Audacity and Vegas wav files

[E King]

I have 2 methods for extracting required audio from an mxf file.

Audacity - open mxf file > make stereo track from channels 1/2, which is the full stereo mix I require > Export as .wav.

Sony Vegas - Render channels 1/2 as a .wav file.

However, I have noticed that there are differences between the two resulting .wav files.

Is there a reason for this that I am missing? The only thing I can think of is that "rendering" the file in Vegas is different to just saving/exporting in audacity? Is it something to be concerned about? I can't really tell any differences when playing the files back.

Here are two pictures:

Audacity: https://images2.imgbox.com/33/a3/PlAmLY5S_o.jpg

Vegas: https://images2.imgbox.com/56/fd/4geHt0WA_o.jpg

At a glance I wouldn't be able to tell the differences between them. I can only tell by opening the first image in windows photos and then pressing the > arrow key to move on to the next one. (Edit: If you have each image open in adjacent browser tabs then clicking between them also highlights the differences).

[johns0]

The difference is if one is saved at 44hz and the other at 48hz.

[E King]

Both are showing up as 48 in Mediainfo. And on those screenshots Audacity is reporting 48 for both. That's after I've saved the files (one via make stereo track > export as 24 bit wav in Audacity, the other by rendering as a wav in Vegas) and then opened them again in Audacity for the purpose of taking a screenshot of the graph.

Media info reports for both:

Audio
Format : PCM
Format settings : Little / Signed
Codec ID : 1
Duration : 29 min 42 s
Bit rate mode : Constant
Bit rate : 2 304 kb/s
Channel(s) : 2 channels
Sampling rate : 48.0 kHz
Bit depth : 24 bits
Stream size : 490 MiB (100%)

Audio
Format : PCM
Format settings : Little / Signed
Codec ID : 1
Duration : 29 min 42 s
Bit rate mode : Constant
Bit rate : 2 304 kb/s
Channel(s) : 2 channels
Sampling rate : 48.0 kHz
Bit depth : 24 bits
Stream size : 490 MiB (100%)

By default Vegas has track fx enabled for audio but I've tried turning that off as well as muting the video before rendering and I still get subtle differences in the two files.

Maybe it's normal for this to happen with different software? Only reason I came across it is because I recently discovered that it's a little bit quicker for me to open the mxf files I'm working with in audacity and save the required audio that way than it is for me to render a wav through vegas. I wanted to make sure the result was the same before I proceeded with this method and the differences have got me a bit stumped.

The mxf files I'm working with have a single audio track containing several audio channels, most of which are just digitally encoded silence, but if this silence is kept alongside Ch1/2 it distorts the sound. So I have to extract the full stereo mix. I then encode this to a Nero AAC file and mux it into an mp4 file produced by Handbrake. (Handbrake can't deal with audio at channel level e.g. choosing which audio channels to encode afaiw)

[Cornucopia]

Both of these methods are Reading (and Interpreting):Rendering:Exporting. Neither would be a bit-for-bit identical copy of the WAV data, because both are interpreting and then re-rendering, esp. based on whatever the settings are for the timeline.
Assuming both truly get rendered as 48k same as source, the main WAV sample data should be equivalent even if the render/export "re-encodes", assuming its uncompressed LPCM. But there could very likely be differences in the LSB (if dithering is added by default in an app, for example), or with the packetizing and header infos, metadata, etc. IOW, the raw LPCM "payload" might match, but how it is stored in the file/container might not.

You want a real indicator of the file, and the differences, do an extract/demux with ffmpeg, which should NOT re-encode (if done properly).

Take that and either copy from before and open as pairs in something like audacity and compare by inverting (polarity switch) one of the files.

Or to find out about the packetizing, headers, metadata, etc, you could just do a bitwise comparison. WinDiff, HexEdit, etc.

*************

Note, however, true digital silence would not distort another sound when mixed in, so that other track/channel you have might not be silent. Or could be your mixer is dropping the level of the stereo mix in anticipation of having to mix it with another track (which wouldn't be necessary with a true digital silence of all zeros).

*************

<edit>I found this tool which should be quite helpful in comparing the real content...https://sourceforge.net/projects/fourierrocks/.</edit>

Scott

[johns0]

I thought the op was talking about files size,different encoders give different results,kinda like taking a different route to get to the same spot.

[E King]

Both of these methods are Reading (and Interpreting):Rendering:Exporting. Neither would be a bit-for-bit identical copy of the WAV data, because both are interpreting and then re-rendering, esp. based on whatever the settings are for the timeline.
Assuming both truly get rendered as 48k same as source, the main WAV sample data should be equivalent even if the render/export "re-encodes", assuming its uncompressed LPCM. But there could very likely be differences in the LSB (if dithering is added by default in an app, for example), or with the packetizing and header infos, metadata, etc. IOW, the raw LPCM "payload" might match, but how it is stored in the file/container might not.

You want a real indicator of the file, and the differences, do an extract/demux with ffmpeg, which should NOT re-encode (if done properly).

Take that and either copy from before and open as pairs in something like audacity and compare by inverting (polarity switch) one of the files.

Or to find out about the packetizing, headers, metadata, etc, you could just do a bitwise comparison. WinDiff, HexEdit, etc.

*************

Note, however, true digital silence would not distort another sound when mixed in, so that other track/channel you have might not be silent. Or could be your mixer is dropping the level of the stereo mix in anticipation of having to mix it with another track (which wouldn't be necessary with a true digital silence of all zeros).

*************

<edit>I found this tool which should be quite helpful in comparing the real content...https://sourceforge.net/projects/fourierrocks/.</edit>

Scott

Your explanation for the files not being identical (in terms of the visual in Audacity) does make sense. My main concern was a potential quality difference but I've made an mp4 with both versions (encoded to AAC) and I can't really pick up anything switching between the two audio tracks on my TV.

Thanks for the link to the fourierrocks software. It seems it only supports up to 16 bit so I haven't been able to use it thus far, but I will make some 16 bit versions of the wavs to try.

Re the distortion. Some of the channels are not completely silent. There's also an AD track on Ch 3 sometimes or Ch 3/4 is the dialogue only track without background music and effects etc. So encoding with all of the channels isn't really an option. I'm surprised that neither Handbrake or VidCoder has the ability to just encode with channels 1/2. I wonder if this will ever be incorporated?

[johns0]

Vidcoder does have the option to encode right channel only,left channel only.

[ProWo]

I have 2 methods for extracting required audio from an mxf file.

Audacity - open mxf file > make stereo track from channels 1/2, which is the full stereo mix I require > Export as .wav.

Sony Vegas - Render channels 1/2 as a .wav file.

You can use also clever ffmpeg-GUI for extracting audio from mxf files.
The wav audio is copied as is, without rendering/recoding.

[E King]

I have 2 methods for extracting required audio from an mxf file.

Audacity - open mxf file > make stereo track from channels 1/2, which is the full stereo mix I require > Export as .wav.

Sony Vegas - Render channels 1/2 as a .wav file.

You can use also clever ffmpeg-GUI for extracting audio from mxf files.
The wav audio is copied as is, without rendering/recoding.

Thanks.

I'll give that one a go.

[E King]

Vidcoder does have the option to encode right channel only,left channel only.

Does it have the ability to encode Ch1/2 as a stereo track e.g. 1 being left and 2 being the right? I haven't found that option if it does.