End-to-end SDR demo roadmap¶

This page defines the flagship demonstration for the course: a complete route from a reference model to RF observation, IQ recording, offline analysis and final measurement reporting.

Demo goal¶

Show that the course is not only theoretical. The learner should see one complete engineering chain:

reference model -> fixed-point constraints -> HDL/FPGA path -> Zynq/AD9363 RF signal -> RTL-SDR/HDSDR observation -> IQ recording -> offline analysis -> measurement report

Target demonstration¶

The recommended first flagship demo is a narrowband tone or simple QPSK signal because it is easy to validate visually and numerically.

Variant	Benefit	Risk
Single tone	Simple FFT peak validation and RF path debugging.	Too simple for final project demonstration.
Two-tone signal	Adds linearity and intermodulation discussion.	Needs careful gain discipline.
QPSK burst	Demonstrates IQ, constellation, EVM and synchronization.	Requires more receiver processing.
OFDM mini-link	Strong final demo for modern SDR.	More complex and should come after QPSK.

Recommended v1: single tone hardware observation¶

Stage 1: Reference model¶

Generate a complex baseband tone in Python, MATLAB or Simulink.

Required outputs:

time-domain waveform plot;
FFT / PSD plot;
expected frequency offset;
sample rate and duration;
scaling assumptions.

Stage 2: Fixed-point preparation¶

Convert the model to an implementation-friendly representation.

Required notes:

sample width;
amplitude scaling;
clipping margin;
quantization error;
expected frequency bin location.

Stage 3: FPGA / board path¶

Prepare the signal for Zynq/AD9363 transmission or replay.

Required notes:

FPGA or software path used;
interpolation / DAC rate assumption;
TX LO frequency;
TX gain;
expected occupied bandwidth.

Stage 4: RF safety and connection¶

Use the RF safety guide before connecting hardware.

Minimum connection record:

TX source:
TX center frequency:
TX gain:
RF path:
Attenuation:
Receiver:
RX gain:
Sample rate:
Bandwidth:
Overload check:

Stage 5: External observation¶

Use RTL-SDR and HDSDR as an independent observation path.

Required screenshots or exported data:

HDSDR spectrum screenshot;
waterfall screenshot when useful;
IQ recording file;
IQ metadata manifest.

Stage 6: Offline analysis¶

Analyze the recorded IQ in at least two environments.

Recommended v1:

Python for reproducible spectrum and peak detection;
MATLAB or Simulink for educational comparison;
optional C++ reader for portfolio-level engineering.

Required metrics:

Metric	Purpose
Peak frequency offset	Checks frequency plan and sample rate.
Noise floor estimate	Checks receiver gain and environment.
Clipping indicator	Checks RF overload.
DC offset	Checks receiver quality and preprocessing.
IQ imbalance estimate	Optional for advanced analysis.

Stage 7: Final report¶

The final report should include:

model parameters;
implementation parameters;
RF path and safety notes;
IQ metadata;
plots before and after capture;
measurement metrics;
discrepancy analysis;
conclusion and next steps.

Demo artifact checklist¶

Artifact	Path suggestion
Reference model script	`blocks/block_11_integrated_sdr_project/scripts/`
Generated reference plots	`docs/assets/`
IQ manifest	`datasets/manifests/`
Analysis script	`tools/` or block-specific `scripts/`
Measurement report	`docs/lab01-tone-rf-iq-analysis.md` or project page
Safety checklist	Linked from `docs/rf-safety.md`

Mermaid system diagram¶

flowchart TB
    MODEL[Reference model\nPython / MATLAB / Simulink]
    FIXED[Fixed-point constraints\nscaling, word length, clipping margin]
    FPGA[FPGA / SoC path\nZynq-7020 control and streaming]
    RF[AD9363 RF frontend\nLO, gain, bandwidth]
    PATH[Controlled RF path\nattenuation / weak OTA]
    RX[RTL-SDR receiver\nindependent observation]
    HDSDR[HDSDR\nspectrum, waterfall, IQ recording]
    META[IQ manifest\nformat, Fs, Fc, checksum]
    ANALYSIS[Offline analysis\nPython / MATLAB / C++ / GNU Radio]
    REPORT[Measurement report\nplots, metrics, uncertainty]

    MODEL --> FIXED --> FPGA --> RF --> PATH --> RX --> HDSDR --> META --> ANALYSIS --> REPORT
    ANALYSIS -. correction .-> MODEL
    ANALYSIS -. tuning .-> RF

V1 acceptance criteria¶

The first end-to-end demo is accepted when:

the reference model and captured signal use the same documented frequency plan;
the recorded IQ file has a manifest;
the FFT peak is detected within the expected tolerance;
the report includes RF safety notes;
the demo can be repeated from documented commands;
at least one generated plot is produced by a script;
the result is linked from the MkDocs navigation.

Future upgrade path¶

Replace the single tone with a QPSK burst.
Add EVM and BER metrics.
Add synchronization impairments and correction.
Add C++ fixed-point analysis for the recorded IQ.
Add Verilog block comparison for a selected DSP stage.
Add a final IEEE-style measurement report page.