TDD cells share a frequency band but alternate between downlink and uplink. When one cell transmits downlink while its neighbor receives uplink, the transmitting gNB can interfere with the receiving gNB. At the same time, the uplink UE can interfere with the neighboring downlink UE. These are the two cross-link interference paths.
Give two cells different slot patterns, change their relative schedule offset, and inspect the interfering transmitter and victim receiver. The two cells use the same 20 MHz channel with one active UE per cell. The geometry and link budget are illustrative, so the simulator explains the mechanism rather than predicting a deployment.
Sections
Transmission Directions
Cell A / B | Interference paths |
|---|---|
DL / DL | gNB A to UE B; gNB B to UE A. Ordinary downlink inter-cell interference. |
UL / UL | UE A to gNB B; UE B to gNB A. Ordinary uplink inter-cell interference. |
DL / UL | gNB A to gNB B (BS-to-BS); UE B to UE A (UE-to-UE). |
UL / DL | gNB B to gNB A (BS-to-BS); UE A to UE B (UE-to-UE). |
G / any | G means the entire slot is muted. That cell has neither a desired link nor an interfering transmitter during this slot. |
Each repeating pattern contains ten 0.5 ms teaching slots. D means downlink, U means uplink, and G means muted. A positive B offset delays its pattern: at global slot k, cell B uses pattern entry (k − offset) modulo 10. Slot boundaries remain aligned; this is a whole-slot schedule shift, not a sub-slot synchronization-error model.
Mathematical Foundation
gNB A is at (0, 0), gNB B at (L, 0), UE A at (aL/100, 30) and UE B at ((1 − b/100)L, 30), in metres. The UE sliders measure progress from each serving gNB toward its neighbor. Association stays fixed even if a UE crosses the midpoint. The scene is schematic; the distance labels use these coordinates.
+ 10n log10(max(d, 1)) dB
Received power
= transmit power
− PL − isolation
The 43.3 dB intercept approximates free-space loss at one metre and 3.5 GHz. The teaching path-loss exponent n is 2.2 for BS-to-BS, 3.5 for UE-to-UE, and 3.2 for all BS-to-UE links in either direction. Isolation is applied only to the selected cross-link type; desired links and ordinary inter-cell interference have zero extra isolation. Powers are total isotropic transmit powers over the same 20 MHz channel. Antenna gains are absorbed into this simplified budget.
+ 10 log10(20 × 106) + NF dBm
SINR = S
− 10 log10(10I/10 + 10N/10) dB
S is desired received power; I is the received interfering power, both in dBm. Noise figures are fixed at 5 dB for a gNB and 9 dB for a UE, giving noise powers of about −96.0 and −92.0 dBm. If the neighbor is muted, interference is zero in linear units and SINR equals SNR. The table shows SNR alongside SINR to expose the interference penalty. A muted receiver has no SINR or scheduled rate.
For example, with 46 dBm at the neighboring gNB and 500 m between gNBs, BS-to-BS path loss is 43.3 + 22 log10(500) = 102.7 dB. The victim gNB receives about −56.7 dBm of interference without extra isolation, almost 39.3 dB above its noise. This does not specify the SINR until the desired UE signal is also included.
= 20 log2(1 + 10SINR/10)
Cycle average
= sum of slot rates / 10
This is a Shannon-based indicator, not measured throughput or an NR MCS/BLER prediction. Cycle averages include muted slots as zero and split DL and UL contributions for each cell. The mismatch percentage counts slots with one D and one U out of all ten slots. Editing a control recomputes the entire repeating cycle; the plot is not a record of past settings.
Simulation
Click a pattern tile to cycle D, U and G. Select a global slot below the patterns to inspect it, or press Run to advance continuously. All quantities update when you edit a control.
D: downlink U: uplink G: muted. Click a tile to change it.
Cyan: cell A desired; green: cell B desired; red dashed: BS-to-BS; purple dashed: UE-to-UE; orange dashed: ordinary interference. Arrows point toward receivers.
Cell A / Cell B. Hollow top markers: no-neighbor SNR. G slots have no marker. Selected slot has a white vertical line.
Usage
Start with Mixed directions and select global slot 4. Cell A is receiving UL while B transmits DL. Follow the red arrow from gNB B to gNB A and the purple arrow from UE A to UE B. Read the victim receiver names in the budget table.
Press Run. Each step is one 0.5 ms model slot; run speed controls wall-clock playback only. Step Forward/Back pauses playback and moves one slot, stopping at time zero when stepping back. Reset pauses, restores Mixed directions and returns to slot zero.
Choose Aligned patterns or Copy A to B + zero offset. Cross-link mismatch becomes zero, but orange same-direction interference remains. Compare the SINR and average rates with the mixed schedule.
Choose Same pattern, delayed B. The editor rows match, but the global rows differ because B starts two slots later. Move the B delay slider and inspect the effective slot pairs.
Choose BS-BS isolation. This repeats the mixed pattern with 40 dB extra BS-to-BS loss. The uplink victim improves in mismatched slots; the UE-to-UE path stays unchanged. Try the separate UE-UE isolation slider.
Move the UEs closer by increasing both progress values toward 50%. Their mutual separation falls to zero at exactly 50/50; the path loss then uses a one-metre minimum. Watch the downlink victim's SINR. The schematic retains separate labels for coincident UEs.
Choose Mute B in mismatch slots. This prevents cross-link interference in those slots but gives B no service there. Compare whole-cycle DL/UL rates, not just the best instantaneous SINR.
Edit individual pattern tiles for your own experiment. Edits and presets remain live while Run is active. Slot inspection pauses playback. Each preset restores the parameter defaults as well as its pattern and returns the cursor to zero.
Parameters
Control | Range and interpretation |
|---|---|
Patterns / B delay | Ten D/U/G entries per cell. Delay 0–9 whole slots rotates B's schedule. G is a full muted slot, not an NR special-slot format. |
gNB separation / UE positions | 100–1500 m; each UE moves 5–95% of that separation toward the neighboring gNB, with a fixed 30 m perpendicular offset. |
Transmit power | Independent gNB powers 10–46 dBm; UE powers −10 to 23 dBm. No automatic uplink power control. Changing power changes both desired power and that transmitter's interference. |
Cross-link isolation | BS-BS 0–60 dB; UE-UE 0–40 dB. Abstract extra coupling loss, representing measures such as shielding or antenna discrimination. Applied in both directions for that link type. |
Run speed | 1–100 slots per real second. The model slot duration stays 0.5 ms. At high speeds the display may skip intermediate visual frames. |
Limitations
This is a deterministic, co-channel, fully overlapping, full-buffer two-cell model. There is no fading, shadowing, antenna pattern, beam scheduling, traffic adaptation, power control, HARQ, MCS, adjacent-channel leakage, receiver blocking or nonlinear saturation. Link powers add incoherently. Very strong interference remains a linear SINR calculation.
The fixed path-loss exponents are teaching choices, not calibrated 3GPP propagation models. Full-slot D/U/G patterns do not validate NR slot configurations or switching gaps. Propagation delays, timing advance, partial-slot overlap and over-the-air time synchronization are omitted. The ideal rate excludes overhead and implementation loss and should not be interpreted as achievable NR throughput. Pattern alignment illustrates one interference-management mechanism, not a complete coordination procedure.
References
Ericsson: Managing cross-link interference in TDD networks. Background on cross-link interference and deployment constraints.
Ericsson Technology Review: 5G synchronization requirements and solutions. Background on TDD time-domain isolation and synchronization. The equations and example values here form a separate illustrative model.