Medium access: SDMA, FDMA, TDMA and CDMA
Medium access methods let multiple users share the same wireless resource; this note explains SDMA, FDMA, TDMA and CDMA, their principles, trade-offs and worked numerical examples.
Drafted with Aria, reviewed by the AiCanCode.org team · Updated . How we write and check our notes. Spotted an error? Use Give Feedback at the bottom of the page.
Why it matters
Mobile and wireless systems must let many users share limited radio resources (space, frequency, time, code). Engineers design base stations, radios and MAC layers using SDMA, FDMA, TDMA and CDMA to maximise capacity, meet quality-of-service and reduce interference in cellular systems, Wi‑Fi, satellite links and IoT networks.
Key ideas
Start from first principles: a wireless medium carries electromagnetic waves; simultaneous transmissions overlap and interfere at receivers. To allow many users we split the medium into separate resources in different domains: space, frequency, time or code. Each splitting method is a multiple-access technique.
Multiple access — a class of methods that let multiple users share a communication medium without unacceptable interference. (Defined here: a way to partition available resources so users can transmit simultaneously and be recovered at receivers.)
Channel — a distinct resource allocated to a user (e.g., a frequency band, a time slot, a spatial beam, or a spreading code). Channel capacity depends on bandwidth, signal power and interference.
Multiplexing — combining multiple signals to share one physical medium. Multiple access is multiplexing applied to multiple users.
Guard band / guard time — small frequency or time separation introduced between channels to reduce interference from imperfect filters or timing mismatches.
Now the four principal methods:
SDMA (Space Division Multiple Access) — separate users by spatial signature (different directions or beams). Implemented using directional antennas, antenna arrays and beamforming. Assumption: signals from different users arrive through separable spatial channels (angles or independent channels) so spatial filtering reduces interference. Exact statement: if two users have non-overlapping spatial signatures as seen by the receiver antenna array and the array has at least as many independent elements as simultaneous users, linear spatial filtering (beamforming) can, under ideal conditions, separate them with arbitrarily small cross-interference (limited in practice by antenna aperture, scattering and angle spread).
- What to draw: draw a base station with a multi-element antenna array and two mobile users at two distinct angles. Label beams, main lobes, side lobes, and indicate desired and interference signals.
FDMA (Frequency Division Multiple Access) — assign each user a distinct frequency sub-band. Exact statement: total available bandwidth B_total is partitioned into disjoint sub-bands B_i for users; adjacent sub-bands are separated by guard bands g to limit adjacent-channel interference. Assumption: filters can separate bands sufficiently; channel is linear and time-invariant over the allocation.
- Draw: a frequency axis with contiguous blocks labelled channel 1, guard band, channel 2, ...
TDMA (Time Division Multiple Access) — assign each user a distinct time slot on a repeating frame. Exact statement: users share a time-multiplexed channel; a frame of duration T_frame contains N_slots each of duration T_slot, with T_frame = N_slots · T_slot. Assumption: tight time synchronisation to avoid slot overlap, and guard time between slots to accommodate timing errors.
- Draw: a time axis showing repeating frames, slots labelled for users, and guard time between slots.
CDMA (Code Division Multiple Access) — all users transmit simultaneously in the same frequency band and time but use different spreading codes. Exact statement: each user multiplies its data symbols by a user-specific spreading code (sequence of chips) forming a spread-spectrum signal. If users' codes are (near-)orthogonal and power control is used, a receiver can separate users by correlating with the target user's code. Assumption: codes have low cross-correlation (or are orthogonal in synchronous systems), and receiver has sufficient knowledge of codes and timing.
- Important defined terms here:
- Spreading code — a binary (or multi-level) sequence used to spread each data symbol into many chips.
- Chip — one element of a spreading sequence; the chip rate R_c is the rate at which chips are transmitted.
- Processing gain — ratio of spread bandwidth to original data bandwidth (or R_c/R_s). Higher processing gain means better ability to separate signal from noise/interference.
