Study Guide — Medium Earth Orbit (MEO) Satellite Systems

Unit Overview

Between the crowded, drag-swept low orbits and the distant geostationary ring lies a vast and under-appreciated territory: Medium Earth Orbit, 2 000–35 786 km. MEO hosts humanity's most precise infrastructure — the Global Navigation Satellite Systems (GPS, GLONASS, Galileo, BeiDou) — and a growing family of communications constellations (O3b/mPOWER). MEO's physics is distinctive: orbits of hours not minutes, multi-hour passes without handover, and a harsh radiation environment that makes satellites expensive. This guide develops the orbital mechanics, the GNSS design doctrine, the radiation problem, link budgets, and the system trade-offs that make MEO simultaneously the best and the hardest place to put a satellite.

Learning Outcomes

After studying this unit and completing the tutorials you should be able to:

Compute MEO orbital parameters and explain the half-sidereal-day orbit design.
Design GNSS constellations: Walker geometry, 4-satellite visibility, DOP reasoning.
Explain pseudoranging, the clock-bias problem, and GNSS signal structure (L1/L2/L5).
Describe the Van Allen belts' impact on MEO spacecraft design.
Build MEO link budgets and compare them with LEO and GEO.
Compare MEO communications constellations (O3b) against LEO and GEO alternatives.
Quantify MEO latency, coverage, and lifetime characteristics.
Select between MEO/LEO/GEO for a given mission with a defensible trade argument.

Indicative Study Time

Activity	Time
Guided reading (Sections 1–8)	5 hours
Figures studied / examples reworked	2 hours
Tutorial problems (Section 9)	3 hours
Revision of key equations & glossary	1 hour

Prerequisite knowledge: Kepler's third law and vis-viva equation; coverage geometry (arccos formula with elevation mask); link budgets (EIRP, G/T, C/N₀); basic orbital elements (a, e, i, Ω, ω).

Section 1 — MEO Fundamentals

1.1 The Regime

MEO: 2 000 km < h < 35 786 km ⇔ periods from ~2.2 h up to 23 h 56 min (sidereal day)

Altitude	Period	Velocity	Resident system
8 062 km	4.79 h	5.16 km/s	O3b / mPOWER
19 130 km	11.26 h	4.24 km/s	GLONASS
20 180 km	11.97 h ≈ ½ sidereal day	4.19 km/s	GPS
21 528 km	12.9 h	4.10 km/s	BeiDou MEO
23 222 km	14.2 h	3.95 km/s	Galileo

1.2 The Half-Sidereal-Day Design

All four GNSS chose periods near half a sidereal day (GPS: 11 h 58 min exactly). Consequences:

Repeating ground track: Earth rotates 180° during one orbit, so after two orbits the satellite retraces the same ground track to within a kilometre — predictable, repeatable coverage geometry.
Constellation symmetry: with satellites evenly phased, the 4-satellite-minimum geometry repeats daily — simple almanacs, simple receivers.
Averaged perturbations: resonant gravitational harmonics average out symmetrically over the half-day cycle.

Fig. B — FSPL (1.5 GHz) and one-way vacuum delay versus altitude. MEO buys a ~17 dB path-loss reduction and ~100 ms latency reduction versus GEO, at the price of moving through the radiation belts and losing GEO's fixed-pointing simplicity.

1.3 Coverage Geometry

ρ = arccos( (R_E/(R_E+h)) cos ε_min ) − ε_min ; at 20 180 km, ε = 5°: ρ ≈ 71°

A single GPS satellite at 5° mask covers ~38% of the globe; the constellation geometry (Section 2) exploits this enormous footprint with only 24–30 satellites.

Check Your Understanding 1

Compute the period of a circular orbit at 23 222 km (Galileo).
Verify GPS's half-sidereal-day period from Kepler's third law.
Compute the service half-angle at 23 222 km with a 5° mask, and the single-ring constellation minimum.

Answers

a = 29 593 km → T = 2π√(a³/μ) ≈ 51 183 s ≈ 14.2 h.
a = 26 551 km → T = 2π√(26 551³/3.986×10⁵) ≈ 43 082 s ≈ 11 h 58 min ✓.
ρ = arccos((6371/29 593)cos5°) − 5 = arccos(0.2141) − 5 = 77.6° − 5° = 72.6°; ring minimum ≈ 360/145.2 ≈ 2.5 → 3 satellites.

