24, 60, or 77 GHz? How to Choose a Radar for Human Sensing

Choosing a human-sensing radar is not a contest in which the highest carrier frequency automatically wins. For most new indoor projects involving human activity recognition, gesture sensing, occupancy, tracking, or contactless vital-motion research, a programmable 60 GHz platform is the strongest default. It usually combines several gigahertz of usable bandwidth, compact antenna arrays, good sensitivity to small movements, and hardware and software designed for indoor sensing.

Choose 24 GHz when the task is simpler—such as motion detection, velocity measurement, or coarse single-target ranging—and cost, propagation margin, or an established 24 GHz design matters more than fine range or angular detail. Choose 77 GHz when its automotive-grade ecosystem, smaller wavelength, or an existing compliant platform provides a concrete advantage. For a generic indoor experiment, however, 77 GHz can create a regulatory problem without providing better range resolution than a well-configured 60 GHz sensor.

Short answer: Start from the measurement you need, then check usable bandwidth, antenna aperture, raw-data access, timing, link budget, and regional authorization. Carrier frequency comes after those questions—not before them.

The decision at a glance

This is a starting matrix, not a substitute for a link budget or a regulatory review.

Strictly, 24 GHz lies below the conventional 30–300 GHz millimetre-wave definition, although vendors and application literature often discuss 24, 60, and 77 GHz sensing platforms together.

The most important correction: frequency is not range resolution

For an ideal FMCW radar, the approximate ability to separate two targets along the range axis is

\Delta R \approx \frac{c}{2B},

where c is the propagation speed and B is the swept bandwidth actually used by the chirp.

Carrier frequency f_c does not appear in this expression. A 60 GHz radar sweeping 4 GHz and a 77 GHz radar sweeping 4 GHz have the same ideal range resolution:

\Delta R = \frac{299{,}792{,}458}{2(4\times10^9)} \approx 3.75\ \text{cm}.

By contrast, a 24 GHz system restricted to a 250 MHz sweep has

\Delta R \approx 0.60\ \text{m}.

That does not mean the 24 GHz radar cannot detect millimetre-scale motion or estimate one target’s range more precisely than 60 cm. Range resolution, range-bin spacing, range accuracy, and displacement sensitivity are different quantities. Texas Instruments makes the same distinction between resolving two nearby objects and estimating the position of one object in its range and angular resolution guide.

Windowing, signal-to-noise ratio, calibration, leakage, target extent, and the estimator also affect practical resolution. The formula is a design baseline, not a promised field result.

Representative numbers—not universal band specifications

The bandwidth row deliberately uses representative implementations. For example, TI specifies 4 GHz continuous bandwidth for the 60–64 GHz IWR6843, 7 GHz for the 57–64 GHz IWRL6432, and 4 GHz for the 76–81 GHz IWR1443. Hardware tuning range is not the same as the bandwidth legally permitted in every country or application.

What the carrier frequency actually changes

1. Doppler sensitivity

For a monostatic radar observing radial velocity v,

f_D=\frac{2v}{\lambda}=\frac{2vf_c}{c}.

The same 1 m/s radial motion produces about 160 Hz at 24 GHz, 400 Hz at 60 GHz, and 514 Hz at 77 GHz. A higher carrier therefore spreads a given set of body-part velocities over a wider Doppler-frequency interval. That can benefit micro-Doppler analysis when the slow-time sampling, coherent processing interval, signal-to-noise ratio, and oscillator stability support it.

To see how torso, arm, leg and foot velocities turn into time-varying Doppler structure, read our physics-first guide to human micro-Doppler.

Notice the scale of the difference: moving from 24 to 60 GHz multiplies the Doppler shift by 2.5, but moving from 60 to 77 GHz adds only about 28%. That incremental gain alone rarely justifies changing an otherwise suitable 60 GHz platform.

2. Phase sensitivity to small displacement

A radial displacement x changes the round-trip phase by approximately

\Delta\phi=\frac{4\pi x}{\lambda}.

For a 1 mm displacement, the ideal phase changes are approximately 58°, 144°, and 185° at 24, 60, and 77 GHz. This is why microwave and millimetre-wave devices are attractive for breathing, cardiopulmonary-motion, tremor, and small-gesture research.

