WaveLab 24bit Multichannel DVD-A Problem

July 02, 2010

This note describes a problem using WaveLab 6.1 to author DVD-Audio content with 24 bit resolution and with multichannel data. The problem is the incorrect values in the DVD-Audio content files (*.AOB) in the low order byte of the 24 bit samples. Low order bytes are incorrectly arranged in some of the samples, with interchannel swapping with some bytes in fact being missing completely.

To demonstrate the nature of the problem, multichannel (5.1) test 24bit/48kHz LPCM wave files were synthesized with pure 2 kHz sine wave content with peak digital amplitude of 127 (~ -96dBFS). Therefore the sample data only creates content in the lowest order byte with the two higher-order bytes of all samples assigned zero value. Six wave files were synthesized, each with the 2 kHz sine wave content occurring only in one of the 6 channels: FL, FR, C, Lfe, SL, SR. The six tracks, each with content in a different channel, were then rendered to a DVD-A structure in WaveLab 6 using the multichannel option setting for cga of 20 (LF, Rf, Ls, Rs / C, LFE). (The problems discussed here occur with any configuration option of WaveLab).

Two other DVD-A authoring approaches were also used to show the nature of the problem. Extraction of the wave files from the authored DVD-A structure confirms the problem with WaveLab.

WaveLab results

The first 6 spectral displays below show the FL/FR 2-channel output signal from a high-quality 7.1 surround sound card for the 6 different tracks being played back. The spectral content is analyzed using RMAA in record mode. If the original wave content is properly represented in the DVD-A structure, one would expect a spectrally pure 2 kHz signal in monitoring the FL/FR channels only when tracks with content in the FL or FR channels are being played back. As can be seen, this is not the case. Detailed inspection of other synthesized test pattern sample data show that in fact some low-order bytes of the samples are missing which leads to harmonic distortion content as is clearly seen.

Next, playback of the first track, which contains only content in the FL channel is shown below with WaveLab Level Meter, followed by the corresponding EXTRACTED wave file using a popular DVD-A wave file extraction program. As can be clearly seen, the input wave used for authoring has content only in the FL channel at the -96 dB level, as expected, but the extracted wave from the DVD-A structure contains spurious content in the SL channel. Futhermore, note that the peak and RMS meter levels are not equal, implying a non-pure sine wave, a sure sign of incorrect data sample packing. Similar results are observed for extracted waves corresponding to the other tracks. Monitoring of other OUTPUT channels also shows similar problems.

dvda-author 9.05 results

For comparison purposes, the same six wave files were used to author a DVD-A structure using dvda-author 9.05. The results below, again measuring the output of the FL/FR channels of the same sound card show a pure 2 kHz signal, which is only present, when the playback track has content in the FL or FR channel indicating no channel mixing. No significant harmonic content is seen indicating proper storage of the lowest byte of multichannel 24 bit audio data. Similar results were obtained with DW Chrome authoring.

The wave files extracted from this dvda-author 9.05 authored content shows the expected pure 2 kHz signal only in the corresponding channel.

Using other simple patterned sample 5.1 24bit data, the nature of the incorrect packing of the lowest-byte data is shown below for the case of cga = 20. As can be seen, the problem only impacts the lowest order byte of the 24 bit data. The sketch shows the DVD-A pack size of 36 bytes which encompasses two complete samples of all channels. Swapped bytes causes the harmonic distortion which might be audible in listening in very quiet sound passages.