Перейти к содержанию

Block 11 Hardware Bring-up Summary

This page keeps detailed Block 11 hardware notes separate from the top-level course status matrix.

Purpose

Block 11 is the integrated SDR project track. Its goal is to connect the reference model, fixed-point and HDL handoff, board-side implementation, capture evidence, metrics and final report.

Current evidence map

Evidence Current state Notes
Baseline board path reviewed The vendor Linux/IIO path is the known-good reference.
External monitor flow measured WAV/IQ monitor capture and offline replay are documented.
Connected RTL-SDR witness measured 868.3 MHz stock DDS tone measured at +201.965 kHz, SNR 40.30 dB, no clipping.
Fallback reference path reviewed The fallback path has a zero-error reference result.
Runtime overlay control path reviewed Register readback and re-init helpers show that the overlay is alive after reload.
Runtime on-chip loopback measured The promoted loopback point reaches zero errors for the current deterministic frame.
QPSK RTL and bridge path executable Canonical simulation reaches BER=0 for 140 symbols / 280 bits through the dual-modem bridge.
Runtime QPSK fabric loopback measured CDC-fixed payload: 4/4 boot sessions and 13/13 attempts at offset 62 reached BER=0; selected ExtraTiming payload: 10/10 attempts reached BER=0 for 140 symbols / 280 bits.
Integrated FPGA implementation measured signoff candidate Hardware-correlated snapshot is fully routed; strategy sweep selected Performance_ExtraTimingOpt with WNS +0.096 ns, TNS 0 and 0 failing endpoints.
External QPSK RTL-SDR path measured cross-session Three independent OTA sessions detect 90/90 bursts with 0/25,200 bit errors, median EVM 19.04%, median CFO +1.969 kHz and no clipping.
Final two-board differential QPSK link reviewed Board A → 30 dB → board B in-fabric RX. Differential coding plus a 24-symbol preamble remove every whole-burst rotation failure (gross 0) at payload BER 4.45×10⁻⁴ over 800 frames; fabric loopback stays at BER < 5.34×10⁻⁷ over 5.6 M bits. The shipped differential image closes timing at WNS +0.036 ns — a separate build from the WNS +0.096 ns canonical strategy sweep above. Lab 11.4 is the final measurement report; Labs 11.41–11.45 carry the investigation.

Status alignment note

The earlier version of this page described an old blocker with high error rate and run-to-run variability. The internal loopback and three 30-burst external sessions are now measured. A 2026-07-08 implementation-strategy sweep improved the canonical routed snapshot to WNS +0.096 ns and the selected payload passed fabric loopback. The integrated modem is now closed on a two-board differential link (see the evidence map above); the remaining work is repeat-build/seed timing robustness, raw-data publication and stable long-run external capture.

Remaining engineering statement

The integrated modem is now closed on a two-board 915 MHz RF link: differential coding plus a 24-symbol preamble give zero whole-burst rotation failures at payload BER 4.45×10⁻⁴, on top of repeated zero-error fabric loopback and a manifest-backed 3/3-session external RTL-SDR baseline. The remaining proofs are optional hardening: repeat-build/seed timing robustness, publication of the measured raw WAV, and longer-duration external statistics.

Next experiments

Priority Experiment Expected evidence
Done Repeat the promoted loopback point from a clean boot 4/4 clean boot sessions and 13/13 attempts on the timing-clean payload
Done Add a controlled cabled measurement variant two-board SMA + 30 dB link with attenuation, gain settings and the Lab 11.4 metric table
P1 Compare baseline and runtime captures with matched monitor settings paired manifests, plots and conclusions
Done Improve timing-closure margin with implementation strategies 6/6 timing-clean runs, selected WNS +0.096 ns and timing-sweep plot/report
P1 Verify repeat-build/seed timing robustness repeat builds or seed sweep retain stable positive WNS
Done Extend QPSK from digital loopback to measured RF OTA manifest, constellation, EVM/SNR and BER summary
Done Measure external QPSK burst repeatability 30/30 detected zero-error bursts with BER/EVM/CFO distributions and confidence bounds
Done Repeat QPSK across independent RF sessions 3/3 sessions, 90/90 bursts, 0/25,200 bits and safe stock restore
P1 Stabilize long RTL-SDR capture strict longer series detects all commanded bursts or uses a more stable capture backend
Done Promote the best result into a final project report Lab 11.4 model-to-measurement report with pass/fail table and limitations

Done criteria

Block 11 should become Portfolio-ready only when it has a clean command path, manifest-backed data, configuration notes, plots, BER/EVM/SNR metrics, a short conclusion and an explicit limitations section.