Jaguar3 TX beamforming apply — closed-loop steering (experimental, self-gating)#204
Open
josephnef wants to merge 1 commit into
Open
Jaguar3 TX beamforming apply — closed-loop steering (experimental, self-gating)#204josephnef wants to merge 1 commit into
josephnef wants to merge 1 commit into
Conversation
…lf-gating) Completes the TX-beamforming-in-monitor-mode loop on Jaguar3 (the #203 follow-through): after devourer sounds a peer, steer injected TX toward it via the hardware-built V matrix for a free link-budget gain (no PA power). The WMAC builds V from the peer's Compressed Beamforming Report autonomously — software only configures the entry and flips the apply toggle (no linear algebra). - src/BeamformingSounder.h: bf::arm_beamformer_entry (the beamformer-entry regs arm_sounder doesn't write — BFMER0_INFO, TX_CSI_RPT_PARAM dims, BBPSF_CTRL, CSI_RRSR) + bf::apply_vmatrix (TXBF_CTRL[11:9] per BW + DIS_NDP_BFEN). Ported byte-faithfully from reference/rtl88x2cu/hal/rtl8822c/rtl8822c_bf_monitor.c. - DEVOURER_TX_NDPA=N: periodic NDPA (N>1 = every Nth frame; 1 = every frame, the original self-sounding capture, unchanged). Periodic sounding measured to scale cleanly (period 1/8/64 -> 16363/6556/831 CBRs). Periodic NDPA frames are forced to VHT-2SS (a 1SS NDP yields no CBR — measured) while the interleaved data keeps its 1SS/HT radiotap rate (what gets steered). - DEVOURER_BF_TXBF=<peer_mac>: configure the beamformer entry at InitWrite, then enable the apply toggle from the RX loop ONLY after ingesting a CBR from the peer (a blind apply with no V degrades the link — measured -7 dB). Self-gating and off by default: inert until a genuine peer CBR arrives. VALIDATION STATUS (honest): the register recipe is byte-exact to the vendor (incl. the 8822E RF-table), the sub-components are individually validated (sounding, beamformee, RF-mode, CBR capture — 50k reports/20s), periodic NDPA + rate-split are validated on-air, and the apply toggle demonstrably affects TX. The end-to-end steering GAIN is NOT yet measured: on the available bench the beamformer never ingested a CBR in the closed-loop config — dominated by USB contention (both Jaguar3 adapters on one bus), flaky multi-chip init, and the 8822E TX+RX RX-desense (0x41e8) — so a positive RSSI A/B could not be produced here. Shipped as an experimental, off-by-default, self-gating knob (inert without a captured CBR, so safe); tests/txbf_apply_onair.sh is the A/B harness for a bench that can run two Jaguar3 chips concurrently without contention. Default behavior unchanged (DEVOURER_TX_NDPA unset + DEVOURER_BF_TXBF unset = byte-identical). Full build + ctest (10/10) green. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Completes the TX-beamforming-in-monitor-mode loop for Jaguar3 — the exact remaining work from #203 (rate-split + CBR-gating + beamformer-entry config). Closes #203.
What
The WMAC builds the steering V matrix from the peer's Compressed Beamforming Report autonomously (no software linear algebra —
rtl8822c_bf_monitor.c); devourer only configures the entry and flips the apply toggle.bf::arm_beamformer_entry— the beamformer-entry regsarm_sounderdoesn't write:BFMER0_INFO(0x06E4),TX_CSI_RPT_PARAMdims (0x06F4),BBPSF_CTRL(0x06DC),CSI_RRSR(0x1678). Byte-faithful vendor port.bf::apply_vmatrix— the toggle:TXBF_CTRL 0x042C[11:9]per BW +DIS_NDP_BFEN.DEVOURER_TX_NDPA=N— periodic NDPA (N>1= every Nth frame;1= every frame, unchanged). Measured to scale cleanly (period 1/8/64 → 16363/6556/831 CBRs). Periodic NDPA frames forced to VHT-2SS (a 1SS NDP yields no CBR — measured); interleaved data keeps its 1SS/HT rate (what gets steered).DEVOURER_BF_TXBF=<peer_mac>— configure the entry at InitWrite, enable the apply toggle from the RX loop only after ingesting a CBR from the peer. A blind apply with no V degrades the link (measured −7 dB), so this is the load-bearing gate. Off by default, self-gating (inert until a real CBR arrives).Validation status — honest
The end-to-end gain A/B could not be produced on the available bench: the beamformer never ingested a CBR in the closed-loop config — dominated by USB contention (both Jaguar3 adapters on one bus), flaky multi-chip init, and the 8822E TX+RX RX-desense (
0x41e8). This is a bench limitation, not a code defect — reproducing the gain needs two Jaguar3 chips on separate USB controllers running concurrently.Shipped as an experimental, off-by-default, self-gating knob: inert without a captured CBR (so it can't degrade a link like the blind version), with
tests/txbf_apply_onair.shas the A/B harness for a suitable bench. Default behavior byte-identical (both envs unset). Full build +ctest(10/10) green.🤖 Generated with Claude Code