Blender has 7 repositories available. Follow their code on GitHub. Blender github.
# 2. Audacity
The optional FFmpeg library allows Audacity to import and export a much larger range of audio formats including M4A (AAC), AC3, AMR (narrow band) and WMA and also to import audio from most video files. Because of software patents, Audacity cannot include the FFmpeg software or distribute it from its own websites. Prudys problemthird grade reading streets. Contribute to audacity/audacity development by creating an account on GitHub.
Two filter units in 8 different configurations. Feedback loop available in 7 of those configurations. Available filter algorithms: Lowpass (12 and 24 dB/oct, each with 3 variations), Legacy Ladder Lowpass (6-24 dB/oct), Vintage Ladder Lowpass (2 variations, each with and without gain compensation), Highpass (12 and 24 dB/oct, each with 3 variations), Bandpass (12 and 24 dB/oct.
Sound recording and editing software. The architecture is described as 'like a city', there are some impressive buildings, and some less impressive buildings. It is hindered by licencing issues but gets around these by supporing standard APIs - so said licenced code may be plugged in by a user. Thinking in layers, the base is 3rd party cross platform libs that expose general audio (PortAudio) and GUI (wxWidgets) APIs. These are in seperate DLL files (rather than being compiled into the core file). Size and speed suffer for this, but it allows other DLLs to use them directly at run time - greater extensability.
An abstraction mistake. There's a small story about the original implementation being for mono and not stereo. So the code defaults to one channel, but has been modified to check for 2 channels later on in many places. Better if originally designed for n channels. This issue was also locked in by exposing a GetLink method to get the pair of a channel if it had one, this method has been used throughout and somewhat locked them in. The GUI architecture also has problems - mixes between plugin code, application code, and special cases with absolute positioning etc. They are hoping for a rewrite with greater abstractions - UI components with no required awareness of the greater context.
A drawback of the cross-platform libs: it runs 3 threads, each in their own way. A GUI thread created by wxWidgets and updated every 50ms. An audio thread created by PortAudio for recording and playback. And a buffer thread handled by the application code to keep thins snappy. This creates complexity but it is forced by the expectations of the plugins, had they rolled their own and abstraction could have been made but without all the cross-platform benefits.
Some audio is too big to be handled with speed, or even to fit in RAM. Audacity splits audio into ~1MB 'block files' which are coordinated in an xml based .aup file. Edits only affect the relevant blocks. Block files also hold summary info which is used when zoomed out in place of processing the entire audio track. One issue on windows was lag with more than ~100 files were placed a single directory. So a hierarchy was created with never more than 100 files to a folder.
Audacity has grown organically, no plan was laied down, hence the city like code.
The most applicable thing I've picked up from this is something like 'non-array like arguments make code smell.. a bit', or at least indicate that some thinking / explanation is required. Also, writing cross platform creates restrictions - glad I'm in the web space, going to stick here as much as I can!
This month’s digital recording tool excursion is into the realm of frequency spectrum analysis. As an organist who took lessons early in life on a Hammond B3, played in rock bands using a chopped M2 with percussion added, and now plays services on a model 906, I can’t avoid trying to guess drawbar settings whenever I listen to someone play a Hammond organ.
With the free digital audio tool “Audacity” (Audacity audio software site) I no longer have to guess at drawbar settings. If I can find a relatively clear section of audio in a recording that sustains a single note, Audacity’s “Plot Frequency” analysis tool can tell me exactly what I need to know.
As an example, here is a screen shot of Audacity displaying a section of Jimmy Smith’s recording of “The Cat” (Jimmy Smith, The Cat):
In this case, we have an “F” in the clear for analysis. By opening the “Plot Spectrum” window under the “Analyze” menu, we can visualize the audio frequency spectrum calculated via an FFT:
In addition to numerous options for presentation (here we use the “Log Frequency” mode), the “Frequency Analysis” window has a tool for locking onto peaks and displaying their frequency, equivalent keyboard note, and volume level. Here are the significant peaks for this analysis plot:
346 Hz (F4) = -13.4 dB
696 Hz (F5) = -18.5 dB
1046 Hz (C6) = -18.3 dB
2077 Hz (C7) = -43.4 dB
Now we turn to a reference page on the Hammond site (Introduction to Drawbars) that gives us the spectral layout of drawbars:
With this, and a reference on frequency vs. keyboard notes (Piano key frequencies), we can lay out a quick spreadsheet that gives us the equivalent drawbar frequencies for any given note:
Audacity Github Online
Since each drawbar has a volume range from 0 to 8, we can use a rough estimation of the relative value for each drawbar position:
0dB 8
-3dB 7
-6dB 6
-9dB 5
-12dB 4
…
Audacity Github Plugin
Audacity Source Code
We can now guess that based on the spectrum of this limited section shown above, Jimmy had his drawbars set to something in the range of “87 7020 000”. So if you’re willing to spend a little time in analysis mode, Audacity provides yet another way of learning by listening to the greats.