Project 12.4 — Full SDR Measurement Report¶
Goal¶
Prepare a complete measurement-driven SDR report that combines modeling, implementation, RF setup, IQ recording, synchronization and metrics.
Why this project matters¶
The other three projects each exercise one skill. This one tests whether you can hold the whole chain together and report it honestly: a system diagram that matches what you actually built, an RF setup someone else could reproduce, a dataset with provenance, metrics with stated definitions, and a conclusion that follows from the numbers rather than from hope. It is the deliverable a colleague or an employer would actually read.
Required report structure¶
| Section | Content |
|---|---|
| Requirements | goal and success criteria |
| Architecture | system diagram and data formats |
| Implementation | model, fixed-point, RTL or scripts |
| RF setup | frequencies, gains, attenuation |
| Dataset | IQ file, metadata and registry entry |
| Results | figures and metrics |
| Discussion | limitations and error sources |
| Reproducibility | commands and environment |
What each section must contain¶
| Section | Minimum a reviewer expects |
|---|---|
| Requirements | a falsifiable sentence and numeric targets frozen before the measurement |
| Architecture | one diagram; the numeric format at every interface; which parts are hardware and which are software |
| Implementation | which code produced the numbers (repository path and commit); model-versus-implementation agreement stated in numbers |
| RF setup | centre frequencies, sample rate, TX and RX gains, attenuation, cable or antenna path, and the safe-shutdown procedure; for a bench with a transmitter, evidence that it was RF-safe (a cabled attenuated path or a shielded setup) |
| Dataset | provenance, format, checksum, storage location and why the raw file is or is not in Git |
| Results | every figure has axes, units and the run it came from; every ratio comes with the raw counts it was computed from |
| Discussion | at least one limitation and one rejected hypothesis; error sources ranked by how much they could move the result |
| Reproducibility | one command that regenerates or validates the metrics without the original bench |
Minimum deliverables¶
- final report;
- dataset registry entry;
- metadata JSON;
- figure set;
- metrics table;
- reproducibility commands;
- limitations and next steps.
Pass/fail table¶
| Criterion | Target | Measured | Status |
|---|---|---|---|
| Frequency error | |||
| SNR | |||
| EVM | |||
| BER | |||
| Clipping fraction |
Fill the Target column from the frozen requirements before you measure. Status is PASS, FAIL or NOT MEASURED; a criterion you could not measure is reported as such, not omitted and not filled with a plausible number. A run that fails a target is a valid result if the measurement is valid and the cause is localized.
Evidence discipline¶
The course separates three kinds of evidence, and the report must say which one each number is:
| Kind | Meaning | How to label it |
|---|---|---|
| synthetic | produced by a model or a generated dataset | say "synthetic" next to the figure and in the manifest |
| replayed | analysis of a stored capture | give the manifest and checksum |
| measured | acquired live on hardware in this project | give the setup, the settings and the number of repeated runs |
Never present a synthetic or replayed result as a hardware measurement. Repeated runs matter: one lucky BER of 0 over a few thousand bits says little; state how many bits were compared and across how many independent runs (compare Lab 8.7).
Verification gates¶
- G0. Plan: requirements, targets and the pass/fail table skeleton are committed.
- G1. Model: the reference model meets the targets in simulation.
- G2. Implementation: the implementation agrees with the model within a stated tolerance.
- G3. Setup: the RF setup is documented and safe; a sanity capture (a known tone) confirms the frequency axis and the channel order.
- G4. Measurement: the dataset is recorded with metadata and checksum; quality checks (DC, clipping) pass or their failure is explained.
- G5. Analysis: metrics are computed with stated definitions and filled into the table.
- G6. Reproduction: a clean checkout validates the metrics from the stored artifacts.
Reference implementation (Block 11)¶
Lab 11.4: Final Measurement Report is a completed instance of exactly this deliverable: a full, hardware-validated SDR measurement report for the two-board QPSK link (architecture, setup, filled pass/fail table, honest limitations, reproducibility). Use it as a worked exemplar of the structure and the standard of evidence (the figures, metric units, attached metadata and reproducible commands) not as the project you must copy.
Final acceptance checklist¶
- Targets were written down before the measurement.
- Every section of the required structure is present.
- Every number is labelled synthetic, replayed or measured.
- Every ratio can be recomputed from committed raw counts.
- The RF setup and safe-shutdown procedure are documented.
- The pass/fail table has no empty Status cell; unmeasured items say NOT MEASURED.
- At least one limitation and one rejected hypothesis are stated.
- A clean-checkout command validates the metrics.
- The conclusion says meets or does not meet the frozen criteria.
Write the report from the Block 12 report template and grade it with the final-project grading rubric.
Report conclusion template¶
The full SDR measurement project achieved ______. The key measured results were ______.
The project meets / does not meet the defined success criteria because ______.