Greater phase sensitivity is useful only if coherence is preserved. Phase noise, sampling jitter, target movement, multipath, I/Q imbalance, phase wrapping, and changes in the dominant scattering point can overwhelm the desired displacement. A higher carrier does not remove those problems; in some cases, it makes phase tracking more demanding.

Google’s 60 GHz Soli research demonstrates that compact millimetre-wave hardware can support both fine gesture sensing and non-contact heart-rate estimation. Those results reflect complete sensing systems and validated algorithms, not frequency alone.

For the complete experimental context behind the gesture result—including its controlled evaluation and same-participant limitation—read our breakdown of what Google Soli actually proved.

3. Antenna size and angular aperture

Shorter wavelengths allow smaller antenna elements and closer element spacing. Approximate half-wavelength spacing is 6.25 mm at 24 GHz, 2.50 mm at 60 GHz, and 1.95 mm at 77 GHz. More elements can therefore fit within a fixed physical width at the higher frequencies.

The important variable for angle discrimination is the effective aperture, array geometry, and estimator. For a simplified uniformly illuminated aperture of width D, beamwidth scales roughly with

\theta \propto \frac{\lambda}{D}.

But if two arrays have the same number of elements and both use λ/2 spacing, their physical widths shrink with wavelength and their conventional beamwidths can be similar. Higher frequency improves angular capability mainly when the design uses the shorter wavelength to fit a larger electrical aperture or more virtual channels into the available space.

This is why “3 transmitters and 4 receivers” is still incomplete information. The physical and virtual antenna positions, simultaneous or time-multiplexed operation, azimuth/elevation layout, field of view, mutual coupling, calibration, and beamforming method all matter.

4. Propagation and link budget

For a simplified monostatic radar,

P_r=\frac{P_tG_tG_r\lambda^2\sigma}{(4\pi)^3R^4L},

where P_t is transmitted power, G_t and G_r are antenna gains, σ is radar cross section, R is range, and L represents losses.

If antenna gains and target radar cross section were held constant, the λ² term would favour the lower carrier. Real hardware does not hold those quantities constant: a fixed physical aperture can provide more gain at shorter wavelength, permitted EIRP differs by band and use, target scattering is frequency-dependent, and processing gain varies with the waveform. Maximum range must therefore be calculated from the complete sensor, target, waveform, and environment—not inferred from 24, 60, or 77 GHz alone.

The oxygen absorption feature near 60 GHz is real and is included in the ITU-R P.676 atmospheric model. At normal indoor distances, however, furniture, people, wall materials, antenna pattern, and multipath are usually more consequential. NIST measurements at 60.5 GHz found substantial, material- and angle-dependent losses through doors, plasterboard, and glass in an office environment; the reported ranges were from several decibels to tens of decibels depending on the path and construction (NIST study).

The practical conclusion is not that 60 GHz “has no range.” It is that a line-of-sight room-scale sensor and a through-wall sensor are different engineering problems.

When 24 GHz is the right choice

24 GHz remains useful because many human-sensing tasks do not require a dense point cloud or centimetre-scale range separation.

Strengths

• A longer wavelength can offer a useful propagation and tolerance margin in some cluttered or partially obstructed settings.
• Simple CW and FMCW front ends can support motion, velocity, direction-of-motion, and coarse range measurements.
• Mature modules and discrete front ends are available for researchers who need access to analogue I/Q rather than a heavily processed point cloud.
• The lower frequency makes RF layout and antenna fabrication somewhat less unforgiving than at 60 or 77 GHz.
• For a basic trigger or single-target tracker, the simpler sensor may produce a better overall system than an unnecessarily complex MIMO device.

Infineon’s current DEMO DISTANCE2GOL, for example, uses a 24 GHz BGT24LTR11 front end in software-controlled FMCW mode for human range and one-dimensional tracking. Its 1-TX/1-RX architecture also illustrates the trade-off: it is approachable for range and velocity experiments but is not a substitute for a multi-channel two-dimensional or three-dimensional imaging array.

Limitations

• In the United States, the commonly used unlicensed band under 47 CFR §15.249 is 24.0–24.25 GHz. A full 250 MHz FMCW sweep corresponds to about 60 cm ideal range resolution.
• Antenna arrays occupy more physical area for the same number of half-wavelength-spaced elements.
• Low-cost modules often expose only motion flags, speed, or a proprietary target list.
• A legacy GUI, limited SDK, or discontinued capture path can cost more research time than the hardware saves.

