We have developed a solution for recording from a very high-resolution digital audio source (192k/24bit) but need a recorder which can actually record the digital stream in a bit-perfect manor. We have looked into a few solutions, such as Audacity, but it appears that many programs process the data at 32-bits and then write the data back at 24-bits, thus corrupting the bit-perfect nature of the files. (Spin-it-Again also does this).

We are looking for a recording solution (software) which can:

Record up to 192k/24bit
Output/Save as FLAC files
Record/Save a bit-perfect digital stream
Split a recording session into multiple files (prefered, though not required)

It is possible to modify the Audacity source to do this, but we are not software engineers, we do the hardware. Bit-perfect is needed in order to preserve the data integrity. Once transfered to our servers the original audio sounds much better than the source due to our hardware capabilities.

Can anyone recommend a software solution for us?

Regards,

Blue Smoke Entertainment Systems

I guess you could look at KRISTAL if you are specifically looking for a free product, as I've had a little better luck with that than with Audacity with avoiding needless 16-bit truncations, but if you seriously believe that quantization error going from 32-bit to 24-bit for recording purposes is audible, you've got another thing coming.

If done correctly, 24-bit integer to 32-bit float and back again to 24-bit integer is lossless (bit-perfect). The 32-bit float number actually has the equivalent of 25 bits of integer resolution.

Of course, one would want to check if the software chain was actually doing everything correctly, but that should be fairly easy to test.

Actually, it IS very audible. Not because there is normally a difference, but because there is a digital signature which has to come through properly or the audio won't play.

This is the issue for bit-perfect.

Almost all recording software can do that. Most of the "work" is done by the soundcard and driver, and the recording software mostly serves as an interface. Of course, you need a 24-bit soundcard. Your choice of recording software generally has no effect on sound quality (assuming the same settings).

Most recording software can do that too. Usually the FLAC encoding is done as a separate step (i.e. Save As after recording). I've seen software that can record directly to MP3, but I don't know if there's any software that can record to directly to FLAC.

I don't know much about this, but it might be called "capture" rather than "recording". And, it would depend on the format and medium. If you have a soundcard with a S/PDIF input, it might have come with a utility to capture a S/PDIF stream.

I think it's actually easier to do a bit-perfect capture. For example, if you are capturing a Dolby AC-3 stream, it's easier to directly capture the (bit-perfect) stream and save it to disk than to decode the AC-3 on-the-fly and save the uncompressed PCM/WAV data.

Any audio editor can do that (including Audacity). Yes, almost all audio editors use 32-bit (or 64-bit) floating point for internal processing. It gives you essentially infinite dynamic range, and from I've read DSP is easier with floating point. But as bryant said, the conversion should not introduce any errors, and certainly not any audible errors.

A quick search turned-up this WAV/MP3 Splitter. Splitting a WAV file should be trivial (for a programmer). You just have to create/write a new header for each new file, and there is no need to "touch" the audio data. I don't know much about the FLAC data structure, but I'd guess that splitting a FLAC isn't too difficult.

That could make things difficult. If you split the WAV (or FLAC) the signature won't match. Depending on how the signature is calculated/generated and depending on how the reference signature is stored/transmitted, the FLAC conversion may corrupt it. (FLAC won't corrupt the audio data.)

Is this DRM? You can't edit a DRM'd file or make a bit-perfect copy (unless you own the key, or unless you "crack" the copy protection.)

If you can open the file with Audacity, and Audacity can play the file, then you should be able to save the file (in any format of your choice) and that file should play (without any tricky signature checking or anything like that).

It is a form of DRM. There is an offset convolved into the signal which corrupts playback if it is not played back from our hardware which performs an opposite operation. If played back on a standard audio output, the audio is corrupted.

If there is a way to get Audacity to "capture" as you say, please let me know how. As far as I can tell it is manipulating the data being sent to it (over a proprietary I2S connection and converting the 24-bit data to 32-bit - thus destroying the data integrity.

What I need to do is to capture the data coming in over the I2S (which is set as a digital record in through the driver, so recording software does work).

Any suggestions?

That's bullshit. I have proven in another thread, that 24 bit int to 32 bit float to 24 bit int conversion is lossless.

You proved it can be lossless - did you prove audacity's implementation is lossless? For all I think foxm01 is trying to say is their attempts with audacity have failed. I don't think he is blaming that on 24fixed-32float-24fixed in general.

I must admit, I haven't tested the latest Audacity versions for 24 bit int to 32 bit float conversion, but the software has shown such a level of maturity over dozens of revisions over the last years that I highly doubt that there should be a problem. It is more probable that there is a driver issue regarding the sound card or chipset.

First, let me respond.

We have tried Audacity - unless there are some settings I am unfamiliar with, it doesn't work.

We have tried "Spin-it-Again". Interestingly, the temp files it creates are perfect. However, it then reads them into it's engine (32-bit) and it spits out files that do not work. Even if set perfectly.

What we are looking for is a recorder that will simply look at an input and write the contents to multiple files (allow splitting). At best, writing to FLAC, however, writing to WAV and us having to convert to FLAC would be a second best choise. I find it hard to believe, but we have been unable to find a single program which does not alter the file in some significant fashion as to render it unusable.

As for 32-bit recreating 24-bit perfectly, perhaps in theory. However, we can accurately test this and have been unable to find a program which can do it.

However, looking for suggestions of things to try. 192k/24bit - perfect recording?

A 24-bit/192kHz PCM WAV file opened in Audition 3.01 (implicit 32 bit float conversion) and saved again as 24 bit integer WAV. The result is bit identical:



I repeat, I seriously doubt, that the problem is the application. Maybe you haven't unchecked to apply dither when saving.

We will check. Have not looked at Audition (Adobe).

However, opening and closing a file is different than capturing the data correctly via the digital input. This it may not do correctly.

Blue Smoke Entertainment Systems

I don't use audacity much but I do seem to remember that one time when I did it applied dither under some circumstances when I wasn't expecting it to. I'm pretty sure I had to turn it off in the preferences.

LOL, I just realized that you are talking about Audacity not Audition. Sorry, sloppy reading on my side. Just saw that I even wrote "Audacity", but I was meaning Audition the whole time. Sometimes the brain works funny. I hate Audacity's ugly interface, do you really want your customers to work with that? Anyway, dithering may be the issue here. Check if there's a buried option to disable it.

If there is some demo version, you can try Sony Soundforge.