Skip to main content

Hi all.

I recorded a live show the other night, with three acts. The last two went smoothly, but unfortunately towards the end of the opening act the guy shooting the video managed to trip over my power cable and pull it out of the wall... my laptop DAW just switched to its internal battery of course, but it wasn't very happy with the sudden loss of both the audio interface it was recording from and the external HD it was recording to!

I assumed at the time that the recording was lost forever, however: on examining the contents of the HD I found the wav files for that recording, all of which show a healthy size, so that got me wondering whether it might be possible to rescue it... the files refuse to open in any of my DAWs or audio editors, except if I use Soundforge and open the files as "RAW" audio, but so far I have not been able to find the correct settings to open the files properly: most settings just produce full-scale noise.

However, when I set the samplerate to 96k (it should be 44.1) and the header size to 4 bytes, I do get recognizable audio. But its at the wrong gain setting (its a lot louder than the original recordings, of a poet speaking into a mic with the gain set for a singer) and is shot through with regular spikes of noise. And of course, it plays at slightly over twice the correct speed!

So, can anyone suggest the correct settings to recover my original audio files? They should be 24 bit 44.1 KHz mono.

The options are as follows:
Samplerate: I set 44.1KHz
Sample type: I chose 24 bit PCM
(Format: unsigned, signed or sign bit, greyed out with 24 bit setting.)
Byte order: I chose Little-endian (Intel) rather than Big-endian (Motorola)
Channels: my files are mono.
Header (0 to 536,870,912 bytes): I chose 4 bytes, on the basis that afaik wav files have a 32 bit header... could this be where I'm going wrong?
Trailer (0 to 536,870,912 bytes): Haven't a clue, so far I have left this at 0 on the basis that my DAW (Tracktion 3) never finished writing them anyway...


Topic Tags


RemyRAD Tue, 05/26/2009 - 02:42

On your hard disk drive, do these corrupted files have a ".wav" suffix? Or do they just show as generic files without a suffix? If that's the case, I would just try adding the ".wav" suffix to those files. They should be .wav files anyhow. 24 bit or 32-bit float? Perhaps it's 16-bit float? Playing too fast would indicate sample rate. But it can also indicate 2 multiplexed channels on a single stereo bitstream? Many analog to digital converters handled channels as pairs even if you are recording mono tracks. So even when overdubbing there is a write before read going on to the adjacent channel not being overdubbed. A scary thought I know but this is carried on with many different devices like DA 88's & HD 24's from what I recently read.

I know this was not recorded with a Alesis HD 24 by your description but you just might want to try utilizing "HD 24 Tools"? A handy program that originally sprung from Linux. It has the ability to create new headers for corrupted audio files. Not sure it would do anything for you in this instance but it's a free download. Something else to try.

We've all had this happen. People tripping on cords & acid, simultaneously.
Ms. Remy Ann David

IIRs Tue, 05/26/2009 - 03:20

Thanks Remy.

Yes the files have a .wav suffix. The only sign there is a problem with them is that their file sizes are all very slightly different: had I pressed stop on the recording they would have been identical lengths and sizes.

I've been using (and beta testing) Tracktion for years, so I'm pretty confident the files were being recorded with the settings I specified: 24/44.1/mono.

Downloading HD24 Tools now...

IIRs Tue, 05/26/2009 - 03:26

BTW, playing at the wrong speed is not a problem. It only happened when I told Soundforge to treat the 'raw' files as 96KHz audio: I mentioned it because that's the closest I got to recovering the audio! If the resulting file had not also had gain issues and dropouts I could have simply reset the SR back to 44.1 in SF and patted myself on the back...

Codemonkey Tue, 06/09/2009 - 18:47

Well, I had something like this happen on Sunday.
The PC died about 40 minutes in and now I have myself a 492MB file, and a 170MB file starting a few minutes later when the PC was running again.

Trouble is the 492MB file appears to my DAW as 0.5 seconds of nothing, and to my music playing programs as 2 seconds of proper audio, then 45 minutes of silence.

Currently I'm actually reading up on wav headers, I'm probably gonna have to find/make a tool to edit it.

Codemonkey Tue, 06/09/2009 - 19:24

Update: I've just looked at the contents with a hex editor.

Apparently, it's not glitching after 2 seconds and playing silence, but rather it has an array of data and then IS in fact nothing but 0000000000000000 all the way.

THAT is a strange one, I would've assumed that it would dump data to disk for every few MB recieved, and it ran for 45 minutes between the start of the silence in the audio track, and the point where the PC died.

*still no idea why it died*

In the meantime, OP you could email me this file of yours, and I'll fix your header :P.

Codemonkey Tue, 06/09/2009 - 19:57

Well, since Kristal is a bag of crap I use a VST Plugin (I know, no plugins while recording, but this one is FOR recording).
It's called TapeIt, it offers me stereo-level-monitoring and free choice of where I put the file (due to my convoluted input setup I don't have this from Kristal or the mixer). You set a file, and it records to that file, dumping whatever is passed to the input.

However, due to the screwup, I brought the drive home with me to defrag + sort it out, so I'll check the temp folder right now.

Nope, nothing. Good point though.

IIRs Wed, 06/10/2009 - 06:21

I don't know anyone with Wavelab. Everyone I know uses macs... do you know if Wavelab "Essential" would do? Even that would be more than I want to spend on this project, but it would be good to know for future reference...

It still seems to me like I should be able to do it with Soundforge, if I could only work out what the proper "raw" settings should be!

Codemonkey Wed, 06/10/2009 - 13:33

Well from what I understand, the header (resolution, sample rate, etc.) is written first, the data-length is left blank, and then as it's recorded, the data is dumped. When you finish, it writes the length.

This tool: [=""]XVI32[/]="http://www.chmaas.h…"]XVI32[/] is a pretty solid (and fast) hex editor.
This page: [[url=http://="http://technology.n…"]Wav Format[/]="http://technology.n…"]Wav Format[/] helped me navigate the header.

If you can't make sense of it, you could open the file, copy the first 100 characters or so, and PM me, i'll see if I can make sense of it.

IIRs Fri, 06/12/2009 - 02:01

that worked!!

nice one Codemonkey, looks like I owe you a drink :-)

Of course this project wasn't worth the effort that required, but its good to know this stuff for future reference... eg; I had another problem with recordings from the same night that I didn't bother to mention on the thread as I had already resolved it myself: I had actually been quite impressed that having restored power after the power-cut I returned to my seat and found Tracktion had rediscovered both the HD and the interface, and was ready to record again... I had been half expecting to need a re-boot, or at least to have to restart the audio device... however as the night progressed it became apparent that the system hadn't recovered quite as gracefully as it appeared, as the headline act's ~1hr set read as over two hours on Tracktion's timeline: basically my audio interface had reset itself to 96KHz while Tracktion thought it was still running at 44.1KHz. Result: a collection of huge files that play back at slightly less than half speed...

My solution was to set up a batch process in Soundforge, using the resample algo with the 'only set the sample rate' option ticked. This mostly worked, except that Soundforge won't work with wav files larger than 2GB so I had to go back to Tracktion and convert all the stereo tracks to dual mono and try again. Quite a lot of hassle, and a lot of duplicated data, when all I really needed to do was to edit the bits of the files that specified the samplerate...!