Two Radios in Every Pocket

Your phone contains at least two radios. One connects to your earbuds. The other connects to the internet. Both use electromagnetic waves in roughly the same frequency band. Both transmit digital data through the air. Yet one drains almost no battery and reaches ten meters, while the other burns through power and reaches across a building. They solve fundamentally different problems with different physical strategies, and understanding why they differ reveals beautiful signal processing, information theory, and physics.

This article focuses on the physical layer—what actually happens to electrons, photons, and electromagnetic fields when your device sends a wireless signal. We will largely ignore the upper protocol layers (pairing, authentication, IP routing) and instead ask: how does a 0 or 1 become a radio wave, and how does the receiver extract it back?

Part I: Electromagnetic Waves and the Radio Spectrum

What is a Radio Wave?

A radio wave is an electromagnetic wave—oscillating electric and magnetic fields propagating through space at the speed of light. It is the same phenomenon as visible light, infrared radiation, and X-rays, differing only in frequency.

An oscillating current in an antenna produces a changing electric field, which produces a changing magnetic field, which produces a changing electric field, and so on—self-propagating outward at $c = 3 \times 10^8$ m/s. Maxwell’s equations describe this precisely:

The key relationship between frequency and wavelength:

At 2.4 GHz ($2.4 \times 10^9$ Hz), the wavelength is about 12.5 centimeters—roughly the width of your hand. At 5 GHz, it’s about 6 centimeters. These wavelengths are convenient: small enough for compact antennas (an efficient antenna is roughly $\lambda/4$, which is 3 cm at 2.4 GHz), but large enough to diffract around furniture and pass through walls.

The ISM Band: Shared Spectrum

Both Bluetooth and WiFi operate in the ISM band (Industrial, Scientific, and Medical)—a portion of the radio spectrum that governments worldwide have designated for unlicensed use. The most important ISM band for wireless communication sits at 2.400–2.4835 GHz. This is not a coincidence. It was originally reserved for devices like microwave ovens (which heat food at 2.45 GHz by exciting water molecules), and when engineers needed spectrum for short-range digital communication, they chose this already-unlicensed band.

WiFi additionally uses the 5 GHz band (5.15–5.825 GHz) and, since WiFi 6E, the 6 GHz band (5.925–7.125 GHz). Bluetooth has always lived exclusively in the 2.4 GHz band.

  2.400 GHz                                          2.4835 GHz
  |                                                         |
  |  Bluetooth: 79 channels × 1 MHz                        |
  |  [01][02][03][04]...........................[77][78][79] |
  |                                                         |
  |  WiFi 2.4 GHz: 3 non-overlapping channels × 20 MHz     |
  |  [=====Ch 1=====]    [=====Ch 6=====]    [====Ch 11====]|
  |                                                         |

  5.15 GHz                                           5.825 GHz
  |                                                         |
  |  WiFi 5 GHz: up to 25 non-overlapping 20 MHz channels   |
  |  (or fewer wider channels: 40, 80, 160 MHz)             |
  |  [==20==][==20==][==20==][=====40=====][========80======]|
  |                                                         |

Notice the fundamental difference already visible in this diagram. Bluetooth slices the same 83.5 MHz of spectrum into 79 narrow 1 MHz channels and hops between them. WiFi claims a single wide channel (20, 40, 80, or even 160 MHz) and pours data through it. This is the central architectural divergence, and everything else follows from it.

Part II: How Data Becomes a Radio Wave

Modulation: Encoding Information in a Wave

A pure sine wave at 2.4 GHz carries no information. It is a single tone—like a tuning fork. To transmit data, you must modulate the wave: vary some property of the carrier signal in a pattern that encodes your bits. There are three properties you can vary:

Both Bluetooth and WiFi use phase-based modulation as their primary mechanism, but they do it very differently.

Bluetooth: Gaussian Frequency-Shift Keying (GFSK)

Bluetooth Classic uses one of the simplest digital modulation schemes: GFSK (Gaussian Frequency-Shift Keying). To send a 1, the transmitter shifts the carrier frequency slightly up. To send a 0, it shifts the frequency slightly down. The deviation is $\pm 115$–$175$ kHz from the carrier center.

The “Gaussian” in GFSK refers to the shape of the pulse-shaping filter applied to the data before modulation. A Gaussian filter has the unique property that its impulse response and its frequency response are both Gaussians (the Gaussian is an eigenfunction of the Fourier transform). This makes it the optimal filter for minimizing the product of time duration and bandwidth—it achieves the minimum possible time-bandwidth product, meaning the signal is as compact as possible in both time and frequency simultaneously.