Best fit

Choose 24 GHz when you can state all three of the following:

1. The task does not require closely separated range bins or a dense spatial point cloud.
2. The chosen module exposes the signal or measurement needed by your algorithm.
3. A concrete advantage—cost, range, analogue access, existing design experience, or certification—outweighs the benefits of a newer 60 GHz platform.

Why 60 GHz is usually the best default for indoor human sensing

60 GHz occupies a productive middle ground: it offers much smaller antennas and greater motion sensitivity than 24 GHz, while current devices can provide range resolution comparable to or better than many 77 GHz development platforms.

Strengths

• Current sensors offer several gigahertz of programmable FMCW bandwidth. TI specifies 7 GHz continuous bandwidth, three receivers, and two transmitters for the IWRL6432; the IWR6843 provides four receivers, three transmitters, and 4 GHz bandwidth.
• Compact antenna-on-package and antenna-on-PCB options support wall, ceiling, bedside, appliance, and desktop placements.
• Phase and Doppler sensitivity are strong enough for small gestures and periodic micro-motion.
• The ecosystem includes examples for occupancy, people counting, tracking, gestures, and vital-motion sensing.
• In the United States, operation from 57 to 71 GHz is addressed by 47 CFR §15.255, including specific provisions for field-disturbance sensors. The allowed power and duty-cycle conditions depend on the operating sub-band and use.

Limitations

• Wide bandwidth can create a large radar data cube and demanding capture, storage, and processing requirements.
• Integrated devices differ substantially in on-chip memory, DSP/accelerator support, raw-streaming interfaces, and antenna layout.
• Blockage and material loss can be severe; “works through clothing” should never be generalized into “works through walls.”
• Close-range sensing can be limited by TX–RX leakage, ramp settling, IF filtering, antenna coupling, and the device’s configured minimum range.
• A compact evaluation board may have a broad field of view but modest conventional angular resolution.

Best fit

For a new university or R&D project on indoor HAR, gestures, occupancy, fall detection, in-cabin sensing, or contactless vital motion, evaluate 60 GHz first. Move away from it only when a measured requirement points elsewhere.

When 77 GHz is justified

The 76–81 GHz ecosystem was shaped strongly by automotive radar. It can also be technically attractive for industrial sensing, fine ranging, displacement measurement, and high-channel-count MIMO research.

Strengths

• The smallest wavelength of the three bands gives the largest Doppler and phase response for the same radial motion.
• Automotive platforms can offer sophisticated chirp scheduling, interference handling, calibration, packaging, and array designs.
• Broad FMCW sweeps support centimetre-scale ideal range resolution.
• A fixed physical aperture can accommodate more antenna elements than at 24 GHz.

The TI IWR1443, for example, covers 76–81 GHz, provides 3 TX and 4 RX channels, and supports 4 GHz continuous bandwidth. Its ideal bandwidth-limited range resolution is therefore the same 3.75 cm as a 4 GHz 60 GHz device. The extra carrier frequency does not improve that number.

The regulatory trap

In the United States, current 47 CFR §95.3331 permits 76–81 GHz systems as vehicular radars or as fixed/mobile radars in airport air-operations areas. That is not a general authorization for indoor university, office, or home human-sensing experiments. Europe also has application-specific standards; ETSI EN 302 264 addresses short-range radar in 77–81 GHz for transport and traffic telematics.

A development board’s availability, tuning range, or “industrial” description does not establish that a particular transmission is permitted. Check the finished device, antenna, EIRP, occupied bandwidth, duty cycle, location, and use with the relevant authority or qualified test laboratory. The same caution applies to 24 and 60 GHz; their rules are simply more aligned with many common short-range indoor sensing cases.

Best fit

Choose 77 GHz when at least one of these is true:

• the application is vehicular and already has a compliant 77 GHz path;
• your laboratory has the authorization and facilities needed for the planned transmissions;
• you need compatibility with an existing 77 GHz dataset, array, radar front end, or production platform;
• a specific device-level feature—not merely the higher number—beats the available 60 GHz alternatives.

What about lower-frequency and UWB radars?

The choice is not limited to 24, 60, and 77 GHz. Radars operating around 3–10 GHz can be attractive when penetration through common building materials, low-cost motion detection, or ultra-wideband ranging matters more than compact antennas and high micro-motion sensitivity.

