On Monday, October 14, 2002, at 05:43 PM, pltsai@globalunichip.com wrote:
Is there anybody with the experience to do the compliance test using ver 0.14.2b code ?! I use madplay to decode the compliance bitstream : he_mode.bit My output file is he_mode_intel_24.hex The output file generated by L3dec.exe is he_mode.hex.
The compliance test result : 251137 samples are compared. and the maximum difference :0.489849 the root mean square is 0.0619873
You can find the 11701th sample: the he_mode_intel_24.hex is 123809 the he_mode.hex is f1e918 It has exceeded the maximum difference accoring to the compliance test.
Also in 11702th ....
The result can't meet the full compliance request.
Please note: The only bitstream whose PCM output must match within tolerance to be compliant is the compl.bit stream. The other bitstreams help to detect potential problems, but they are only informative, not normative.
Also I doubt why the file size of output hex file is different with the hex file generated by l3dec.exe
The he_mode.bit bitstream tests the various stereo encoding modes. The reason you find different file sizes is likely because of these changing modes: frames 1-10 and 111-131 are single-channel; frames 11-20 are dual channel; frames 21-110 are stereo.
By default, 'madplay' only outputs the first channel in a dual channel stream, so you are losing the second channel output for comparison. You can force it to be output using -S (--stereo), but then the single-channel frames will have duplicate samples.
If you account for the expected number of channels in each frame, you should find MAD's output to be much closer to that of l3dec.