GFSK is simple. The receiver only needs to determine whether the frequency is above or below the carrier center—a binary decision. This simplicity means cheap radios, tiny silicon area, and very low power consumption. The tradeoff: it encodes only 1 bit per symbol. At a symbol rate of 1 Msym/s in a 1 MHz channel, that’s 1 Mbit/s raw—the original Bluetooth Classic data rate.

Enhanced Data Rate: Shifting Phase

Bluetooth 2.0+ introduced Enhanced Data Rate (EDR) using more sophisticated modulation. Instead of GFSK, EDR uses $\pi/4$-DQPSK (2 bits/symbol, 2 Mbit/s) and 8-DPSK (3 bits/symbol, 3 Mbit/s). These are differential phase-shift keying schemes: data is encoded in the change in phase between consecutive symbols rather than in the absolute phase. DQPSK rotates the phase by one of four offsets (each representing 2 bits), and 8-DPSK by one of eight offsets (3 bits).

Higher-order modulation packs more bits per symbol but requires the receiver to make finer distinctions—telling apart 8 phase states instead of 2 frequency states. This needs a better signal-to-noise ratio, which is why EDR is only used when the link quality is good.

WiFi: OFDM — A Symphony of Subcarriers

WiFi uses a fundamentally different and far more sophisticated scheme called OFDM (Orthogonal Frequency-Division Multiplexing). Where Bluetooth uses one narrow channel with one modulated carrier, WiFi divides its wide channel into many narrowband subcarriers and transmits data on all of them simultaneously.

A 20 MHz WiFi channel (802.11a/g/n) contains 64 subcarriers spaced 312.5 kHz apart. Of these, 52 carry data or pilot signals and 12 are unused guard bands at the edges. Each subcarrier is independently modulated using QAM (Quadrature Amplitude Modulation)—the same technique used in cable TV and DSL.

Now for the key insight. Each OFDM subcarrier carries a QAM symbol. With 52 data subcarriers and 64-QAM (6 bits each), a single 20 MHz OFDM symbol carries $52 \times 6 = 312$ bits. The symbol duration is $3.2\;\mu\text{s}$ (plus a $0.8\;\mu\text{s}$ guard interval), so the raw rate is $312 / 4\;\mu\text{s} = 78$ Mbit/s. This is already nearly $80\times$ faster than Bluetooth—and WiFi can use wider channels (40, 80, 160 MHz), higher-order QAM, and multiple spatial streams to go much further.

Why OFDM Works: The Fourier Transform

OFDM is, at its mathematical heart, the Fourier transform in action. Each OFDM symbol is generated by taking the Inverse Fast Fourier Transform (IFFT) of the data symbols. The transmitter does not generate 52 separate sine waves with 52 separate oscillators. Instead, it places the QAM symbols into a frequency-domain vector, computes the IFFT to produce a time-domain waveform, and sends that waveform through a single digital-to-analog converter.

This is a profound connection. The WiFi chip in your phone computes a 64-point or 256-point FFT for every single OFDM symbol it receives—millions of FFTs per second. The FFT, one of the most important algorithms in all of mathematics, is what makes modern wireless communication practical. Without it, OFDM would require dozens of separate oscillators and demodulators; with it, the entire operation is a single digital computation.

The Guard Interval and Multipath

OFDM has another critical advantage: robustness against multipath. In any indoor environment, a transmitted signal bounces off walls, floors, furniture, and people. The receiver gets multiple copies of the signal arriving at different times—the direct path plus delayed reflections. For a narrowband system like Bluetooth, these multipath reflections cause inter-symbol interference (the echoes of one symbol overlap with the next symbol).

OFDM solves this by making each symbol much longer in time. A Bluetooth GFSK symbol lasts $1\;\mu\text{s}$. An OFDM symbol lasts $3.2\;\mu\text{s}$ plus a $0.8\;\mu\text{s}$ guard interval (the cyclic prefix), for a total of $4\;\mu\text{s}$. Indoor multipath delays are typically under $0.5\;\mu\text{s}$, well within the guard interval. The cyclic prefix absorbs the echoes, and the FFT cleanly separates the subcarriers as if there were no multipath at all.

Part III: Bluetooth’s Physical Strategy

Frequency Hopping Spread Spectrum (FHSS)

Bluetooth’s defining physical-layer strategy is frequency hopping. Rather than occupying one channel continuously, a Bluetooth link hops across all 79 channels in a pseudo-random sequence, switching 1,600 times per second (every $625\;\mu\text{s}$).