However, carrier frequency alone is not enough to classify these systems. A 5.8 GHz continuous-wave motion detector and a multi-gigahertz UWB radar may operate in a similar frequency region while offering completely different measurements.

A low carrier frequency does not necessarily mean poor range resolution

Carrier frequency and bandwidth play different roles. The wavelength is determined by the carrier:

\lambda=\frac{c}{f_c}.

For an FMCW or UWB radar, ideal range resolution is primarily determined by usable bandwidth:

\Delta R=\frac{c}{2B}.

A narrowband radar sweeping only 50 MHz has an ideal range resolution of approximately 3 m, regardless of whether its carrier is 3, 10, or 60 GHz. A UWB system occupying 2 GHz of bandwidth can theoretically separate reflectors approximately 7.5 cm apart.

This is why a low-GHz UWB radar can provide much better ranging than a narrowband 10 GHz motion detector. It achieves that resolution through bandwidth, not because of its nominal center frequency.

MIT’s Vital-Radio research is a useful example: it used wideband FMCW measurements to separate reflections by range and monitor breathing and heart motion, including demonstrations involving walls. That result should be understood as the capability of a carefully designed wideband system—not of every low-frequency radar module. (Vital-Radio paper)

Why lower frequencies can help around obstructions

At equal antenna gains and distance, lower-frequency signals have lower free-space path loss. They may also experience less attenuation through certain walls, furniture, clothing, and non-metallic covers.

That does not make “lower frequency penetrates better” a universal rule. Performance depends on:

• Material composition and moisture
• Wall thickness and internal reinforcement
• Incidence angle and polarization
• Antenna placement
• Occupied bandwidth
• Transmit-power limits
• Receiver sensitivity and clutter suppression

Metal surfaces remain strong blockers or reflectors across these bands. Reinforced concrete, foil insulation, coated glass, appliances, and plumbing can also create severe attenuation or multipath. Any through-wall requirement should therefore be tested using the actual wall construction and sensor installation.

The cost of using a longer wavelength

Lower-frequency radars sacrifice some sensitivity to small displacement. For a monostatic radar, the round-trip phase change caused by displacement x is:

\Delta\phi=\frac{4\pi x}{\lambda}.

Approximate values for a 1 mm displacement are:

A lower-frequency radar can still measure breathing and other periodic motion, especially with coherent phase processing. However, the same physical displacement produces less phase change, while human micro-Doppler features are compressed into a narrower frequency region.

Array size is another consideration. Antenna spacing is commonly kept near half a wavelength to avoid spatial ambiguity. Half-wavelength spacing is about 47 mm at 3.18 GHz, 26 mm at 5.8 GHz, 14 mm at 10.525 GHz, and only 2.5 mm at 60 GHz. Consequently, achieving useful angular resolution at low frequencies can require a much larger sensor.

Be careful with the stated frequency

“3.18 GHz,” “5.6 GHz,” and “10 GHz” are not always the exact frequencies at which a commercial radar is authorized to transmit.

In the United States, the FCC field-disturbance-sensor bands include 5.785–5.815 GHz and 10.500–10.550 GHz. This is why practical modules are commonly described as 5.8 GHz and 10.525 GHz, rather than exactly 5.6 or 10.0 GHz. (FCC 47 CFR §15.245)

Similarly, a device described as operating at 3.18 GHz is not automatically a compliant UWB radar. U.S. indoor UWB rules cover systems whose UWB bandwidth is contained within 3.1–10.6 GHz and impose low spectral-density limits and operational restrictions. Compliance depends on the complete emitted spectrum and equipment authorization—not one frequency printed on a product page. (FCC 47 CFR §15.517)

Other countries use different allocations and power limits. Regulatory approval should therefore be checked for the intended market before selecting a module.

When should you choose one?

Choose a lower-frequency CW radar around 5.8 or 10.525 GHz when the requirement is inexpensive, low-data-rate detection of motion, approach, departure, or possibly basic respiratory movement. Confirm whether the module exposes coherent I/Q data; a digital motion output alone leaves little room for advanced processing.

Consider low-GHz UWB when fine range separation or operation around obstructions is central to the research question and a larger antenna, custom signal processing, and regulatory constraints are acceptable.

