have a signal strength of approximately half that of a nearby station to even exhibit any noticeable interference. Two years later, engineers at General Electric took this research further and determined the zone of interference between stations was very narrow and wideband FM receivers would only demodulate the stronger signal. In fact, these engineers found that simply moving a few steps in any direction inside this zone of interference would occasionally make the receiver switch entirely to demodulating one station or the other without any noticeable interference. Today we understand this phenomenon as the capture effect, and this suppression of weak signals allowed FM stations on the same channel to be spaced closer together than would be possible with AM. In North America, however, regulatory agencies typically do not allow code channel stations to be close enough for anyone to experience the capture effect in practice. By early 1939, Armstrong had received experimental authorization from the FCC to operate a high-power FM station under the callsign W2XMN. Broadcasting Magazine detailed the new system in their February issue, and by 1941, the FCC officially designated a block of 40 channels between 42 and 50 MHz for FM broadcasting in the United States. In 1945, to avoid unpredictable radio propagation in the low VHF band, the FCC designated a new high band for FM broadcasting from 88 to 108 MHz, which is the same radio frequency range used in most regions of the world today for FM radio broadcasting. So, how does all this work? Armstrong's wideband FM system greatly increased the distance the light bulb moves so that a greater change in its position is required to produce sound waves, and makes it easier for receivers to follow changes in its position, both of which work to reject interference. You may be asking how something moving side to side produces sound, but maybe using something that actually produces sound in this manner instead of a light bulb will help. How about a guitar? You can imagine the instantaneous frequency of an FM station's carrier wave as a guitar string. A real guitar string rests at a certain position when it hasn't been plucked or picked and does not produce sound. Accordingly, when there is silence at the input of an FM transmitter, the carrier rests at its assigned center frequency, which is the frequency displayed on your radio when you tune into an FM station, and does not change. When a guitar string is plucked, it vibrates at a specific audible frequency and produces sound. An FM transmission does the same thing when modulated by a sound recording, except it can vibrate at many different frequencies because it's being controlled by an electrical signal rather than the physical properties of an object. To show this in action, I've gathered some computers, a software-defined radio receiver, a USB sound interface, a broadcast audio processor that can provide an input to an FM transmitter, an actual FM transmitter, and a dummy load to absorb any radio signals coming from the transmitter. I've also brought my oscilloscope and some other measuring equipment. Dummy loads are really important when testing transmitters. They provide an electrical load that looks just like a tuned antenna to a transmitter, so the RF energy is totally absorbed and converted into heat. A dummy load also keeps nearly all of that RF energy from radiating. The FM band is a licensed band, and while I do have an amateur radio license, I don't have a license to broadcast in the FM band, so when doing stuff like this, you need to stay legal. To get an idea of how good a job this dummy load does, this is what I had to do to get a usable signal on the SDR across the room. This is a high-gain antenna sitting next to the dummy load, and I have the transmitter power output set to nearly 30 watts. Some real low-power FM stations don't have a TPO this high. First, let's turn the transmitter on and look at a graphical representation of the RF power it's producing on a frequency axis. This FM transmitter is tuned to 95.7 MHz and produces a single radio wave oscillating 95.7 million times per second, but a graph like this is more useful in analyzing the frequencies than looking at the actual RF sine wave. Graphs like this are produced by a mathematical operation called a Fast Fourier Transform, or FFT. The software I'm using also helpfully takes snapshots of the FFT and plots the signal intensity over time below the FFT, sometimes called a waterfall display. This will be helpful to show how the FFT changes over time, which demonstrates exactly how FM works. This transmitter doesn't have anything connected to its input, so what you're seeing here is an entirely unmodulated carrier. An FM station broadcasting silence, or dead air, would look just like this. If you recall the guitar string from a few moments ago, an unmodulated FM signal is like the guitar string at rest. What I'll do now is use a function generator to send a 1 Hz sine wave to the