BBC Radio 5 Live Data Recovery

The above tweet is about the corrupted audio which the radio producer had extracted by his own data recovery attempt. It is not a sample rate problem but instead relates as to how the specific audio recorder stores data to the SD card. The portable audio recorder in use was a Zoom H4n audio recorder. This has the capability to record up to four audio streams simultaneously. It can be used with personal radio microphones so users can move freely whilst their individual voices are recorded. In the production edit, the separate audio files are played back simultaneously so that each person has the same level of volume.

The data loss was reported due to lack of battery power and the audio recorder suddenly losing power. As a consequence of this, the File Allocation Table (FAT) had become corrupt. On inspection of the SD card, it was clear that there were no folder or file name entries associated with the required data. Full sector-by-sector scans could not locate any reference to the required files in the FAT. The FAT defines the folder and filename structure, but also (critically) defines the location of files by LBA values and deals with any splitting of files if these are not written in sequential blocks in the SD card. Without the FAT it was necessary to perform a 'RAW' recovery and to locate files by file type (e.g. .WAV, .MP3 etc).

The garbled high-speed audio which can be heard on the radio show is because the recovery attempt made by the show's producer failed to take into account that the three audio streams were interleaved within the raw data. Whilst their 'RAW' recovery found the start of each file it could not take into account the method in which data was written to the SD card by the Zoom H4n audio recorder. Below are examples of working audio files from the Zoom audio recorder, which show this interleave between files.

To obtain normal audio playback it was necessary to take each of the "RAW" .WAV files in the recovery, and then write a script to separate the three streams of audio. Essentially the first 64 sectors of data were extracted to create audio 'stream 1', the following 64 sectors of data to create 'stream 2', and the third 64 sectors to create 'stream 3'. This process was repeated over and over again until each of the three audio streams were fully assembled and would playback normally.

In the radio show, you can hear how the producer tried to "Wake up" the SD card by copying some additional data to the SD card after the data loss. The golden rule of data loss/recovery applies here: if you lose data turn the device off and do not attempt to use it. The audio recordings they wished to recover had not been marked as 'allocated' within the FAT. Consequently, when new data was copied to the SD card it wrote to areas of the SD card which contained the required data. Once data has been overwritten with new data there is no method to reverse this. As a consequence, it was only possible to recover the first 15 minutes of audio associated with the third podcast recording.

Until February 2021 you can listen to the full BBC Radio 5 Live Elis James and John Robins episode from the BBC's website.