Choose 24, 60, or 77 GHz FMCW when the system needs compact arrays, range–Doppler maps, angle estimation, gesture recognition, multiple-person separation, or detailed micro-motion analysis.

The practical conclusion is that lower-frequency radar is not an inferior version of millimetre-wave radar. It is a different engineering trade-off. Judge it by waveform, bandwidth, antenna aperture, exposed data, regulatory approval, and the materials in the deployment environment—not by carrier frequency alone.

Seven specifications that matter more than the number on the box

1. Usable bandwidth

Record the chirp start frequency, stop frequency, linear usable sweep, ramp time, and restrictions imposed by the device and local rules. Do not substitute the front end’s tuning range for the bandwidth your configuration can actually use.

2. Raw complex-data access

For algorithm research, determine whether the platform exposes raw ADC samples for every receive channel, range profiles, radar cubes, point clouds, tracks, or only application decisions.

A board that outputs polished point clouds may be excellent for rapid prototyping. It is a poor choice if the research question concerns clutter suppression, CFAR, weak limb returns, phase processing, beamforming, or learned front ends. Our radar-based human activity recognition guide explains how each representation discards different information.

3. Antenna geometry—not only channel count

Ask for the physical TX/RX coordinates and the resulting virtual array. Check whether the array measures azimuth, elevation, or both; which channels can operate simultaneously; the unambiguous field of view; sidelobes; polarization; and calibration support.

4. Chirp and frame timing

Maximum unambiguous velocity, Doppler resolution, coherent processing interval, frame rate, and data throughput are coupled. A configuration for slow breathing may alias a fast hand movement. A high-frame-rate gesture configuration may not observe enough cycles for stable respiratory-frequency estimation.

If micro-Doppler is the target representation, the slow-time sampling and window duration must support the desired time–frequency trade-off. See the ArthaVedya STFT and radar micro-Doppler guide before locking the chirp schedule.

5. Minimum and maximum practical range

Ignore an unqualified “up to 20 m” claim. Ask:

• For what target radar cross section?
• At what detection probability and false-alarm rate?
• With which antenna, chirp, integration time, and mounting height?
• In an empty chamber or a furnished room?
• Does the application require detecting a person, resolving two people, or classifying their activities?

The maximum range for torso detection is not necessarily the maximum range for reliable limb micro-Doppler or vital-motion extraction.

6. Processing and capture architecture

Check on-chip CPU/DSP/accelerator resources, RAM, host interfaces, raw-data throughput, capture-board requirements, supported operating systems, and whether examples expose source code. A $100 sensor that requires an additional capture card, proprietary tool, and older operating system is not a $100 research setup.

7. Regulatory, RF-exposure, and product status

Verify the exact orderable board, antenna, region, and intended use. FCC §15.255 explicitly subjects 57–71 GHz devices to RF-exposure requirements, and certification guidance stresses that developers remain responsible for the finished equipment. TI’s regulatory compliance guide is a useful engineering introduction, but the current rule and competent regulatory advice take precedence.

A practical hardware-selection workflow

Step 1: Write the observable before naming a sensor

Use a statement such as:

Detect and classify walking, sitting, standing, and falling for one person at 1–5 m, over ±45° azimuth, with subject-independent evaluation and access to raw multi-channel complex samples.

This is actionable. “Build a 77 GHz AI radar” is not.

Step 2: Convert the task into minimum requirements

Define:

• range interval and minimum range separation;
• maximum radial speed and required velocity resolution;
• field of view and angular separation;
• smallest motion of interest;
• number of people;
• required frame rate and continuous recording duration;
• indoor, outdoor, vehicle, or behind-material geometry;
• raw data, maps, points, tracks, or edge decisions;
• power, enclosure, and compute limits.

Step 3: Derive—not guess—the first waveform values

For example, a desired 5 cm ideal range resolution requires approximately

B\geq\frac{c}{2\Delta R}=\frac{299{,}792{,}458}{2(0.05)}\approx3.0\ \text{GHz}.

That immediately rules out a conventional 250 MHz 24 GHz sweep, regardless of classifier quality.

Step 4: Check authorization before purchase

Confirm the band and use in every location where the system will transmit. University research status does not automatically waive spectrum rules. If compliance is unclear, resolve it before building a dataset around the hardware.

