The mmWave Promise and the Coverage Challenge

Millimeter-wave (mmWave) spectrum in FR2 -- specifically the 24.25--29.5 GHz (n257, n258, n261) and 37--43.5 GHz (n260) bands -- offers enormous bandwidth: up to 400 MHz per carrier and the potential for multi-gigabit throughput per user. 3GPP defines FR2 in TS 38.104 Table 5.2-1, with maximum channel bandwidths of 400 MHz and subcarrier spacings of 60 kHz and 120 kHz.

However, mmWave faces fundamental propagation challenges that limit its practical coverage radius to 100--300 meters in most urban deployments. Understanding these challenges is essential for any engineer involved in 5G FR2 planning, whether for enhanced mobile broadband (eMBB), fixed wireless access (FWA), or venue coverage.

Propagation Mechanisms at mmWave

Free-Space Path Loss

The Friis free-space path loss (FSPL) increases with the square of frequency:

FSPL (dB) = 20log10(4pidf/c)

At a reference distance of 100 m:

  • 3.5 GHz (FR1): FSPL = 20log10(4pi1003.5e9/3e8) = 83.3 dB
  • 28 GHz (FR2): FSPL = 20log10(4pi10028e9/3e8) = 101.4 dB
  • 39 GHz (FR2): FSPL = 20log10(4pi10039e9/3e8) = 104.3 dB

The 18 dB additional path loss at 28 GHz compared to 3.5 GHz means the signal at 100 m is 63 times weaker. At 39 GHz it is 21 dB worse -- 126 times weaker.

Path Loss Comparison Across Bands

FrequencyBandFSPL at 100 mFSPL at 200 mFSPL at 500 mAdditional vs 3.5 GHz
700 MHzn2869.3 dB75.3 dB83.3 dB-14.0 dB (better)
1800 MHzn377.5 dB83.5 dB91.5 dB-5.8 dB
3500 MHzn7883.3 dB89.3 dB97.3 dB0.0 dB (reference)
28 GHzn257/n261101.4 dB107.4 dB115.4 dB+18.1 dB
39 GHzn260104.3 dB110.3 dB118.3 dB+21.0 dB

This table uses free-space only. Real-world urban environments add significant additional losses from clutter, diffraction, and blockage.

3GPP Channel Models for FR2

The 3GPP channel model in TR 38.901 defines specific path loss equations for frequencies up to 100 GHz across multiple scenarios. The key models for urban mmWave are:

UMi-Street Canyon (LoS): PL = 32.4 + 21log10(d) + 20log10(f) (d in meters, f in GHz) UMi-Street Canyon (NLoS): PL = 22.4 + 35.3log10(d) + 21.3log10(f) - 0.3*(h_UT - 1.5)

The NLoS path loss exponent of 3.53 (from the 35.3*log10(d) term) compared to 2.1 for LoS means that non-line-of-sight conditions add approximately 20--30 dB of additional loss at typical urban distances.

Blockage and Penetration Loss

The most severe challenge for mmWave is blockage. At 28 GHz and 39 GHz, common materials that are semi-transparent at sub-6 GHz frequencies become nearly opaque.

MaterialPenetration Loss at 3.5 GHzPenetration Loss at 28 GHzPenetration Loss at 39 GHzSource
Clear glass (single pane)2--4 dB4--8 dB5--10 dBTR 38.901 Table 7.4.3-1
Low-E coated glass8--15 dB25--40 dB30--45 dBNYU WIRELESS measurements
Brick wall (single)8--12 dB25--35 dB30--40 dBTR 38.901
Concrete wall (15 cm)12--18 dB35--50 dB40--55 dBTR 38.901
Human body3--5 dB15--25 dB20--30 dBTR 38.901 Sec 7.6.1
Foliage (tree canopy)5--10 dB15--25 dB per tree20--30 dB per treeITU-R P.833
Vehicle (side window)5--8 dB20--30 dB25--35 dB5GAA measurements

The human body blockage of 15--25 dB at 28 GHz is a critical issue for handheld devices. A user's hand and head can block the direct path from the base station to the phone antenna, causing sudden signal drops. 3GPP models this as a self-blockage zone defined in TR 38.901 Section 7.6.4, with a blockage cone of approximately 120 degrees around the UE body.

Rain and Atmospheric Absorption

Rain fade at mmWave frequencies follows the ITU-R P.838 model. At 28 GHz, rain attenuation is approximately 1 dB/km for light rain (5 mm/hr) but increases to 10--12 dB/km for heavy rain (50 mm/hr). At 39 GHz, the attenuation is roughly 30% higher.

For typical urban mmWave cell radii of 100--200 m, rain fade adds 0.1--2.4 dB -- generally manageable. However, for FWA links of 300--500 m, heavy rain can add 3--6 dB of attenuation, which may break the link margin.

