Topic: PMD 661 improper shutdown leaves file damaged; samples, screenshot, incl.

[Massive Dynamic]

I had battery failure while taping with my PMD 661, but I have a 1.31GB file saved to the card. I was able to import the file as raw data in Audacity, and the wave form looks right, but the sound is masked by a lot of hiss. I am not sure if the file can be salvaged, or if it is irreparable. Here is a link to a 1 minute sample.
http://www.sendspace.com/file/kw8kvc

Attachments don't seem to be working, so here is a link to a 124KB screen shot.
http://www.sendspace.com/file/hcwdpe

There are six options in Audacity when importing raw. Here is what I chose:

Selected 24 bit PCM
Big-endian
2 Channels (Stereo)

Start offset: 0 bytes
Amount to import: 100%
Sample rate: 44100

I have tried changing the offset value, but all other tries have made the sound worse. Anyone have an idea if this recording could be salvaged?

EDIT:
I tried to open the file in Wave Editor a couple of times, and its First Aid function had a couple of identified issues:
"The IFF size is smaller than the content of the file. The number of bytes described in the IFF size (910) is smaller than the size of the file. This means there are unused bytes in this file, which can pose a security risk." WE couldn't seem to fix this last night, so I gave up.

"A chunk was found with an invalid name. Chunk names can only contain printable characters. One was found with the name 'õ“ˇŒ'. This usually means the chunk previous to this has an invalid length." WE found this issue today and was able to fix this file (I'm listening to it right now). Thumbs up to WE, though I am still curious if using Audacity to import raw data could work on other files.

[live2496]

Hello Clive,
I got the same result here and I think that there has been some error when converting the file to float format. (Audacity uses float internally). I did try different offsets in Samplitude and experimented with both float and 24-bit possibilities.

In reviewing your notes I think that it is a good bet that the source file is little-endian, and if so then reimporting the original file and specifying little-endian in the dialog should correct it. I think you may be getting the noise because the sample bytes are in the wrong order.

Gordon

edit: in viewing the file in UltraEdit it appears that the data chunk starts at byte 84.  -gg