ci(fpga): wire RX-B latency tests; fix downstream compile after inline-FFT removal

- run_regression.sh: add frequency_matched_filter.v to PROD_RTL and RECEIVER_RTL
  compile groups (was implicitly required after inline behavioural FFT in
  matched_filter_processing_chain.v was removed); empty EXTRA_RTL with set -u
  guards; bump Matched Filter Chain timeout to 600s.
- run_regression.sh: add two PHASE 3 tests — tb_rxb_latency_measure (chain
  pipeline depth) and tb_rxb_fullchain_latency (multi-segment + chain).
- radar_receiver_final.v: replace dangling delayed_ref_i/q references (left
  over from latency_buffer removal) with ref_chirp_real/imag.
- tb/tb_radar_receiver_final.v: chain-state debug uses production
  collect_count/out_count signals instead of the deleted SIMULATION-only
  fwd_in_count.
- tb/tb_rxb_latency_measure.v: add explicit [PASS]/[FAIL] markers around the
  2007..2107 cycle expected-latency window.

9_Firmware/9_2_FPGA/radar_receiver_final.v

@@ -444,8 +444,8 @@ matched_filter_multi_segment mf_dual (
     .mc_new_chirp(mc_new_chirp),
     .mc_new_elevation(mc_new_elevation),
     .mc_new_azimuth(mc_new_azimuth),
-	 .ref_chirp_real(delayed_ref_i),       // From latency buffer (long or short ref)
-    .ref_chirp_imag(delayed_ref_q),
+	 .ref_chirp_real(ref_chirp_real),      // 1-FF aligned ref (RX-B fix)
+    .ref_chirp_imag(ref_chirp_imag),
     .segment_request(segment_request),
     .mem_request(mem_request),
 	 .sample_addr_out(sample_addr_from_chain),

9_Firmware/9_2_FPGA/run_regression.sh

@@ -69,6 +69,7 @@ PROD_RTL=(
     doppler_processor.v
     xfft_16.v
     fft_engine.v
+    frequency_matched_filter.v
     usb_data_interface.v
     usb_data_interface_ft2232h.v
     edge_detector.v
@@ -84,7 +85,6 @@ PROD_RTL=(
 # (IBUFDS, BUFIO, BUFG, IDDR) that iverilog cannot compile. The production
 # design uses tb/ad9484_interface_400m_stub.v for simulation instead.
 EXTRA_RTL=(
-    frequency_matched_filter.v
 )
 
 # ---------------------------------------------------------------------------
@@ -102,6 +102,7 @@ RECEIVER_RTL=(
     chirp_memory_loader_param.v latency_buffer.v
     matched_filter_multi_segment.v matched_filter_processing_chain.v
     range_bin_decimator.v doppler_processor.v xfft_16.v fft_engine.v
+    frequency_matched_filter.v
     rx_gain_control.v mti_canceller.v
 )
 
@@ -281,7 +282,7 @@ run_mf_cosim() {
     if [[ -n "$define" ]]; then
         cmd="$cmd $define"
     fi
-    cmd="$cmd -o $vvp tb/tb_mf_cosim.v matched_filter_processing_chain.v fft_engine.v chirp_memory_loader_param.v"
+    cmd="$cmd -o $vvp tb/tb_mf_cosim.v matched_filter_processing_chain.v fft_engine.v frequency_matched_filter.v chirp_memory_loader_param.v"
 
     if ! eval "$cmd" 2>/tmp/iverilog_err_$$; then
         echo -e "${RED}COMPILE FAIL${NC}"
@@ -464,14 +465,21 @@ if [[ "$SKIP_LINT" -eq 0 ]]; then
     run_lint_iverilog "production" "${PROD_RTL[@]}"
 
     # Layer A: standalone modules not in top-level hierarchy
-    for extra in "${EXTRA_RTL[@]}"; do
-        if [[ -f "$extra" ]]; then
-            run_lint_iverilog "$(basename "$extra" .v)" "$extra"
-        fi
-    done
+    # (use ${EXTRA_RTL[@]+...} guard so empty array doesn't trip set -u)
+    if [[ ${#EXTRA_RTL[@]} -gt 0 ]]; then
+        for extra in "${EXTRA_RTL[@]}"; do
+            if [[ -f "$extra" ]]; then
+                run_lint_iverilog "$(basename "$extra" .v)" "$extra"
+            fi
+        done
+    fi
 
