Avoiding an overflow in the LTP quantization