transmitter. Humans can't hear sound at frequencies this low, but the sound wave will change slowly enough so that you can see the FM carrier swing back and forth through the channel at the same frequency as the sound waveform. If you take the waterfall display and rotate it 90 degrees clockwise, you'll see that the FM carrier wave's spectral intensity traces that exact same waveform. This one appears inverted because higher frequency radio waves were on the right in the original orientation, but a radio receiver would understand it correctly. This is the very definition of an analog modulation method, because the modulation of the carrier wave is analogous, or a continuously variable direct copy of the modulating signal. From this, you can assume that the amount of frequency deviation in the FM signal is directly proportional to the amplitude of the modulating signal. Louder sound equals wider swing. You can see this in the FFT and waterfall display as I adjust the amplitude of the sound wave up and down. If I apply a sound waveform representing music to the transmitter's input, you'll see the FFT showing the FM carrier swing back and forth through its channel wildly, tracing the same waveform right there on the waterfall. A radio receiver uses circuitry and, more recently, software to follow that carrier wave and convert this frequency modulation back into the sound waveform that produced it. The 75 kilohertz deviation limit in FM broadcasting is defined as 100% modulation, and the amplitude of the audio signals going to the transmitter's input should be carefully adjusted so the modulation doesn't exceed 100%. But this is a regulatory limit, and not a technical one. There's nothing stopping me from turning up the gain and over-modulating the transmitter, but government regulators consider that harmful interference, and will levy fines or other sanctions against broadcasters who use excessive modulation. Broadcasters commonly use devices to handle the adjustment of these signal levels automatically. These are called audio processors, and I made a video back in March of 2022 that explores these in greater detail. I've linked it in the description so you can go watch later if you haven't already. This is an Omnia 6 broadcast audio processor. The final processing stage in a device like this is typically a clipper that ensures the modulating signal going into the transmitter absolutely does not go over a set limit. There are usually a few limiter and automatic gain control stages before the clipper so that it's a last resort, and broadcasters also employ some form of dynamic range compression so that very quiet sounds have their levels boosted a bit in the broadcast. When I played the music a moment ago, I was going through the audio processor, but I only had the final clipper enabled. You could tell that the FM deviation was only using a fraction of the bandwidth of the channel, which is shown highlighted on the FFT. Now, let's do it again, but with a sound enhancement preset enabled. The music sounds bigger, and more importantly, the modulation level is higher. The movement of the carrier wave takes up more of the channel. If you look at real FM stations on the FFT, you can see the exact same thing. I've tuned to my local NPR station which has a lot of spoken word content, so we have plenty of opportunities to see short periods of silence. Sure enough, the carrier wave comes to rest at the center frequency during these periods of silence. But what are all these extra spikes? Is that some sort of interference? You see it on this commercial station too. What is that? I have a confession to make. I disabled some features of the audio processor. Let me turn them back on. There, that's better. So, what are these extra spikes? Howard Armstrong had a trick up his sleeve when proposing a 75 kHz deviation limit. As it turns out, you don't need to go that hard when making sure your deviation is more than the highest sound frequency you plan to reproduce. Humans can, in the best cases, only hear sounds up to around 20 kHz, and that upper limit gradually goes down as you age. This process is called presbycusis, and it's the reason teenagers sometimes like to use notification sounds on their phones that are about 16 to 18 kHz. They can still generally hear sounds of those frequencies, while by the age of 35, you typically can't. With that in mind, it's entirely possible that a maximum deviation of 25 to 30 kHz would have been sufficient, and both broadcasters and consumers alike would be happy with the improvement. Armstrong's 1936 paper also detailed the concept of multiplex operation, which is sometimes abbreviated to MPX. Hey, there's the term that sent me down this rabbit hole in the first place! In multiplex operation, the signal used to modulate the transmitter not only contains the frequencies audible to humans, but additional signals at what he termed super-audible frequencies, that is, higher than humans are capable of hearing. Today, NBC is known