Step 5: Shortlist complete platforms

Compare evaluation boards, not isolated RF chips, unless your team is prepared to design millimetre-wave RF hardware. Include antennas, capture hardware, cables, software versions, licences, storage, and host compute in the comparison.

Step 6: Run a proof-of-measurement test

Before collecting a large dataset, test whether the platform can retain the required information:

1. record an empty scene;
2. record one stationary person;
3. record the smallest and fastest target motions;
4. repeat at the nearest, farthest, and most oblique positions;
5. inspect raw channels and the intended representation;
6. verify dropped frames, phase continuity, saturation, and metadata;
7. repeat on another day and after remounting the sensor.

Only then should the project commit to large-scale acquisition.

Three example decisions

Example A: A room-occupancy sensor

The requirement is presence and coarse location within 6 m, including nearly stationary occupants. A low-power 60 GHz module is a strong first choice because fine phase changes, compact arrays, and existing occupancy pipelines are useful. A 24 GHz module can still win if binary presence is sufficient and it offers lower system cost or better coverage. A 77 GHz platform adds little unless already integrated into a compliant product.

Example B: A publishable activity-recognition dataset

The project needs walking, sitting, falling, and bending across people and viewpoints. Select a 60 GHz board with multi-channel raw complex capture, at least roughly 3–4 GHz usable bandwidth, programmable chirps, stable timestamps, and documented antenna geometry. Do not choose a cheap “AI presence” module that exposes only labels. The ability to revisit range–Doppler, range–angle, phase, and CFAR processing is worth more than a longer advertised detection range.

Example C: Fine hand gestures at a desk

The operating range is 0.15–1 m and movements are rapid and small. Start with 60 GHz and prioritize minimum range, leakage behaviour, high frame rate, phase stability, and a suitable wide-field antenna. A 77 GHz device offers only about 28% more Doppler/phase scaling than 60 GHz and may impose a harder regulatory path. A 24 GHz system can sense motion, but its larger antennas and commonly narrower sweep are less attractive for a compact, spatially detailed interface.

Common purchasing mistakes

1. Buying frequency instead of bandwidth. A 77 GHz label does not promise finer range resolution than 60 GHz.
2. Confusing resolution with accuracy. A sensor may estimate one strong target accurately without resolving two nearby scatterers.
3. Counting antennas without inspecting their coordinates. A 3-TX/4-RX device does not guarantee a useful 12-element rectangular virtual array.
4. Choosing processed output for a raw-signal research question. Lost phase and sub-threshold energy cannot be recovered later.
5. Using maximum range as the primary specification. Human classification range is usually shorter and more condition-dependent than detection range.
6. Assuming higher frequency always means worse—or better—propagation. Antenna gain, EIRP, target scattering, noise figure, processing gain, and environment must be included.
7. Ignoring minimum range. Close gestures and bedside sensing can fail because of coupling, leakage, and IF limits.
8. Ignoring spectrum rules until deployment. The intended application can determine whether a band is permitted at all.
9. Treating a research vital-sign estimate as a medical device. Clinical claims require appropriate reference measurements, populations, protocols, and regulatory validation.
10. Buying before testing the data path. SDK restrictions, capture-board dependencies, and dropped samples can define the project more than RF performance.

Final recommendation

For a new indoor human-sensing project, begin by evaluating a 60 GHz programmable development platform with documented antenna geometry and access to the lowest-level data your research needs. It is usually the best-balanced option for range–Doppler processing, compact arrays, gestures, HAR, occupancy, tracking, and micro-motion.

Choose 24 GHz deliberately for simpler motion or coarse-ranging systems, established analogue access, or a demonstrated link-budget and cost advantage. Choose 77 GHz deliberately for compliant vehicular or specialized platforms, high-channel-count automotive hardware, or compatibility with an existing research system—not because 77 is numerically greater than 60.

The defensible question is therefore not “Which GHz is best?” It is:

Which legal, measurable, and supportable platform preserves the information needed to answer this specific human-sensing question?

