spectrum: (fixes #1314) Use std::conj for RX beamforming in long-term component

Author: Sandra Lagen

CHANGES.md

@@ -63,6 +63,8 @@ fixes, use `./ns3 run "clang-tidy -fix"`.
   Procedure A updates are applied; if it exceeds 1 m, channel parameters are re-generated (new realization).
   For stationary links, the channel remains unchanged. See `spectrum.rst` for ns-3 update triggering and
   1 m step-size handling.
+* (spectrum): Correct long-term combining in the 3GPP channel model to use RX Hermitian combining (conjugate
+  transpose of RX beamforming weights); previously the RX beamforming vector was not conjugated (issue #1314).
 
 ## Changes from ns-3.46 to ns-3.46.1
 

RELEASE_NOTES.md

@@ -62,6 +62,7 @@ been tested on Linux. As of this release, the latest known version to work with
 - (wifi) #1295 - Avoid floating point precision issues in wifi-error-rate-models-test.cc
 - (internet) !2624 - Fix TCP LEDBAT congestion window growth calculation
 - (wifi) #1284 - Fix radiotap for received packets
+- (spectrum) #1314 - Fix missing complex conjugation of RX beamforming vector in 3GPP long-term channel component calculation.
 
 ## Release 3.46.1
 

src/spectrum/model/three-gpp-spectrum-propagation-loss-model.cc

@@ -176,7 +176,7 @@ ThreeGppSpectrumPropagationLossModel::CalculateLongTermComponent(
         auto uIndex = startU;
         for (size_t rIndex = 0; rIndex < uPortElems; rIndex++, uIndex++)
         {
-            rxSum += uW[uIndex - startU] * params->m_channel(uIndex, sIndex, cIndex);
+            rxSum += std::conj(uW[uIndex - startU]) * params->m_channel(uIndex, sIndex, cIndex);
             const auto testV = rIndex % uElemsPerPort;
             if (const auto ptInc = uElemsPerPort - 1; testV == ptInc)
             {

src/spectrum/model/three-gpp-spectrum-propagation-loss-model.h

@@ -192,6 +192,13 @@ class ThreeGppSpectrumPropagationLossModel : public PhasedArraySpectrumPropagati
     /**
      * @brief Computes a longTerm component from a specific port of s device to the
      * specific port of u device and for a specific cluster index
+     *
+     * The beamformed (long-term) effective channel coefficient:
+     * \f$ h_{\rm eff} = \mathbf{w}_{\rm rx}^H \mathbf{H} \mathbf{w}_{\rm tx} \f$.
+     *
+     * @note The RX beamforming vector is stored on the antenna as \f$\mathbf{w}_{\rm rx}\f$.
+     *       This function applies it as a Hermitian inner product (i.e., uses the complex
+     *       conjugate of the RX weights) when forming \f$\mathbf{w}_{\rm rx}^H\f$.
      * @param params The params that include the channel matrix
      * @param sAnt pointer to first antenna
      * @param uAnt uAnt pointer to second antenna