- Important defined terms here:
Key theoretical relations often used in exams:
- FDMA isolates users in frequency; required total bandwidth = sum of user bandwidths + guard bands.
- TDMA provides isolation in time; each user's effective data rate is reduced by the fraction of frame they occupy; guard time reduces usable fraction.
- CDMA uses spreading: for synchronous orthogonal codes, cross-correlation is zero in ideal conditions; in asynchronous or multipath channels, codes are not perfectly orthogonal and interference appears.
Practical trade-offs and notes:
- SDMA requires antenna arrays and good spatial separation; capacity scales with number of independent spatial channels (MIMO ideas link here but are separate topics in the course list).
- FDMA is simple but wastes spectrum in guard bands; good for analog or simple radios.
- TDMA is spectrally efficient but needs tight timing and introduces latency (frame duration).
- CDMA is robust to multipath and soft-capacity (users can be added until interference rises); requires power control (to handle the near‑far problem) and careful code design (Walsh, Gold, Kasami codes).
Where assumptions break: in rich scattering, SDMA separation is harder; in mobile fast fading, TDMA slot timing and CDMA synchronisation requirements tighten.
Formulas
B_total = Σ B_i + Σ g — B_total is total available bandwidth (Hz), B_i are individual channel bandwidths (Hz), g are guard bands between adjacent channels (Hz); applies to FDMA planning.
T_frame = N_slots · T_slot — T_frame is frame duration (s), N_slots number of slots per frame (unitless), T_slot slot duration (s); applies to TDMA framing.
Bits_per_slot = R_physical · T_slot — bits carried in one slot by the physical channel (bits); and Required_bits_per_frame = R_user · T_frame — bits a user must receive each frame to sustain average rate R_user (bits).
R_physical ≈ R_user · N_slots — physical channel rate required (bit/s) when N_slots users time‑share one carrier; increase this to account for guard time and control overhead by dividing by the time‑efficiency factor.
Processing_gain (G_p) = R_c / R_s — R_c chip rate (chips/s), R_s symbol (or data) rate (symbols/s or bit/s if one bit per symbol); G_p unitless. In dB: G_p(dB) = 10 · log10(G_p).
Number_of_Walsh_codes = L where L = R_c / R_s for synchronous CDMA using Walsh codes; L is the code length (unitless) and equals the available orthogonal codes, but L must be a power of two for Walsh–Hadamard codes (choose L = 2^m ≈ R_c/R_s); this applies under synchronous, single-cell conditions.
SINR ≈ S / (Σ_{j≠i} S_j + N0·B) — SINR (signal-to-interference-plus-noise ratio) at a CDMA receiver when S is desired signal power (W), S_j are interfering user powers (W), N0 is one-sided noise spectral density (W/Hz), B is system bandwidth (Hz); applies to link-budget style CDMA analysis.
Notes: bandwidths in Hz; chip/symbol rates in s^-1; bit rates in bit/s; power in watts (W); noise spectral density N0 is one-sided in W/Hz.
Worked examples
Example 1 — FDMA channel packing (numerical) Problem: You have B_total = 20 MHz. You must allocate 8 equal channels. Each channel needs 2 MHz for signals, and you must place a guard band g = 0.1 MHz between adjacent channels. Does the allocation fit? If not, how much extra bandwidth is needed? Solution steps:
- Sum of user bandwidths = 8 × 2 MHz = 16 MHz.
- Number of guard bands between 8 channels = 7. Total guard = 7 × 0.1 MHz = 0.7 MHz.
- Required B_total = 16 + 0.7 = 16.7 MHz.
- Compare with available 20 MHz: 20 − 16.7 = 3.3 MHz spare. Answer: Yes it fits; 3.3 MHz remains unused.
Example 2 — TDMA slot calculation (numerical, bits carry) Problem: A physical carrier supports 8 Mbps. A TDMA frame has N_slots = 8 slots per frame. Each user wants R_user = 1 Mbps sustained. Frame length is T_frame = 10 ms. Find slot duration, bits per slot, and whether user rate is satisfied. Solution steps:
- Slot duration T_slot = T_frame / N_slots = 0.01 s / 8 = 0.00125 s = 1.25 ms.