Key takeaways

• FMCW range resolution is set primarily by swept bandwidth, not carrier frequency.
• Higher carrier frequency increases Doppler and displacement-phase sensitivity and reduces antenna scale.
• Angular performance depends on electrical aperture, array geometry, calibration, and processing—not frequency or channel count alone.
• 60 GHz is the strongest general default for new indoor human-sensing research.
• 24 GHz remains valuable for simpler and potentially longer-margin sensing tasks.
• 77 GHz is technically capable but can be application-restricted; verify authorization before purchase or transmission.
• Raw-data access, chirp timing, antenna geometry, minimum range, software support, and regulatory compliance can matter more than the headline band.

MATLAB code for Figure 1

The accompanying script plot_radar_frequency_tradeoffs.m creates the four-panel comparison at publication resolution and writes the numerical table used in this article. It requires only base MATLAB.

% plot_radar_frequency_tradeoffs.m
% Original ArthaVedya figure: 24, 60 and 77 GHz human-sensing trade-offs.
% Base MATLAB only. The calculations are ideal theoretical comparisons.

clear; close all; clc;

c = 299792458;                    % propagation speed (m/s)
fcGHz = [24 60 77];
fc = fcGHz * 1e9;
lambda = c ./ fc;
labels = {'24 GHz','60 GHz','77 GHz'};
colors = [31 78 104; 26 145 140; 222 145 47] / 255;

% Derived reference values
wavelength_mm = lambda * 1e3;
half_spacing_mm = wavelength_mm / 2;
doppler_Hz_per_mps = 2 ./ lambda;
phase_deg_per_mm = (4*pi*1e-3 ./ lambda) * 180/pi;

T = table(fcGHz(:), wavelength_mm(:), half_spacing_mm(:), ...
    doppler_Hz_per_mps(:), phase_deg_per_mm(:), ...
    'VariableNames', {'Carrier_GHz','Wavelength_mm','HalfLambda_mm', ...
    'Doppler_Hz_per_mps','Phase_deg_per_mm'});
disp(T);
writetable(T,'radar_frequency_reference_values.csv');

fig = figure('Color','w','Position',[100 100 1500 950]);
tl = tiledlayout(fig,2,2,'TileSpacing','compact','Padding','compact');
% Reserve only a narrow footer: enough to separate the note from the
% lower x-axis labels without creating excessive blank space.
tl.OuterPosition = [0.01 0.035 0.98 0.945];
title(tl,'What Changes Between 24, 60 and 77 GHz?', ...
    'FontWeight','bold','FontSize',20);

% (a) Range resolution: independent of carrier frequency
nexttile;
B_GHz = logspace(log10(0.1),log10(8),500);
dR_cm = c ./ (2*B_GHz*1e9) * 100;
semilogx(B_GHz,dR_cm,'Color',[0.18 0.22 0.27], ...
    'LineWidth',2.5); hold on;
Bmark = [0.25 4 7];
dRmark = c ./ (2*Bmark*1e9) * 100;
rangeMarkerColor = [0.18 0.48 0.55];
scatter(Bmark,dRmark,85,repmat(rangeMarkerColor,numel(Bmark),1), ...
    'filled','MarkerEdgeColor','w', ...
    'LineWidth',1.0);

% Use deliberately staggered annotation positions. The 4 and 7 GHz
% markers are close in both x and y, so automatic proportional offsets
% make their labels overlap near the lower-right corner.
labelX = [0.27, 3.80, 7.55];
labelY = [65.0, 15.0, 6.5];
labelAlignment = {'left','right','right'};
for k = 1:numel(Bmark)
    plot([Bmark(k),labelX(k)],[dRmark(k),labelY(k)],'-', ...
        'Color',rangeMarkerColor,'LineWidth',0.8,'HandleVisibility','off');
    text(labelX(k),labelY(k), ...
        sprintf('%.2g GHz: %.1f cm',Bmark(k),dRmark(k)), ...
        'Color',rangeMarkerColor,'FontWeight','bold','FontSize',10, ...
        'HorizontalAlignment',labelAlignment{k}, ...
        'VerticalAlignment','bottom','Clipping','on');
end
xlabel('Usable FMCW sweep bandwidth (GHz)');
ylabel('Ideal range resolution (cm)');
title('(a) Bandwidth controls range resolution');
grid on; box on; xlim([0.1 8]); ylim([0 155]);

% (b) Monostatic Doppler versus radial velocity
nexttile;
v = linspace(0,3,300);
for k = 1:numel(fc)
    plot(v,2*v/lambda(k),'LineWidth',2.4,'Color',colors(k,:));
    hold on;
end
xlabel('Radial speed (m/s)');
ylabel('Doppler frequency (Hz)');
title('(b) Higher carrier gives larger Doppler shift');
legend(labels,'Location','northwest','Box','off');
grid on; box on; xlim([0 3]);

