Trouble implementing BFGS

Hello,

I am currently working on a system where I have to find the best signal other a certain time. In this particular problem, the variables to optimize are the amplitudes of the signal for each time. I have implemented the gradient and conjugate gradient method with success but BFGS doesn't seems to work. I use the following code just for the "direction problem": (I need to MAXIMIZE my function that is why the Hessian is negative)

%Define DIRECTION of improvement

HesLength = numel(obj.newGradient);

%Direction for first iteration
if(obj.oldGradient == 0)
  obj.newDirection = obj.newGradient;
  if(strcmp(type,'BFGS')) 
      obj.oldHessianInv = -eye(HesLength);
  end
else                
switch(type)
  case 'Gradient'
      obj.newDirection = obj.newGradient;

    case 'Conjugate'
        %Polak-Ribiere method, auto direction reset
        Dgrad = obj.newGradient - obj.oldGradient;

        beta = (obj.newGradient(:).'*Dgrad(:))/(obj.oldGradient(:).'*obj.oldGradient(:));
        beta = max(beta,0);
        %beta must be positive, if negative than direction becomes gradient

        obj.newDirection = beta*obj.oldDirection + obj.newGradient;

    case 'BFGS'
        Dgrad = obj.newGradient - obj.oldGradient;
        divGrad = obj.deltaPulse(:).'*Dgrad(:);

        if(divGrad < 0) %Definite negative condition
            QHes = eye(HesLength) - (Dgrad(:)*obj.deltaPulse(:).')/divGrad;
            obj.newHessianInv = QHes.'*obj.oldHessianInv*QHes + (obj.deltaPulse(:)*obj.deltaPulse(:).')/divGrad;

            obj.newDirection(:) = -obj.newHessianInv*obj.newGradient(:);

            obj.oldHessianInv = obj.newHessianInv;

        else %If divide by zero use simple gradient
            obj.newDirection = obj.newGradient;
            obj.oldHessianInv = -eye(HesLength); 
        end

        %direction tends to increase close to convergence.
        %When equal inf, convergence is achieved
        if(isinf(max(obj.newDirection(:))))
            obj.oldDirection = 0;
            obj.newDirection = 0;
        end

    otherwise
        ERROR('METHOD NAME IS UNKNOWN');
end

Is there anything wrong for this subpart of the optimization? Also, later in the algorithm the direction is constrained between [-max,+max].

Thanks