This is an old idea—it was patented by Hedy Lamarr and George Antheil in 1942 as a way to prevent the jamming of radio-controlled torpedoes. The principle: if the transmitter and receiver agree on a hopping sequence, they can communicate despite interference on any individual channel, because they never stay on an interfered channel for long.

Time →  t=0    t=625μs  t=1250μs  t=1875μs  t=2500μs  t=3125μs
         ___
Ch 23   |___|
                  ___
Ch 71            |___|
                            ___
Ch 05                      |___|
                                        ___
Ch 48                                  |___|
                                                    ___
Ch 12                                              |___|
                                                                ___
Ch 67                                                          |___|

Each hop: 1 MHz wide, 625 μs long, transmits one packet.
Sequence is pseudo-random, determined by master device’s clock and address.

The hopping sequence is generated by a deterministic algorithm seeded by the Bluetooth device address and the master clock. Both endpoints know the seed, so they hop in lockstep. To an eavesdropper or interferer who doesn’t know the sequence, the signal appears as brief, unpredictable bursts scattered across 79 MHz of spectrum—very hard to jam or intercept.

Adaptive Frequency Hopping (AFH)

Bluetooth 1.2 introduced Adaptive Frequency Hopping. The radio monitors which channels have persistent interference (typically from WiFi, which parks on a 20 MHz swath of the same band). Channels with high error rates are removed from the hopping sequence, and the remaining good channels are used more frequently. This is how Bluetooth coexists with WiFi despite sharing the same spectrum: it learns to avoid the frequencies WiFi occupies.

With a 20 MHz WiFi channel occupying roughly 20 of the 79 Bluetooth channels, AFH reduces the available hop channels to about 59—still more than enough for robust communication. This is a direct consequence of the narrow-channel architecture: losing 25% of your channels costs you nothing in bandwidth (each channel is independent), whereas losing 25% of a single wideband OFDM channel would be catastrophic.

Bluetooth Low Energy (BLE): An Even More Minimal Radio

Bluetooth Low Energy (BLE), introduced in Bluetooth 4.0, redesigned the physical layer for extreme power efficiency. BLE uses only 40 channels (2 MHz wide each, vs. Bluetooth Classic’s $79 \times 1$ MHz), with 3 dedicated advertising channels and 37 data channels. It still uses GFSK modulation but with a slightly different set of parameters.

The power savings come from duty cycling. A BLE device can sleep for seconds or minutes, wake up for a few milliseconds to send or receive a packet, and go back to sleep. A typical BLE sensor (heart rate monitor, temperature sensor) transmits a few dozen bytes every second. At 1 Mbit/s, that takes about 0.3 milliseconds. The radio is on for 0.03% of the time. The average current draw can be under 10 microamps.

Part IV: WiFi’s Physical Strategy

Wide Channels and High Throughput

WiFi’s strategy is the opposite of Bluetooth’s. Instead of hopping across narrow channels, WiFi claims a wide, continuous block of spectrum and maximizes the data rate through it. The evolution of WiFi is largely the story of making channels wider and modulation denser:

The throughput gains come from three sources: wider channels (more subcarriers = more parallel data streams), denser modulation (more bits per subcarrier), and MIMO (multiple antennas, discussed below). Each generation pushes all three.

MIMO: Multiple Antennas, Multiple Streams

Starting with WiFi 4, access points and devices use MIMO (Multiple-Input Multiple-Output)—multiple antennas that transmit and receive simultaneously. The key insight from information theory is that in a multipath environment (which is nearly always the case indoors), multiple antennas can create independent spatial channels.

With $N_T$ transmit antennas and $N_R$ receive antennas, the channel can support up to $\min(N_T, N_R)$ independent spatial streams, each carrying the same data rate as a single-antenna system. This multiplies throughput linearly. WiFi 5 supports up to 8 spatial streams; in practice, consumer devices have 2–4 antennas.

This is a remarkable result. Multipath reflection, which is a problem for simple single-antenna systems, becomes an advantage for MIMO. The reflections create independent propagation paths, and MIMO exploits them. The more complex the environment (more walls, furniture, scatterers), the richer the channel matrix H, and the more spatial streams can be supported.

WiFi’s Power Cost

All of this comes at a steep power cost. A WiFi radio must:

• Compute FFTs continuously (64–4096 point, at every symbol)
• Run multiple analog-to-digital converters at high speed (one per antenna, at 20–320 MHz bandwidth)
• Perform channel estimation, equalization, and error correction per subcarrier
• Drive a power amplifier with high linearity (QAM requires precise amplitude control, unlike GFSK)
• Stay awake to listen for incoming packets (even in power-save mode, the radio wakes up for every beacon interval, typically 100 ms)

Typical WiFi transmit power is 15–20 dBm (30–100 mW)—100 times more than Bluetooth. The receive path draws 100–300 mW even when not transmitting. This is why WiFi drains your phone battery far faster than Bluetooth.