% (c) Unwrapped round-trip phase versus displacement
nexttile;
x_mm = linspace(0,5,300);
for k = 1:numel(fc)
    phase_deg = (4*pi*(x_mm*1e-3)/lambda(k))*180/pi;
    plot(x_mm,phase_deg,'LineWidth',2.4,'Color',colors(k,:));
    hold on;
end
xlabel('Radial displacement (mm)');
ylabel('Unwrapped round-trip phase change (degrees)');
title('(c) Higher carrier gives stronger phase response');
legend(labels,'Location','northwest','Box','off');
grid on; box on; xlim([0 5]);

% (d) Physical wavelength and nominal half-lambda spacing
nexttile;
b = bar(fcGHz,[wavelength_mm(:),half_spacing_mm(:)], ...
    'grouped','LineStyle','none');
b(1).FaceColor = [0.20 0.43 0.55];
b(2).FaceColor = [0.74 0.80 0.83];
xlabel('Carrier frequency (GHz)');
ylabel('Distance (mm)');
title('(d) Antenna scale shrinks with wavelength');
legend({'Wavelength','Half-wavelength spacing'}, ...
    'Location','northeast','Box','off');
grid on; box on;

annotation(fig,'textbox',[0.08 0.002 0.84 0.022], ...
    'String',['Ideal comparisons only. Practical performance also depends on ' ...
    'usable bandwidth, antenna geometry, SNR, calibration, waveform and regulation.'], ...
    'EdgeColor','none','HorizontalAlignment','center', ...
    'VerticalAlignment','middle','FontAngle','italic','FontSize',9, ...
    'Color',[0.30 0.34 0.38]);

% Publication outputs
exportgraphics(fig,'radar_frequency_tradeoffs.png','Resolution',300);
exportgraphics(fig,'radar_frequency_tradeoffs.pdf','ContentType','vector');
savefig(fig,'radar_frequency_tradeoffs.fig');

References

1. Federal Communications Commission, 47 CFR §15.249: Operation in the 24.0–24.25 GHz band.
2. Federal Communications Commission, 47 CFR §15.255: Operation within 57–71 GHz.
3. Federal Communications Commission, 47 CFR §95.3331: Permissible 76–81 GHz radar uses.
4. ETSI, EN 302 264 V2.1.1: Short-range radar equipment in 77–81 GHz, 2017.
5. Texas Instruments, IWRL6432 57–64 GHz industrial radar sensor.
6. Texas Instruments, IWR6843 60–64 GHz intelligent mmWave sensor.
7. Texas Instruments, IWR1443 76–81 GHz mmWave sensor.
8. Texas Instruments, Understanding Range and Angular Resolution in mmWave Radar Devices, revised January 2026.
9. Texas Instruments, mmWave Radar Device Regulatory Compliance Guide, revised December 2021.
10. Infineon Technologies, DEMO DISTANCE2GOL 24 GHz FMCW evaluation board.
11. Infineon Technologies, BGT60TR13C 60 GHz radar sensor datasheet.
12. International Telecommunication Union, Recommendation ITU-R P.676-13: Attenuation by atmospheric gases and related effects, 2022.
13. S. Y. Jun et al., “Penetration Loss at 60 GHz for Indoor-to-Indoor and Outdoor-to-Indoor Mobile Scenarios”, 2020.
14. J. Lien et al., “Soli: Ubiquitous Gesture Sensing with Millimeter Wave Radar”, ACM Transactions on Graphics, 2016.
15. A. Bernstein et al., “Soli-enabled Non-Contact Heart Rate Detection for Sleep and Meditation Tracking”, Scientific Reports, 2023.
16. Federal Communications Commission, 47 CFR §15.245: Operation within the bands 902–928 MHz, 2435–2465 MHz, 5785–5815 MHz, 10500–10550 MHz, and 24075–24175 MHz.
17. Federal Communications Commission, 47 CFR §15.517: Technical requirements for indoor UWB systems.
18. F. Adib et al., Smart Homes that Monitor Breathing and Heart Rate, Proceedings of CHI, 2015.