GUI is based on Omnirom DSP Manager and able to run on most Android devices from 5.0 - 8.0 include Samsung, AOSP, Cyanogenmod. This app in order to improve your music experience especially you want realistic bass and more natural clarity. This app don't work too much around with modifying Android framework.
-
Compression
-
Bass Boost
--> 2048/4096 order FIR linear phase bass boost
-
Reverberation
--> GVerb and Progenitor 2
-
FIR Equalizer (Cubic spline interpolated 15 bands equalizer)How smooth it is?
-
Stereo Widen
-
BS2B
-
Partitioned Convolver (Auto segmenting selection)
--> Support mono, stereo, full/true stereo(LL, LR, RL, RR) IR
--> Samples per channels for stereo impulse response should less than 1000000
--> Samples per channels for full stereo impulse response should less than 500000
-
Vacuum tube modelling
A: Float32
A: Effect has 2 options to boost low frequency band, IIR based is obsoleted, equalizer already do the job.
Filtering is work on convolution basis. When user change the bass boost parameter, engine will compute new low pass filter(with gain) impulse response using firls(Function ported from Matlab).
2048/4096 order FIR filter should work on all frequency listed on options.
A: Convolver is a effect apply convolution(a mathematical operation) on audio signal, that perfectly apply user desired impulse response on music, it could simulate physical space.
Effect itself require audio file(.wav/.irs/.flac) to become impulse response source.
For more info: Convolution and Convolution reverb
A: This probably impossible from all algorithms perspective. All algorithms has been precisely measured on time domain, frequency domain, input/output characteristic, etc with mathematics/numerical analysis software.
I put much time on Auto Convolver engine and it's calibration in order to provide accurate result that same(#1) as MATLAB conv(), but not losing any performance.
How about other reasons that bring down quality?
- Your Android audio framework is not supported, this not uncommon on HTC devices.
- Audio amplitude clipped(#2), this is a obvious reason which produce strange sound, try to turn your volume down!
- If you are using convolver, did you try to use smaller normalise(#3) value, which control overall impulse response amplitude level.
- Tube modelling require a high performance CPU to output best quality, so I turn down algorithm precision. **This may produce low amplitude noise.
#1: Floating point rounding issue is not taken into account.
#2: JamesDSP use hard clip, but not any limiter.
#3: This option will be changed to simple amplitude control.
A: Analog Modelling internal work as a vacuum tube amplifier, was designed by ZamAudio. The tube they used to model is 12AX7 double triode. They also provide a final stage of tonestack control, it make sound more rich. However, the major parameters is amplifier preamp, this is how even ordered harmonic come from, but this parameter have been limited at maximum 12.0. Input audio amplitude is decided by user, thus louder volume will generate more harmonics and internal amplifier will tend to clip the audio. Analog amplifier was built from real mathematical model, most notably is nonlinearity of vacuum tube. Original is written in C++, for some reasons I ported it to C implementation.
File reverbHeap is modified Progenitor 2 reverb, memory allocation is using heap not stack, it will be useful when you play around with Visual Studio, because VS have 1Mb stack allocation limit...
A: Because of fftw3 library linked.
A: Audio effects actually is not hard to implement, I don't think close source is a good idea. Many audio effects is exist in the form of libraries, or even in thesis, everyone can implement it... All files in this repository is published under GPLv2.
A: Yes. It is relatively easy to use some effect in other applications. Convolver, reverb, 12AX7 tube amplifier source code is written in similar style, you can look at the how their function is called, initialised, processed, etc. Most of the effect is written in C, so it is easy to port to other platforms without huge modifications.
A: Highest compatibility method(Manual installation)
Effect may get unloaded by Android system if no audio stream for while.
audio_effects.conf is a file specified for system to load effect using known UUID.
- you need to add
jdsp {
path /system/lib/soundfx/libjamesdsp.so
}
bundle {
path /system/lib/soundfx/libbundlewrapper.so
}
jamesdsp {
library jdsp
uuid f27317f4-c984-4de6-9a90-545759495bf2
}
effects {
- copy libjamesdsp.so to /system/lib/soundfx
- Reboot
- Install APK
B: Automatic installation(Only support up to Android 7.1.1)
- Copy zip package to your phone, reboot to TWRP, install package, reboot, done.
- Upmixed 7.1 Spatializer (Non open source)
- Support automatic Android 8.0 installation.
Now work on AOSP, Cyanogenmod, Samsung on Android 5.0, 6.0 and 7.0/7.1(TESTED)
Some tester report latest release is fully functioning on Android Oreo, but only by manual installation method(Q/A 10B)
- See my project release page
This won't require to modify SELinux(in most case), let your device become more safe. Also, it's good to customizing your own rom or even port rom with JamesDSP.
See readme in download link.
Better contact me by email. Send to james34602@gmail.com
The engine frame is based on Antti S. Lankila's DSPManager.
Current project compatibility and new features supported by James Fung.
Android framework components by Google
Advanced IIR filters library by Vinnie Falco, modify by Bernd Porr, with shrinked to minimum features needed.
Source code is provided under the GPL V2
- Matlab, for verifying algorithm correctness
- Various application testers