Section 2 — GNSS: The Flagship MEO Application

2.1 Constellation Design Doctrine

System	Altitude	Planes × sats	Inclination	Design intent
GPS	20 180 km	6 × 4 = 24 + spares	55°	Global ≥4-sat visibility, 24-slot expandable
GLONASS	19 130 km	3 × 8 = 24	64.8°	Higher latitude performance (Russian territory)
Galileo	23 222 km	3 × 8 = 24 + 6 spares	56°	Commercial service, search-and-rescue payloads
BeiDou	MEO 21 528 km + GEO/IGSO	MEO 24 + 3 GEO + 3 IGSO	55°	Hybrid: MEO global + GEO/IGSO Asia-Pacific augmentation

Fig. A — Simulated GPS visibility over 24 h (24 sats, 6 planes × 4, i = 55°): a minimum of 6 satellites above 5° elevation everywhere, averaging 8–9 — comfortably above the 4 needed for a fix. This simulation is the quantitative backing for the "≥4 visible" design claim.

2.2 Why Four Satellites?

ρ_i = √[(x−x_i)² + (y−y_i)² + (z−z_i)²] + c·δt (pseudorange)

Four unknowns (receiver x, y, z and clock bias δt) need four independent pseudoranges. Three satellites with a perfect clock would suffice — but no receiver clock is perfect, and a 1 µs clock error is a 300 m error. Geometry quality is measured by DOP (Dilution of Precision): GDOP ≈ 1 when satellites are well spread; error σ_pos = GDOP × σ_range.

2.3 GNSS Signal Structure (Engineering View)

Band	GPS	Purpose
L1 1 575.42 MHz	C/A (civilian) + P(Y) (military)	Primary ranging; ionosphere correction via dual frequency
L2 1 227.60 MHz	P(Y), L2C (modernized)	Second frequency → civilian ionosphere correction
L5 1 176.45 MHz	L5 (safety-of-life)	Aviation; higher power, wider bandwidth

Code vs carrier: pseudorange from code phase (m–dm accuracy), carrier phase gives cm–mm (surveying, RTK).
Ionosphere: the largest natural error (1–100 m); dual-frequency receivers remove ~99% of it — the reason GPS is "dual-frequency by design".
Relativistic correction: satellite clocks are rate-offset by −4.45×10⁻¹⁰ before launch to compensate gravitational + velocity effects — without it, GPS would drift ~10 km/day. A favourite exam question.

Check Your Understanding 2

Why does GLONASS use 64.8° inclination while GPS uses 55°?
Why can surveying receivers achieve centimetre accuracy with the same signals that give metre accuracy in a phone?
Compute the daily range drift if GPS ignored the relativistic clock offset (offset 4.45×10⁻¹⁰, c = 3×10⁸ m/s).

Answers

Russian territory spans high latitudes; 64.8° keeps satellites higher in the sky there, improving geometry and availability at 60°+ N.
Carrier-phase measurements (2 mm wavelength vs 300 m C/A chip) plus differential/RTK techniques remove common errors — the signals are the same; the measurement technique and geometry differ.
Δt per day = 4.45×10⁻¹⁰ × 86 400 s = 3.85×10⁻⁵ s → c·Δt ≈ 11.5 km/day.

Section 3 — MEO Communications Systems

3.1 The O3b Model

O3b ("Other 3 billion") — now SES's mPOWER generation — flies Ka-band broadband at ~8 062 km: below the navigation belt, above LEO. Design rationale:

Latency: ~125 ms round trip vs GEO's ~500+ ms — acceptable for enterprise/cloud apps that GEO cannot serve.
Coverage: each satellite covers a wide equatorial footprint (~45° half-angle with 5° mask at 8 062 km); 20 satellites in equatorial orbit cover ±45° latitude — the populated band.
Gateway economics: few satellites, fixed-ish pointing, steerable beams follow demand (aircraft, ships, energy sites).

3.2 The Failed 1990s MEO Comms

ICO Global Communications and Odyssey (TRW) proposed telephony at ~10 000 km in the 1990s; both died with the telecom bust. Lessons: MEO comms must (a) target latency-sensitive but bandwidth-moderate markets, (b) avoid duplicating fiber's strengths, and (c) survive on tens — not thousands — of satellites. O3b succeeded by serving trunk/backhaul rather than handsets.

