diff --git a/MMVII/include/MMVII_SysSurR.h b/MMVII/include/MMVII_SysSurR.h
index 07e79c0a6..2f15d2f25 100755
--- a/MMVII/include/MMVII_SysSurR.h
+++ b/MMVII/include/MMVII_SysSurR.h
@@ -354,7 +354,7 @@ template <class Type> class cInputOutputRSNL
           tStdVect                mWeights;  ///< Weights of eq, size can equal mVals or be 1 (cste) or 0 (all 1.0) 
           tStdVect                mVals;     ///< values of fctr, i.e. residuals
           std::vector<tStdVect>   mDers;     ///< derivate of fctr
-	  size_t                  mNbTmpUk;
+	  size_t                  mNbTmpUk;  ///< number of tmp unknown, computed at init
 
           // use a s converter from tREAL8, "Fake" is used to separate from copy construtcor when Type == tREAL8
 	  cInputOutputRSNL(bool Fake,const cInputOutputRSNL<tREAL8> &);
diff --git a/MMVII/src/Matrix/cResolSysNonLinear.cpp b/MMVII/src/Matrix/cResolSysNonLinear.cpp
index c15ad0f42..73a40c500 100755
--- a/MMVII/src/Matrix/cResolSysNonLinear.cpp
+++ b/MMVII/src/Matrix/cResolSysNonLinear.cpp
@@ -388,9 +388,10 @@ template <class Type> void  cResolSysNonLinear<Type>::AddObservationLinear
 {
      SetPhaseEq(); 
      Type  aNewRHS    = aRHS;
-     cSparseVect<Type> aNewCoeff;
+     cSparseVect<Type> aNewCoeff;  // we make a copy because SubstituteInSparseLinearEquation modify aCoeff
 
 #if (WithNewLinearCstr)
+      //     Coef . X =Rhs     Coef(X-X0) = R  - Coeff.(X0)
       for (const auto & aPair :aCoeff)
       {
           aNewRHS -=  mCurGlobSol(aPair.mInd) * aPair.mVal;