Question regarding conversion from 24 BitWav to 24BitFlac

Hello,

Over the weekend I recorded my Doors 1st album SACD to my computer digitally and brought in a 24/88.2 file into my computer which was about 1.4 gb. I split the songs in audacity and exported them to FLAC. The resulting files ending up being about 462 mb which I initially thought was way to small but didn't have time to analyze it. Now that I am playing that file in foobar I see that the bit rate went way down from 4233 in the wav file to 1429 in the flac file.

I guess my conversion to flac level messed something up, is there a proper way to convert 24bit wav to 24bit flac without losing the bitrate.

Thanks for any help you can provide. I can provide more details if needed.

Thanks
Mike

 

I don't think anything got messed up, that's what lossless compression does. If it was the same size it wouldn't be compression.

 

There's an option to select 24-bit when exporting to FLAC in Audacity (default is 16-bit). Maybe you missed that.

 

I find Audacity quite counter intuitive when it comes to sample rate and bit depth import, export, and conversion. I figured it out, eventually.

 

the best thing you can use to convert anything to flac is the latest build of the command-line FLAC.exe, If you're extra paranoid about the results being exact

 

1429 kbps for a 24-bit 88.2 kHz FLAC file is low. It should be in the range of 2000-3000 kbps. Uncompressed 24-bit 88.2 kHz 2 channel audio is 4233.6 kbps.

It's possible that your FLAC file got processed at 16-bits during the conversion instead of maintaining 24-bits, and that lower 16-bit bitdepth is the reason why the FLAC file was able to compress to a lower than expected bitrate. But it does depend on recording level, the style of music and style of recording, and other factors. But 1429 kbps for a 24/88.2 FLAC file is unexpectedly low.

You can check the file properties in Foobar. Load the file in Foobar. Right-click on the file and select Properties. There will be a Metadata tab and a Properties tab. Select the Properties tab. That will show the sample rate, bits per sample, and the number of channels for the file. Verify that the bits per sample is 24 bits and that there are two channels rather than just one channel. However, even if the bits per sample is 24 bits it's possible that Audacity processed the file at 16-bits and that the other 8 bits are just padding. 16-bit files padded to 24-bits compress better with FLAC than true 24-bit files. And that could also help explain the lower than expected bitrate of the FLAC file.

You could try using Foobar to convert the WAV files to FLAC. Foobar will do the right thing during conversions and will do proper conversions with no funny business behind your back. Audacity can be a bit weird.

 

Thanks for the information, I was really only bringing the file into Audacity in order to split it into individual songs. Are there other pieces of free software that will allow me to split the 24bit wav file into individual songs and export out a 24bit wav file. Normally for conversion I use db music converter and not audacity.

Thanks
Mike

 

Audacity's only real problem is that it can't record beyond 16 bits in Windows unless you compile it yourself with ASIO or find such a version somewhere. If you are recording in another program that is capable of 24-bit recording and use Audacity to split the files it will work fine so long as you select "24 bit" in "Options" when exporting to FLAC as mentioned above by Antares.

If you want to export to WAV rather than FLAC, choose "Other uncompressed files" > "Options" and you'll find what you need.

 

You could play with Reaper, great recorder and editor, and technically can be used for free, but well worth the small price, for editing and cutting I use a really old version of wavelab

 

OK so I ran a couple tests last night. I took audacity conversion out of the equation and only labels the tracks in audacity and exported out a 24bit wav file from Audacity with matched the original file size of the single wav file I was starting with.

I did my wav to flac conversion from db music converter using the uncompressed flac option and the bit rate remained the same at 4233. I then did the same test and used flac level 5 conversion option in db music converter and the bit rate for the same file when way down to 1434. In the properties section of foobar2000 it still says sample rate 88200, channels 2, bits per sample 24, and bitrate 1434.

If this is normal then going forward I guess when converting a SACD disc like this, I'll use the uncompressed flac option to keep the full bitrate of the music being brought in.

 

I converted a 24 bit dsf file that was playing about 5500kbps to flac and it ended up playing at 2750kbps or so.

 

It's a bit confusing but your software is reporting the compressed "bitrate" for the FLAC file not the actual bitrate at playback which will be the same as the original WAV. If you have any doubt load both the original WAV and its uncompressed or compressed FLAC into foobar2000 and use its "Binary Comparator" component to find any differences:

http://www.foobar2000.org/components/view/foo_bitcompare

 

I wonder if you're recording the audio at a high enough level? If you look at the recording in a wave editor and look at the waveform how high does the waveform go? If the waveform never extends up to the top area then the recording level is likely too low. A recording with a low or very low recording level will compress more with FLAC and could result in the low bitrate numbers you're getting when compressed to FLAC.

 

There are different levels of FLAC compression. Your default maybe set to 5. I think 8 is the best.

 

That's interesting. If its all lossless, why not compress it as much as possible?

 

I guess if you want the highest quality you'll just get a bigger file size which is why the OP had a reduction in size of the original WAV

 

Well sure, but if there's "highest quality" and there's "less then highest quality", this implies not-really-lossless to me. See what I'm saying?

 

Back in the old days they would do benchmark tests against all of the various lossy and lossless codecs. Comparing compression ratios against encoding time and decoding time. And make graphs comparing all of them.

In order to be competitive in the benchmarks for best compression ratio would mean doing some more CPU and memory intensive compression algorithms. But those high quality (high quality meaning best compression ratios) algorithms would be slower, sometimes very much slower. So the codecs would implement levels of compression. So they could have a fast compression option that results in a lower compression ratio, and they could have slower compression options that result in better compression ratios. For example, FLAC has compression options from 1 to 8. With 1 being faster and 8 being slower. 1 being a larger file, 8 being the smallest file. All FLAC compression levels are lossless. Some just take more CPU or memory to encode or decode.

For an example of a benchmark style comparison look at this data and graphs from the FLAC developers: FLAC comparison

 

Ahhh, good explanation, thanks. I remember the old SHN format took forever, but I suppose that could have just been the late 90's hardware.

 

It was much more important to be memory and CPU efficient back in the 90s when a Pentium was considered a fast CPU. With an old CPU like that there will be quite a difference in encoding time or decoding time for FLAC level 1 vs. FLAC level 8. With a modern CPU like an i7 that difference shrinks to near insignificance. However, decoding CPU efficiency is still important when considering portable players. Portables don't have fast CPUs, and different compression levels could mean more power use which results in less battery life. For portable use I'd tend to avoid using FLAC level 8.