as a broadcast television network in the United States, but in the 1930s, NBC had two radio networks, known as the Red Network and the Blue Network. Armstrong noted that an FM multiplex station would be able to transmit the audio programs of the Red Network and Blue Network simultaneously, with the signal for one of them shifted entirely above the range of human hearing before transmission and the reverse upon reception. He also demonstrated a broadcast radio fax service which would allow newspapers to transmit images of their publication directly to consumers over FM radio stations without interrupting the main program. AM stations had run radio fax services before, but they had to completely interrupt regular programming to do so. Let's go back to my setup here and look at the SDR display, but this time I've swapped the RF and demodulated audio sections. The demodulated signal has an FFT of its own showing the spectral intensity of the audio. As the music plays, you can see the intensity of various audio frequencies in this FFT. Watch what happens when I re-enable multiplex. You see a gigantic spike appear in the demodulated signal, but you can't hear any sudden loud noises. What gives? That huge spike is at exactly 19 kilohertz. It's too high a frequency for most humans to hear, but it's definitely there. This is called a pilot tone, and it has a very specific purpose. At the time of Armstrong's 1936 paper and presentation to the FCC, research into modern stereophonic sound recording was in its infancy. Audio engineers knew all about stereo, and there had even been experiments in broadcasting live symphony performances in stereo using two separate AM transmitters, but the lack of an efficient way to record stereophonic sound left it to a novelty until well into the 1950s. The earliest work in modern stereo sound recording in consumer media was done in vinyl records. On mono records, the groove wobbles in a horizontal motion, once again as a direct analog of the sound waves used to produce it. That's not unlike the carrier wave of an FM radio station when you think about it. The horizontal groove was done out of necessity. The earliest Edison phonograph cylinders had a vertically cut groove, modulating the depth of the groove in response to sound. It was found that a flat disc was a far superior form factor, but the weight of the typical stylus assembly used to reproduce the sound would destroy the groove over time, so horizontal motion was the solution. Over time, as electronic amplification technology matured, the weight of the pickup was reduced, the grooves were able to get smaller, and the material these records were made from eventually shifted from shellac to the lighter, softer, and more durable vinyl. The real breakthrough to put a second audio channel on a vinyl record was to reintroduce vertical motion, not only cutting the groove so the sides move in response to sound, but also applying a modulating signal to the depth of the groove. The very first proposed system to cut stereo vinyl records involved putting the left audio channel on the horizontal plane and the right channel on the vertical plane. This never actually made it into production, and there were good reasons for that. Think of how many people had record players at the time, and all of them only had one speaker and the stylus only understood the horizontal motion of the groove. If they played a stereo recording using this system, they would only get the left channel, and if any sound was panned all the way to the right in the stereo image, that sound would not be played. In 1957, Western Electric introduced a new model of their Westrex system for cutting stereo vinyl records. In this system, the planes for the left and right channels were still at right angles to each other, but offset by 45 degrees rather than entirely vertical and horizontal. This was called the 45-45 method, and was the very method described by Alan Blumlein in his first patent on modern stereo all the way back in 1933. The effect with using this system was that thanks to the 45 degree offset, the absolute horizontal motion of a stylus riding this V-shaped groove reflects a mix, or sum, of both audio channels combined. This ensures backward compatibility with older mono equipment. The absolute vertical motion reflects what's called a stereo difference signal. Not useful on its own, but when combined with the sum using specific vector algebra functions, you can extract each individual channel from these signals. Why did I just spend several minutes talking about how stereo is encoded on vinyl records in a video about radio? Well, that's simple. FM radio took a page from vinyl and uses sum and difference, or left plus right and left minus right, channels to broadcast in stereo. In an FM stereo broadcast, the frequency range from 15 hertz to 15 kilohertz is the sum channel. It's functionally equivalent to a mono signal because that's exactly what it is. Both left and right audio channels mix down to