3.3 Bent-Pipe vs Regenerative at MEO

Architecture	Advantage at MEO	Cost
Bent-pipe (transparent)	Simple, protocol-agnostic	Double hop for ISL-less routing
Regenerative (onboard processing/ISL)	Single-hop mesh, traffic grooming	Complex payload; MEO sats are few and expensive to risk

MEO's small constellations make onboard processing affordable to manage centrally from ground — most MEO comms sats are bent-pipe with ground-routed traffic.

Check Your Understanding 3

Why does O3b's equatorial orbit suffice commercially despite covering only ±45°?
Estimate the round-trip latency for a 4 000 km ISL-free MEO hop (up 8 062 km, down 8 062 km, terrestrial tail 20 ms).

Answers

~90% of humanity lives within ±45° latitude; polar coverage adds cost without revenue for a trunk operator.
Propagation 2×8 062/3×10⁵ = 54 ms + terrestrial 20 ms ≈ 74 ms one-way-ish… for a round trip of user→sat→gateway→sat→user: 4×27 ms + 20×2 ≈ 148 ms RTT — still 3× better than GEO.

Section 4 — The Radiation Environment: MEO's Tax

4.1 The Van Allen Belts

Fig. C — Schematic trapped-particle environment. GNSS altitudes (19 000–23 000 km) sit near the outer electron belt peak: every orbit threads the harshest electron environment in Earth orbit.

4.2 Consequences for Spacecraft Design

Total ionizing dose (TID): MEO GNSS sats accumulate 10–20 krad(Si)/yr behind moderate shielding — radiation-hardened (rad-hard) parts mandatory (10×+ cost of COTS).
Single-event effects: outer-belt electrons cause deep-dielectric charging and SEUs; redundant computers with voting, error-correcting memory, and periodic watchdog resets are standard.
Solar cycle: belt flux varies ~10× over 11 years; design for the worst case — mass penalty paid for the whole mission.
Design consequence: MEO satellites cost 3–10× more per unit than LEO sats of similar capability; mission life 10–15 years amortizes this.

Example 4.1 — The charging hazard. A dielectric (solar-array cover glass, cable insulation) exposed to >1 MeV electrons accumulates charge; if discharge occurs through a circuit, micro-dischays can destroy electronics. Mitigations: conductive coatings, grounding standards, charge-drain resistors — all standard on GNSS buses, all unnecessary on LEO broadband sats at 550 km. This one paragraph explains a visible share of the MEO/LEO cost gap.

Check Your Understanding 4

Why is LEO (550 km) largely immune to the belt problem while MEO is not?
Why must MEO sats be designed for solar-maximum belt conditions even though most of the mission may occur in quieter years?

Answers

The inner belt begins ~1 000 km and peaks ~3 000–6 000 km; 550 km sits below the trapped population — only the SAA tail matters.
Damage is cumulative and irreducible: a single solar-maximum year can deliver most of the mission dose; components cannot be replaced, so the design must survive the worst year it might meet.

Section 5 — MEO Link Engineering

5.1 The Budget

C/N₀ = EIRP + (G/T)_rx − L_FS − L_atm − L_misc + 228.6

Example 5.1 — GNSS L1 budget to a phone. d = 20 180 km+slant, f = 1.575 GHz: L_FS = 92.45 + 3.97 + 20log(25 000 slant) ≈ 184.4 dB (with 25° elevation slant ≈ 24 000–26 000 km). GPS L1 C/A EIRP ≈ 26–28 dBW (incl. 14 dB spreading loss relative to peak); phone G/T ≈ −28 dB/K; L_{atm/iono/misc} ≈ 1.5 dB.
C/N₀ ≈ 27 − 28 − 184.4 − 1.5 + 228.6 ≈ 41.7 dBHz. A GPS receiver processes the C/A code with ~43 dB processing gain — the signal arrives 20 dB below thermal noise, yet is fully recoverable. This "signal buried in noise" property is unique to spread-spectrum links.

5.2 MEO Comms Budget (Ku/Ka)

For trunk links (O3b-class): EIRP 45–55 dBW per beam, gateway G/T 25–30 dB/K, rain margins 6–10 dB via site diversity — C/N₀ ≈ 95–105 dBHz supporting hundreds of Mbps per beam. The FSPL sits ~13 dB below GEO (Fig. B): the same ground antenna, pointed at MEO, closes links GEO cannot — or permits 13 dB more rain margin.