Part V: The Physics of Propagation

Free-Space Path Loss

As a radio wave propagates outward from an antenna, its power density decreases with the square of the distance. This is not absorption—it’s geometry. The same power is spread over a sphere of increasing area.

At 5 GHz, the path loss is about 6 dB higher than at 2.4 GHz for the same distance. This is why 5 GHz WiFi has shorter range—not because higher frequencies are absorbed more (they aren’t, significantly, in air), but because the effective antenna aperture is smaller at shorter wavelengths. A 5 GHz antenna captures less of the passing wave than a 2.4 GHz antenna of the same physical size.

Wall Attenuation and Material Absorption

Indoor propagation is dominated by material absorption and reflection. Different materials attenuate 2.4 GHz signals by different amounts:

Water deserves special mention. The 2.45 GHz frequency is near a rotational absorption resonance of water molecules. This is why microwave ovens operate at this frequency—the water in food efficiently absorbs the energy and heats up. In wireless communication, this means that humid environments, rain, and human bodies (which are ~60% water) attenuate 2.4 GHz signals more than you might expect from their physical density alone.

Link Budget: Why Bluetooth Works at 10m and WiFi at 100m

The link budget determines the maximum range. It is an accounting of all the gains and losses from transmitter to receiver:

WiFi has a 29 dB advantage in link margin over Bluetooth—that’s a factor of nearly 800 in power. About 17 dB comes from higher transmit power, and about 12 dB from better receiver sensitivity (a consequence of the wider channel and more sophisticated signal processing). This is why WiFi reaches across buildings while Bluetooth is room-scale.

Part VI: Shannon’s Theorem and the Fundamental Tradeoffs

All of the design choices in Bluetooth and WiFi are ultimately constrained by one equation—Shannon’s channel capacity theorem, the most important result in information theory:

This single equation explains everything about the Bluetooth/WiFi tradeoff:

Shannon’s theorem is not a guide or approximation—it is a hard mathematical bound. You can approach it with better coding (LDPC, turbo codes), better modulation (higher QAM), and better signal processing (MIMO, beamforming). You cannot exceed it. Every generation of WiFi gets closer to the Shannon limit, and the improvements are getting harder.

Part VII: The Complete Comparison

Part VIII: The Intuition

Here is the mental model that ties it all together.

Bluetooth is a whisper. It speaks softly (low power), uses a narrow voice (1 MHz), and constantly changes the frequency it whispers at (frequency hopping) so that even in a noisy room, the listener can always catch a few clean words. It doesn’t need to say much—a few kilobytes of audio codec data, a heart rate number, a keyboard scancode—so a whisper is enough. The radio is tiny, cheap, and sips power. It trades throughput for simplicity and efficiency.

WiFi is a firehose. It shouts (high power) across a wide bandwidth (20–160 MHz), packs information into every dimension available (many subcarriers $\times$ dense QAM $\times$ multiple spatial streams via MIMO), and uses the most sophisticated signal processing available (FFTs, LDPC, beamforming) to extract every last bit. It needs to carry video streams, file downloads, and web traffic—megabytes and gigabytes. The radio is power-hungry, computationally intensive, and requires precision analog circuits. It trades power and complexity for throughput.

Both are electromagnetic waves at 2.4 GHz. Both are governed by Shannon’s theorem. Both are limited by path loss and multipath. The difference is entirely in how they allocate the fundamental resources: bandwidth, power, and complexity.

The Deeper Physics

At the deepest level, wireless communication is constrained by thermodynamics. The noise floor at the receiver is set by thermal noise: the random motion of electrons at room temperature. This noise power is:

This is a beautiful convergence of physics, mathematics, and engineering. Maxwell’s equations determine how the wave propagates. Boltzmann’s constant determines the noise floor. Shannon’s theorem determines the maximum data rate given the signal-to-noise ratio. Fourier analysis provides the computational tool (the FFT) that makes OFDM practical. And the engineers who designed Bluetooth and WiFi made different, complementary choices about where to sit in this space of physical constraints.

Every time you connect your earbuds and stream a song while browsing the web, two radios in your pocket are independently solving the wave equation, sampling at millions of times per second, computing Fourier transforms in real time, and extracting information from thermal noise—all while consuming less power than a dim LED. The physics is the same. The math is the same. The engineering is exquisitely different.