mono. Radios that only understand mono broadcasts, or that only have one speaker, will get the entire audio content of the broadcast from this portion of the demodulated signal. Where's the difference channel? Does it have something to do with that 19 kilohertz pilot tone? Let's go back to the SDR, but now I've enabled all features of the audio processor. The pilot tone at 19 kilohertz really stands out, but let me play music again and let's watch what this FFT display looks like. You can see the expected audio between zero and 15 kilohertz like before, but along with the pilot tone, you see lots of activity at an even higher frequency than 19 kilohertz. That's the stereo difference signal. These are audio signals, but you can also think of them sort of like RF signals. The zero through 15 kilohertz sum channel is most like a single sideband AM signal. The intensity of the sound at a given frequency is a direct correlation to the amplitude of the sound wave. An interesting effect of AM is that in the course of the amplitude of the carrier being modulated, the signal modulating it gets sort of splattered on either side of the carrier proportional to the component frequencies of that modulating signal. These are called sidebands. For example, if an AM station were broadcasting a one kilohertz tone, the FFT for the radio signal would show the carrier wave and a concentration of RF power on either side of it at an offset of one kilohertz. This effect would cause an imprint of the audio spectrogram to appear on a waterfall display. This means that in a traditional double sideband AM transmission with a carrier, there are three whole copies of the modulating signal, one in the carrier itself and one in each of the sidebands. Single sideband AM gets rid of the carrier and one of the sidebands, keeping just one copy of the modulating signal and using less spectrum. The disadvantage with single sideband is that you still need a carrier to mix with the sideband to demodulate it, but these demodulators will have a local oscillator to generate one artificially. This artificial carrier needs to be the same frequency as a real one would be to work correctly. If you're into amateur radio, this explains why the sounds you hear from a single sideband transmission are shifted up or down in frequency if your radio isn't tuned just right. Considering what I just told you, take a look at the sum channel again from 0 to 15 kilohertz. The similarity to a single sideband AM transmission should stand out a little more to you. An artificial zero hertz carrier is easy to recreate. Zero hertz is just direct current electricity. Now look at the 19 kilohertz pilot tone. Doesn't that sort of look like an unmodulated FM carrier? The pilot tone serves as an indicator that this is a stereo broadcast. If you've ever seen a radio receiver or boom box that has some sort of light or display showing that it's receiving stereo, the presence of this pilot tone is what controls that. The pilot is protected and no other signals are allowed within four kilohertz of it, so the audio has a very aggressive low pass filter at 15 kilohertz. This effect is easier to see on radio stations that have more aggressive audio processing, like this one. What's all this extra stuff at a higher frequency than the pilot? It seems to have some sort of correlation to the audio present in the sum channel, and this is because that's the stereo difference or left minus right signal. The difference channel is a type of double sideband AM signal with a suppressed carrier or DSBSC. That's why it seems to have a mirror image appearance, but with no spike corresponding to the carrier wave in the center. An interesting development in the 2010s saw some stations experimentally broadcasting the difference channel with single sideband modulation, which proponents claimed results in a better signal to noise ratio and more robust decoding of the stereo separation. Newer receivers handled it well, but older radios had some problems. In the end, it was never fully authorized in the United States, and all FM stereo stations went back to DSBSC. Based on what I've said so far, you may be thinking that your FM radio has a local oscillator to recreate the carrier and demodulate this difference channel. That's a great observation, and you would be right, if it was necessary. This may not be obvious at first, but the carrier for the difference channel is included inside the multiplex signal. It's the pilot tone. The difference channel is centered at 38kHz, which is exactly twice the frequency as the pilot. This is sometimes called the second harmonic. All your radio needs to do is lock on to the 19kHz pilot tone, double its frequency, and use that as the carrier. To get both left and right audio channels from the sum and difference signal, you just need simple algebraic equations. The sum channel can be expressed as the binomial L plus R, and the difference channel is expressed as L minus R. To isolate the left channel, simply