Atmospheric gases (oxygen and water vapor) contribute less than 0.5 dB/km at both 28 GHz and 39 GHz, making gas absorption negligible for short-range urban cells. The oxygen absorption peak near 60 GHz (up to 15 dB/km) is the primary reason the 57--71 GHz band (WiGig/802.11ad) is limited to very short range.

Worked Example 1: mmWave Link Budget at 28 GHz

Scenario: Urban small cell deployment at 28 GHz, cell radius target 150 m, outdoor UE.
ParameterValueNotes
Carrier frequency28 GHz (n261)100 MHz bandwidth
gNB Tx power35 dBm (3.2 W per panel)3GPP max for local area BS
gNB antenna gain24 dBi256-element phased array, 8 analog beams
gNB beamforming gain15 dBAnalog beam sweeping, 8 SSB beams
UE antenna gain10 dBi4-element phased array module
UE beamforming gain6 dB4 beam directions
EIRP35 + 24 = 59 dBmBelow FCC limit of 75 dBm EIRP
Path loss (UMi LoS, 150 m)32.4 + 21log10(150) + 20log10(28) = 32.4 + 45.6 + 29.0 = 107.0 dBTR 38.901 UMi-LoS
Shadow fading margin7 dBLog-normal, sigma = 4 dB, 90% edge reliability
Body/blockage margin10 dBSingle obstruction allowance
Rain fade margin1 dBLight rain at 150 m
Total path loss + margins107 + 7 + 10 + 1 = 125 dB
Received power59 - 125 + 10 + 6 = -50 dBm
Thermal noise (100 MHz BW)-174 + 10*log10(100e6) + 7 (NF) = -87 dBmUE noise figure = 7 dB
SNR-50 - (-87) = 37 dB
Required SNR for 256-QAM, 0.93 rate22 dBFrom TS 38.214 CQI table
Margin37 - 22 = 15 dBHealthy link margin

This link budget shows that 28 GHz can deliver 256-QAM at 150 m under LoS conditions with single blockage. However, under NLoS:

NLoS path loss at 150 m: 22.4 + 35.3log10(150) + 21.3log10(28) = 22.4 + 76.8 + 30.8 = 130.0 dB

With NLoS, total loss becomes 130 + 7 + 10 + 1 = 148 dB, giving received power of 59 - 148 + 16 = -73 dBm, SNR = 14 dB. This supports only QPSK or 16-QAM, reducing throughput by 60--80%.

Worked Example 2: FWA Coverage at 39 GHz

Scenario: Verizon-style FWA deployment at 39 GHz with Customer Premises Equipment (CPE) window-mounted unit.
ParameterValueNotes
Carrier frequency39 GHz (n260)400 MHz bandwidth
gNB Tx power38 dBm64T64R AAS panel
gNB antenna gain + BF28 + 18 = 46 dBi512-element array, digital beamforming
EIRP38 + 46 = 84 dBmAt FCC EIRP limit
CPE antenna gain + BF22 dBiExternal phased array, roof/window mount
Target distance300 mTypical FWA range
LoS path loss32.4 + 21log10(300) + 20log10(39) = 32.4 + 52.0 + 31.8 = 116.2 dB
Building penetration (low-E glass)35 dBWindow-mount CPE behind coated glass
Shadow fading8 dB90% reliability
Rain margin2 dBModerate rain at 300 m
Total loss116.2 + 35 + 8 + 2 = 161.2 dB
Received power84 - 161.2 + 22 = -55.2 dBm
Noise floor (400 MHz)-174 + 10*log10(400e6) + 7 = -81 dBm
SNR-55.2 - (-81) = 25.8 dB
Achievable modulation64-QAM, rate 0.77CQI 11
Estimated DL throughput~1.5 Gbps400 MHz, 64-QAM, 4 layers

The critical factor is the 35 dB building penetration loss for low-E glass. If the CPE is mounted externally (zero penetration loss), the SNR jumps to 60.8 dB, enabling full 256-QAM and throughputs exceeding 3 Gbps. This is why Verizon and T-Mobile strongly recommend external CPE mounting for FWA customers.

Beam Management: Compensating for Propagation Challenges

mmWave relies on beamforming to overcome the propagation deficit. 3GPP defines the beam management framework in TS 38.321 (MAC) and TS 38.331 (RRC), with four key procedures:

  1. P1 (Initial beam acquisition): gNB sweeps SSB beams across the coverage area. FR2 supports up to 64 SSB beams per cell (vs 8 in FR1), defined in TS 38.213 Section 4.1.
  2. P2 (gNB beam refinement): Narrows from wide SSB beams to narrow CSI-RS beams for traffic.
  3. P3 (UE beam refinement): UE adjusts its receive beam direction.
  4. Beam failure recovery: If the serving beam is blocked, UE initiates beam failure recovery via PRACH, selecting a new candidate beam from SSB measurements.