5.3 Latency and Doppler at MEO

Latency: 67 ms one-way at 20 180 km (Fig. B): fine for everything except high-frequency trading and interactive gaming at the margin.
Doppler: v ≈ 4.2 km/s → at L1 (1.575 GHz), up to ±4.5 kHz — trivially handled by GNSS receivers (it is, in fact, measured and exploited); at Ka (20 GHz) ≈ ±40 kHz — needs demodulator compensation but far less than LEO's ±350 kHz.

Check Your Understanding 5

Compute FSPL at 1.227 GHz (L2) for a 20 000 km path, and the maximum L2 Doppler at v = 4.2 km/s.
Why does a phone's GPS work indoors where a satellite TV dish (GEO, Ku) cannot?

Answers

L_FS = 92.45 + 1.78 + 20log(20 000) = 92.45 + 1.78 + 86.0 ≈ 180.3 dB; Δf = (4.2/3×10⁵)×1.227×10⁹ ≈ ±4.3 kHz.
Processing gain: the spread C/A code (1 MHz chip rate → 30 dB+ gain) digs the signal out of noise and attenuation; GEO TV uses high-order modulation needing 10+ dB C/N — walls kill it, while GPS tolerates ~25 dB attenuation.

Section 6 — Orbit and Constellation Design Details

6.1 Walker Constellations at MEO

GPS: Walker 55°:24/6/2 — T = 24, P = 6 planes, F = 2 phasing; Galileo: 56°:27/3/1

Design checks for any MEO constellation:

Visibility floor: ≥4 sats above mask at worst latitude (equator usually binds; Fig. A).
Geometry quality: PDOP < 6 at the 95th percentile — achieved by spreading planes, not adding altitude.
Slot strategy: spare satellites pre-positioned in plane ("on-orbit spares") — replacing a 20 000 km satellite costs months; spares hide the failure.

6.2 Station-Keeping and Lifetime

Perturbations at MEO: lunisolar gravity strongly perturbs i and Ω; resonance with Earth's J₂ at half-sidereal-day periods is benign by design.
Station-keeping: modest ΔV (~1–5 m/s/yr), 10–15-year lives; disposal orbits raised ~300 km above constellation at end of life (graveyard).
Debris environment: MEO is sparse — conjunction rates far below LEO; long decay times (centuries) are acceptable precisely because few objects reside there.

Check Your Understanding 6

Why do GNSS operators keep on-orbit spares rather than launching a replacement upon failure?
Compute the minimum number of satellites for single-fold global coverage from 20 180 km using the coverage-fold estimate N ≈ 4.84/sin²(ρ) with ρ = 71° (5° mask), and compare with GPS's 24.

Answers

A launch campaign to 20 180 km takes months and $50–100M+; a spare re-phases within days. MEO economics demand insurance in orbit.
N ≈ 4.84/sin²71° ≈ 4.84/0.895 ≈ 5.4 — but this is an idealized continuous single-fold coverage bound; the 4-visible-satellite requirement with geometry (PDOP), plane phasing losses, and slotting granularity drive the practical count to 24.

Section 7 — Case Studies

Fig. D — The MEO family: communications constellations occupy the lower band (~8 000–10 000 km, below the worst radiation); navigation constellations occupy the upper band (~19 000–23 000 km), accepting the radiation tax for the coverage geometry.

Case	Lesson
GPS (1978–today)	Half-sidereal-day orbit + 24-slot Walker = maintainable global utility for 45+ years; incremental modernization (L2C, L5, L1C) without redesigning the constellation.
Galileo (2011–)	European sovereignty argument; higher integrity via dual-frequency free-of-charge service; demonstrates that 24-sat MEO is replicable, not unique to GPS.
GLONASS (1982–)	FDMA legacy (channel-per-satellite) vs GPS CDMA — receiver complexity lesson; modernization to CDMA ongoing.
BeiDou hybrid	MEO + GEO + IGSO: GEO/IGSO deliver high-elevation geometry and messaging (short-message service) over Asia — a regional augmentation on a global backbone.
O3b → mPOWER (2009–)	MEO comms works when targeting trunk/backhaul with steerable beams; software-defined payloads (mPOWER) follow the LEO industry lesson of upgradability.
ICO/Odyssey (1990s)	Cautionary: MEO telephony failed against terrestrial cellular economics — orbit choice cannot rescue a weak business model.