- Bits per slot for a user = R_physical · T_slot = 8e6 bit/s × 0.00125 s = 10,000 bits. Required bits per frame to sustain 1 Mbps = R_user · T_frame = 1e6 × 0.01 = 10,000 bits (matches).
- Check physical carrying: physical channel at 8 Mbps during one slot carries 8e6 × 0.00125 = 10,000 bits; but that slot is used by one user so user can get 10,000 bits every frame = 10,000 bits / 0.01 s = 1,000,000 bit/s = 1 Mbps. Both ways consistent. Answer: Slot duration 1.25 ms; 10,000 bits carried per slot; user rate satisfied (since each frame delivers 10,000 bits = 1 Mbps).
Example 3 — CDMA processing gain and Walsh code count Problem: In a synchronous CDMA system chip rate R_c = 1.0 Mcps and user symbol (data) rate R_s = 10 kbps. Compute processing gain and number of orthogonal Walsh codes. Solution steps:
- Processing gain G_p = R_c / R_s = 1,000,000 / 10,000 = 100.
- In dB: G_p(dB) = 10 log10(100) = 20 dB (use 10·log10 because G_p is power ratio).
- If using Walsh codes, the code length must be a power of two; since R_c/R_s = 100 is not 2^m, pick L = 128 (closest power of two), yielding up to 128 orthogonal codes and an adjusted symbol rate R_s = R_c/L = 7.8125 kbps. If R_s must remain 10 kbps, use non-Walsh near-orthogonal codes (e.g., Gold/Kasami) rather than a Walsh set. Answer: Processing gain 100 (20 dB); with Walsh codes the code length must be 2^m, so choose L = 128 for a 1 Mcps chip rate (giving up to 128 orthogonal codes and R_s = R_c/L = 7.8125 kbps), or keep R_s = 10 kbps but use non-Walsh near-orthogonal codes (e.g., Gold/Kasami).
(Note: practical CDMA capacity is constrained by interference, power control and code reuse; see Practical trade-offs and notes above.)
Common mistakes
- Forgetting guard bands/time when calculating total bandwidth or frame feasibility.
- Using wrong units: mixing kHz, MHz, kbps, Mbps without conversion.
- In CDMA, claiming codes alone prevent interference without mentioning power control and cross-correlation imperfectness.
- Treating number of Walsh codes as unlimited; it's equal to code length L and limited by chip/symbol rates.
- Using 20·log10 instead of 10·log10 for power ratios (processing gain is a power ratio so use 10·log10).
For your exam
- Diagram questions: draw frequency allocation for FDMA with guard bands; draw TDMA frames; draw base station with beams for SDMA. Label all quantities and mark guard times/bands.
- Derivations: derive T_frame = N_slots·T_slot, derive processing gain G_p = R_c/R_s and show conversion to dB. State assumptions for orthogonality in CDMA.
- Numericals: channel packing (bandwidth), TDMA slot sizing and bits per slot, CDMA processing gain and code count. Show stepwise unit conversions and final bold result.
- Short design/analysis: given user rates and total spectrum, propose allocation (choose FDMA or TDMA or CDMA) and justify trade-offs (latency, hardware complexity, power control).
What full-marks answers include: clear definitions with bolded symbols/units, diagrams with labels, step-by-step calculations with units carried through, and explicit assumptions/limitations.
Quick check
Q1: What does processing gain mean in CDMA? Q2: How many guard bands are there between N FDMA channels? Q3: Write the relation between frame duration, number of slots and slot duration in TDMA. Answers: Processing gain = ratio R_c/R_s (chip rate over symbol rate), number of guard bands = N−1, relation T_frame = N_slots · T_slot.
Finished this topic? Mark it so your progress, study plan and readiness keep up.
Stuck on something here?