add these two binomials. The algebraic equation would then be written as quantity L plus R plus quantity L minus R. By opening the parentheses or brackets to get L plus R plus L minus R and combining like terms to get L plus L plus R minus R, this equation solves to L plus L, or 2L, only the left channel. For the right channel, the expression is quantity L plus R minus quantity L minus R. Opening the parentheses or brackets with the understanding that subtracting a binomial inverts the signs of its terms, you'll get L plus R minus L plus R. Combining like terms yields L minus L plus R plus R, which solves to 2R, or only the right channel. This is how the circuitry in a stereo FM receiver extracts the audio channels. Adding two signals together is pretty easy to imagine, but how do you subtract an electrical signal? For anything with wave oscillations, you'll simply invert the phase of the waveform before combining the signals. This would look like flipping an image of the wave vertically. The peaks of the wave, which would increase the amplitude of the wave it's combined with, are now valleys, which decrease the amplitude of the wave it's combined with. There are a few other interesting properties of FM stereo broadcasting worth mentioning. The pilot tone at 19kHz is typically injected at an approximately 9% modulation level. That, of course, would not cause the FM carrier to deviate from its center frequency very much, which you were clearly able to notice in earlier clips of the waterfall display during periods of This is deliberate, and ironically enough, it makes use of John Carson's 1922 reasoning for dismissing FM. An FM receiver will not try to decode the stereo transmission if it doesn't find the pilot tone. The low modulation level of the pilot ensures that as radio signal reception gets worse, the pilot is one of the first components of the transmission to drop below the noise floor and the receiver falls back to mono mode. Speaking of the noise floor, in FM transmissions, there is more noise at higher frequencies in the demodulated signal. You can sort of see this in the FFT for the audio. The noise floor is visible as this curved line. While it looks curved on this display, the noise profile is more like a diagonal line. Engineers call that a triangular distribution. There is a point at which that noise becomes audible. To compensate for that, FM broadcasters amplify the audio levels more at higher frequencies just before transmission, and radio receivers have a complementary attenuation circuit that brings the audio back to a normal level but also cuts down that noise. This is called pre-emphasis and de-emphasis. Fully grasping how this works requires a deeper understanding of electronics, but you need to be aware it exists. For those squarely on the consumer side, Dolby noise reduction uses a similar method to reduce noise on magnetic tape. There were even experiments with a form of Dolby noise reduction in FM radio broadcasts in the 1970s, but it never really caught on. As I was wrapping up the script for this video, I started reading this book, Empire of the Air, The Men Who Made Radio by Tom Lewis. It's a fantastic account of how broadcasting got its start and follows the stories of the three people who shaped broadcasting the most. Howard Armstrong and David Sarnoff, who I talked about in this video, and Lee DeForest, who invented the triode vacuum tube in 1906. This book has gone straight to the top of my list of reading materials for radio engineers, and you can find this 30th anniversary edition printing at your favorite bookseller. Ken Burns also made a documentary directly based on this book for PBS in 1991, which you can watch with a PBS passport from your local PBS member station if you're in the United States. Elsewhere in the world, it's likely available from a variety of streaming services or digital media stores. I also wanted to thank you for watching this video. I seriously didn't plan on this becoming what it did, but I really enjoyed writing it. Putting this video together helped me learn a few things, and also helped me understand some things I knew about, but didn't know why. This video essay format isn't my forte, but hopefully this comprehensive exploration of how FM works is a resource to broadcasting professionals and radio enthusiasts alike. There's so much more to the story, like the bitter format war that erupted over stereo broadcasting in the late 1950s, the numerous protracted lawsuits that upended the industry and led to Howard Armstrong's tragic and untimely end, and the shift of technical innovation from individual inventors to teams of scientists and engineers employed by giant corporations like RCA. I'll get back to making my usual style of video soon. To support my content, make sure you like, subscribe, and share my videos with your favorite fellow nerds so they can see what broadcasting is all about. You can also visit my landing page at broadcastblueprint.com. I'll catch you next time.