Section 8 — Trade-offs: MEO vs LEO vs GEO

Attribute	LEO	MEO	GEO
Constellation size (global)	tens–thousands	~20–30 (the minimum for global)	3 (excl. poles)
One-way delay	2–4 ms	~67–110 ms	~119 ms
Satellite lifetime	5–7 yr	10–15 yr	15+ yr
Radiation hardness	minimal	severe (outer belt)	moderate (belts skirted at transfer)
Handover	every minutes	none needed (hours-long passes)	none
Coverage per satellite	small	~38% of globe	~42% of globe
Per-satellite cost	low (COTS)	high (rad-hard)	high
Best missions	capacity, latency, imaging	navigation, trunk latency-sensitive	broadcast, fixed services

The MEO thesis: MEO is optimal exactly when a mission needs global geometry from few satellites with acceptable latency — navigation being the perfect case. It loses to LEO on capacity-per-dollar and to GEO on simplicity. Choose MEO when 24 satellites can do what 4 000 (LEO) or 3-with-huge-footprint (GEO) cannot.

Section 9 — Tutorial Problems

Constants: μ = 3.986×10⁵ km³/s², R_E = 6 371 km, sidereal day = 86 164 s, c = 3×10⁵ km/s, FSPL = 92.45 + 20log f(GHz) + 20log d(km).

Compute period, velocity, and coverage half-angle (5° mask) for the O3b altitude of 8 062 km.
A proposed Kenyan regional augmentation uses 3 satellites at 20 180 km, i = 0° (equatorial). Compute coverage, ring geometry, and argue why equatorial MEO suits East Africa.
Verify Galileo's period (23 222 km) and show its ground track repeats every (approximately) how many days given T = 14.2 h.
Build a GNSS L5 (1 176.45 MHz) link budget: EIRP 29 dBW, slant 25 000 km, receiver G/T = −25 dB/K, losses 1.5 dB. Find C/N₀ and the processing gain needed to demodulate at C/N = 6 dB in a 10 Hz data bandwidth.
Estimate the maximum Doppler at L1 for a receiver on the equator directly under a GPS satellite at 20° elevation (use v = 4.19 km/s; maximum radial component occurs at rise/set geometry — take v_rad,max ≈ v·cos(el)·(R_E/(R_E+h)) as seen from the ground).
Compare PDOP reasoning: sketch (describe) why 4 satellites clustered within 30° of the zenith gives poor GDOP, and why MEO geometry naturally spreads satellites in elevation.
Radiation design: a MEO satellite must survive 15 krad(Si) total dose behind 3 mm aluminium. Given that a given COTS part survives 1 krad and a rad-hard equivalent 100 krad (at 10× cost), choose a strategy (shielding thickness trade, part selection, or both) and defend it in cost terms.
Compute round-trip latency for a voice call routed: Nairobi user → MEO satellite (20 180 km) → London gateway → fiber return 7 000 km. Compare with the same call via GEO.
Constellation design: using ρ = 71° at 20 180 km (5° mask), design a 3-plane Walker for a fictional equatorial-region GNSS that guarantees ≥4 sats from 3° S to 10° N. State sats per plane, phasing F, and inclination; justify each choice in ≤2 lines.
Orbit trade essay (1 page): "GNSS belongs in MEO; broadband belongs in LEO; broadcast belongs in GEO." Attack or defend each clause with quantitative arguments (latency, geometry, coverage, radiation, constellation size).