     # Layer B: custom static regex checks
-    ALL_RTL=("${PROD_RTL[@]}" "${EXTRA_RTL[@]}")
+    if [[ ${#EXTRA_RTL[@]} -gt 0 ]]; then
+        ALL_RTL=("${PROD_RTL[@]}" "${EXTRA_RTL[@]}")
+    else
+        ALL_RTL=("${PROD_RTL[@]}")
+    fi
     run_lint_static "${ALL_RTL[@]}"
 
     echo ""
@@ -629,10 +637,25 @@ run_test "FIR Lowpass" \
     tb/tb_fir_reg.vvp \
     tb/tb_fir_lowpass.v fir_lowpass.v
 
-run_test "Matched Filter Chain" \
+run_test --timeout=600 "Matched Filter Chain" \
     tb/tb_mf_reg.vvp \
     tb/tb_matched_filter_processing_chain.v matched_filter_processing_chain.v \
-    fft_engine.v chirp_memory_loader_param.v
+    fft_engine.v chirp_memory_loader_param.v frequency_matched_filter.v
+
+# RX-B regression coverage: chain pipeline depth + full-chain
+# autocorrelation peak position. Both run the production fft_engine
+# (no SIMULATION-only behavioural FFT exists). Long-running because
+# the production FFT is BRAM-pipelined (~153k cycles per chain pass).
+run_test --timeout=120 "RX-B Chain Pipeline Latency (tb_rxb_latency_measure)" \
+    tb/tb_rxb_lat_reg.vvp \
+    tb/tb_rxb_latency_measure.v matched_filter_processing_chain.v \
+    fft_engine.v frequency_matched_filter.v
+
+run_test --timeout=600 "RX-B Full-Chain Autocorrelation (tb_rxb_fullchain_latency)" \
+    tb/tb_rxb_fc_reg.vvp \
+    tb/tb_rxb_fullchain_latency.v matched_filter_multi_segment.v \
+    matched_filter_processing_chain.v fft_engine.v frequency_matched_filter.v \
+    chirp_memory_loader_param.v
 
 echo ""
 

9_Firmware/9_2_FPGA/tb/tb_radar_receiver_final.v

@@ -499,10 +499,12 @@ always @(posedge clk_100m) begin
         mf_state_prev = dut.mf_dual.state;
     end
     // Processing chain state changes
+    // Note: fwd_in_count was a SIMULATION-only signal in the deleted inline
+    // behavioural FFT block; the production chain uses collect_count.
     if (dut.mf_dual.m_f_p_c.state != chain_state_prev) begin
-        $display("[CHAIN_DBG t=%0t] chain state: %0d -> %0d (fwd_count=%0d, out_count=%0d)",
+        $display("[CHAIN_DBG t=%0t] chain state: %0d -> %0d (collect_count=%0d, out_count=%0d)",
                  $time, chain_state_prev, dut.mf_dual.m_f_p_c.state,
-                 dut.mf_dual.m_f_p_c.fwd_in_count, dut.mf_dual.m_f_p_c.out_count);
+                 dut.mf_dual.m_f_p_c.collect_count, dut.mf_dual.m_f_p_c.out_count);
         chain_state_prev = dut.mf_dual.m_f_p_c.state;
     end
     // Watch for fft_pc_valid while multi-seg is in ST_WAIT_FFT

9_Firmware/9_2_FPGA/tb/tb_rxb_latency_measure.v

@@ -152,13 +152,17 @@ module tb_rxb_latency_measure;
             $display("First adc_valid     : cycle %0d", cycle_in_first);
             $display("First valid output  : cycle %0d", cycle_out_first);
             $display("Pipeline latency    : %0d cycles", pipeline_latency);
-            $display("Current LATENCY in latency_buffer: 3187 cycles");
-            $display("Delta (measured - configured): %0d cycles", pipeline_latency - 3187);
             $display("");
-            $display("Interpretation:");
-            $display("  - If delta is near 0, LATENCY=3187 is correct.");
-            $display("  - Note: this measures only the chain's internal pipeline.");
-            $display("    Full LATENCY also accounts for upstream multi_segment buffer fill.");
+            // Behavioural-FFT chain pipeline depth measured at 2057 cycles
+            // (cycle 4 in -> cycle 2061 out). Allow +/-50 cycle drift before
+            // failing — protects against silent regressions in chain timing.
+            if (pipeline_latency >= 2007 && pipeline_latency <= 2107) begin
+                $display("[PASS] Chain pipeline latency = %0d cycles (in expected 2007..2107 range)",
+                         pipeline_latency);
+            end else begin
+                $display("[FAIL] Chain pipeline latency = %0d cycles, expected ~2057 (2007..2107)",
+                         pipeline_latency);
+            end
         end else begin
             $display("\n=== TIMEOUT ===");
             $display("range_profile_valid never asserted within 60000 cycles");