A new coding scheme called radix conversion variable length code (RCVLC) is proposed to represent non-negative integers from small values to very large values. RCVLC obtains lowest code length to represent mid-range and large-range integers than Fibonacci code, Elias codes and other existing schemes. This RCVLC is applied as an entropy coder of the Burrows-Wheeler Transform data compressor to compress Calgary corpus which contains both binary and text files. It is observed from the experimental results that the new RCVLC obtains better compression than the existing coding schemes on an average.