Selected solutions

a = 14 433 km; T = 2π√(a³/μ) ≈ 17 247 s ≈ 4.79 h; v = √(μ/a) ≈ 5.25 km/s; ρ = arccos((6371/14 433)cos5°) − 5 = arccos(0.4400) − 5 = 63.9° − 5° = 58.9°.
3 sats at 120° spacing on the equator at 20 180 km: each covers ρ ≈ 71° — overlapping footprints blanket the equatorial belt ±60° continuously; East Africa sits near the sub-satellite longitudes → near-zenith passes, minimal path loss and maximal availability. A true GNSS needs inclined planes for latitude diversity; an equatorial ring is a regional augmentation, not a global system.
a = 29 593 km → T ≈ 14.23 h. Repeat time: find n·T ≈ m·sidereal day → 14.23×17 ≈ 242 h ≈ 10.07 days — the ground track nearly repeats every ~10 days (Galileo's actual repeat is ~10 days; GPS is exactly 1 sidereal day).
L_FS = 92.45 + 1.40 + 20log(25 000) = 92.45 + 1.40 + 88.0 = 181.9 dB. C/N₀ = 29 − 25 − 181.9 − 1.5 + 228.6 = 49.2 dBHz. Required C/N₀ = 6 + 10log(10) = 16 dBHz → processing gain ≥ 33 dB (L5's 10.23 MHz chip rate provides ~45 dB — ample).
v_rad,max ≈ 4.19 × cos20° × (6371/26 551) ≈ 4.19 × 0.940 × 0.240 ≈ 0.945 km/s → Δf = (0.945/3×10⁵)×1.575×10⁹ ≈ ±4.96 kHz.
Clustered near zenith → line-of-sight unit vectors nearly parallel → small geometry matrix determinant → GDOP blows up; MEO's ~20 000 km altitude puts satellites at elevations from 5° to near-zenith with azimuthal spread by plane phasing — the elevation spread is the geometry gift GEO (all at one elevation band) cannot match.
Strategy: rad-hard for critical functions (computers, power control) at 10× cost is non-negotiable; for the rest, evaluate shielding: dose ∝ e^(−t/λ) with λ ≈ 2.5 mm Al for electrons → going from 3 mm to 6 mm halves dose but adds mass; choose rad-tolerant (10 krad) parts + 3 mm shielding + spot shielding of sensitive volumes; total cost ≈ 2–3× COTS baseline versus 10× full rad-hard — standard MEO practice.
MEO: 2×20 180/3×10⁵ = 0.134 s + fiber 7 000/2×10⁵ = 0.035 s ≈ 170 ms RTT. GEO: 2×35 786/3×10⁵ = 0.239 s + 0.035 ≈ 274 ms — MEO saves ~100 ms, the O3b selling point.
Design: 3 planes × 8 = 24, i ≈ 8–10°, F = 1, RAANs at 0/120/240°. Justification: low inclination concentrates coverage in the target band (no wasted polar coverage); 8 per plane gives 3-h spacing between passes with phasing F = 1 spreading ascending nodes; 24 ensures ≥4 with one satellite failed.
Model answer: GNSS-in-MEO defended by 4-satellite global geometry from 24 sats (LEO needs 100+ for equivalent geometry continuity; GEO gives no elevation spread); broadband-in-LEO defended by path loss/terminals (MEO broadband works only for trunk, not consumer); broadcast-in-GEO defended by fixed dishes and continental footprint (LEO broadcast would need tracking). Each clause is quantitative, not sentimental.

Section 10 — Glossary of Key Terms

Almanac: Coarse orbital parameters for all constellation satellites, broadcast to assist acquisition.
Bent-pipe: Transparent transponder: receives, amplifies, re-transmits without onboard demodulation.
CDMA / FDMA: Code-division / frequency-division multiple access; GPS uses CDMA, legacy GLONASS FDMA.
DOP (GDOP/PDOP): Dilution of Precision: geometry factor multiplying ranging error into position error.
Ephemeris: Precise orbital parameters for a satellite, broadcast for positioning computation.
Graveyard orbit: Disposal orbit ~300 km above GEO (or re-orbited MEO) for end-of-life satellites.
Half-sidereal-day orbit: Period = 11 h 58 min; satellite ground track repeats daily — the GNSS design choice.
IGSO: Inclined Geosynchronous Orbit: 24-h period with nonzero inclination (BeiDou augmentation).
Pseudorange: Code-derived satellite–receiver range including clock bias: ρ = range + c·δt.
Rad-hard: Radiation-hardened electronics qualified for the MEO dose environment.
RTK: Real-Time Kinematic: carrier-phase differential technique delivering cm-level positions.
Walker constellation: Symmetric (i : T/P/F) constellation geometry; basis of GNSS designs.

Section 11 — References and Further Reading

B. Hofmann-Wellenhof, H. Lichtenegger & J. Collins, GPS: Theory and Practice, 5th ed., Springer, 2001.
E. D. Kaplan & C. J. Hegarty (eds.), Understanding GPS: Principles and Applications, 2nd ed., Artech House, 2006.
T. Pratt, J. Bostian & J. Allnutt, Satellite Communications, 2nd ed., Wiley, 2003.
J. R. Wertz & W. J. Larson (eds.), Space Mission Analysis and Design, 3rd ed., Microcosm Press, 1999.
ESA Navipedia and GPS SPS Performance Standard (public web resources) — signal structures and constellation details.

Exam guidance: MEO questions reward the calculation chain: Kepler → period → coverage angle → constellation count → link budget. Show every step; state the elevation mask; quote the relativistic correction when GPS appears. A number without its geometry is a guess, not an answer.