Beam Management Timelines

EventTypical LatencyImpact
SSB beam sweep (64 beams, 120 kHz SCS)5 msInitial access delay
P2 refinement (CSI-RS based)2--5 msImproved gain after initial access
P3 UE beam switch1--2 msTransparent to gNB
Beam failure detection (BFD)20--50 msBased on BFD timer and threshold
Beam failure recovery (BFR)10--30 msVia contention-free PRACH
Total beam switch (blockage event)30--80 msPerceivable as brief throughput dip

The 30--80 ms beam recovery time means that a pedestrian walking past and temporarily blocking the path causes a brief throughput drop. For vehicular UEs, where blockage changes rapidly, beam management overhead increases significantly.

Operator Deployment Data

Verizon 5G Home (FWA at 28 GHz and 39 GHz)

Verizon has deployed mmWave FWA in over 50 markets since 2018, accumulating significant real-world data:

  • Average cell radius: 200 m (28 GHz), 150 m (39 GHz)
  • Median DL throughput (external CPE): 900 Mbps (28 GHz, 400 MHz), 1.5 Gbps (39 GHz, 800 MHz aggregated)
  • Median DL throughput (window CPE): 350 Mbps (28 GHz), 500 Mbps (39 GHz) -- 60% throughput reduction due to glass penetration
  • FWA subscriber density: 6--12 CPEs per mmWave cell
  • Link availability: 99.7% (external mount), 97.2% (window mount) over 12-month period
  • Rain outage events: 0.3% of time in Southeast US (heavy rain markets like Houston)

T-Mobile mmWave Venue Deployment

T-Mobile deploys mmWave primarily for dense venue coverage (stadiums, airports, convention centers):

  • Typical deployment: 8--16 mmWave small cells per stadium, covering 30,000--70,000 seats
  • Per-user throughput during events: 100--300 Mbps DL (compared to 10--20 Mbps on mid-band alone)
  • SoFi Stadium (Los Angeles): 72 mmWave small cells, peak aggregate capacity 120 Gbps
  • Key learning: ceiling-mounted units with 15-degree downtilt provide best seat-level coverage while minimizing body blockage from standing spectators

Reflection, Diffraction, and Scattering at mmWave

Unlike sub-6 GHz, where diffraction around building corners provides usable signal, mmWave diffraction is extremely weak. The diffraction loss at a knife-edge obstacle scales with frequency -- at 28 GHz, the diffraction loss around a building corner is 25--35 dB compared to 10--15 dB at 3.5 GHz.

However, mmWave benefits from strong specular reflection off smooth surfaces (glass, metal, concrete). Reflected paths can provide viable links when the direct path is blocked, with reflection coefficients of 0.7--0.9 for glass and metal at near-normal incidence. This is why urban street canyon environments often provide better mmWave coverage than open areas -- the reflected paths create a rich multipath environment that beamforming can exploit.

Scattering from rough surfaces (brick, stucco, vegetation) disperses the signal in many directions, reducing the usable reflected power. The Rayleigh criterion defines a surface as "rough" when surface height variations exceed lambda / (8 * cos(theta_i)). At 28 GHz, lambda = 10.7 mm, so surfaces with roughness exceeding ~1.3 mm appear rough -- this includes most building facades except glass and polished metal.

Design Guidelines for mmWave Coverage

Based on propagation physics and operator deployment experience, the following guidelines apply:

  1. Assume 100--200 m cell radius for urban outdoor eMBB at 28 GHz; 80--150 m for 39 GHz
  2. FWA requires external CPE for reliable service beyond 200 m -- window-mount CPE loses 25--40 dB through modern low-E glass
  3. Deploy at street level (6--10 m height) for pedestrian coverage; higher mounting increases LoS probability but reduces street-level signal due to tilt
  4. Plan for 3--4x site density compared to mid-band (n78) coverage at the same QoS target
  5. Leverage reflections: mount antennas facing building facades across the street to create reflected coverage around corners
  6. Indoor coverage from outdoor mmWave is generally not viable -- use indoor small cells or mid-band for in-building coverage

Key Takeaway: mmWave at 28 GHz and 39 GHz delivers multi-gigabit throughput but faces 18--21 dB additional free-space path loss compared to 3.5 GHz, 15--40 dB material penetration loss, and 15--25 dB human body blockage. Practical cell radii are 100--200 m outdoors. Verizon's FWA data shows 60% throughput loss for window-mounted vs external CPE, confirming that building penetration is the dominant coverage limiter. Beam management with up to 64 SSB beams and beam failure recovery in 30--80 ms partially compensates, but mmWave fundamentally requires dense small-cell deployment and careful site placement to exploit LoS paths and reflections.