rt-plugins: LADSPA plugins for Active Loudspeakers

This is a collection of IIR digital filter plugins, designed for using a computer to perform crossover and equalization for active loudspeakers.  The plugins are written for LADSPA, the Linux Audio Developers Plugin API, so they can be used in any LADSPA-capable host: I mostly use ecasound.  The code is released under the GNU Public License, i.e. this is free software.

What are they for?

I developed these for implementing DSP versions of the Linkwitz Labs Pluto and LX521. Now I use them for rapid-prototyping audio reproduction systems. But the filters are generic building blocks that could be used for other things.  It's hardly an exhaustive collection, just what I've needed myself and have bothered to write.  I'll probably add more over time.

Why not use existing plugins?

Indeed. Most of the filters here have already been implemented elsewhere. But since existing implementations lack clear documentation, it was difficult to discover e.g. the order of a filter and what its parameters mean (does the frequency of a highpass give the -3dB point? or the -6dB point? Is it 6dB/oct? or 12?) I ended up writing these plugins myself so I could be certain that the parameters have their standard meanings in the audio engineering world, and that the filters have the expected magnitiude and phase responses.

Are they audiophile grade?

I've tested the filters extensively for distortion: it's neglibible. You can run the tests yourself -- I've included them in the testing folder. The filters do what they're supposed to while adding barely measurable distortion or digital "noise". Beyond that, questions as to whether these filters are more (or less) "transparent" than others are misguided. But yes, the filters are "audiophile quality" if that matters.


Source Package

Release Date




Plugins Description





RTlowpass1 9015 fc[Hz]

1st-order low-pass, -3dB at fc

RThighpass1 9016 fc[Hz]

1st-order high-pass, -3dB at fc

RTlowpass 9006 fc[Hz], Q

2nd-order low-pass, -3dB at fc

RThighpass 9007 fc[Hz], Q

2nd-order high-pass, -3dB at fc

RTlr4lowpass 9020 fc[Hz]

Linkwitz-Riley 4th-order low-pass, -6dB at fc

RTlr4hipass 9021 fc[Hz]

Linkwitz-Riley 4th-order high-pass, -6dB at fc

RTlowshelf 9002 gain[dB], fc[Hz], Q

low shelving (unity gain at high frequency)

RThighshelf 9003 gain[dB], fc[Hz], Q

high shelving (unity gain at low frequency)

RTparaeq 9001 gain[dB], fc[Hz], Q

parametric eq (notch/peak) at fc

RTallpass1 9010 fc[Hz]

1st-order all-pass (non-inverting)

RTallpass2 9005 fc[Hz], Q

2nd-order all-pass (non-inverting)

The parameters take floating point values.  The gain parameter can be positive or negative.  The fc parameter specifies the corner frequency for the high- and low-pass filters, and the center of the transition band for the shelving and all-pass filters.

All of the plugins are designed for real-time use.  They are implemented as digital biquads, designed by applying a bilinear transform to the corresponding analog transfer function, with pre-warping to account for the frequency-warping inherent to biquads.  The biquad coefficients and accumulators are kept in double precision (53 bits of precision) to make quantization distortion negligible (except maybe for extremely high Q or low fc).

I have a blog post that includes some typical magnitude and phase response measurements for all of these, plus some examples of how to use the filters in ecasound.

Note: Each plugin is supposed to have a unique index, assigned by the ladspa.org people.  I haven't requested these indices yet as things are still in flux.  If you have other LADSPA plugins with the same index values I don't know what will happen.


  1. Make sure you have the LADSPA header ladspa.h installed on your system.  You can get it here, or on Ubuntu (and probably other Debian-based distros) you can just do: sudo apt-get install ladspa-sdk
  2. Unpack the source tarball: tar xfz rt-plugins-x.x.x.tar.gz (where x.x.x is the version number)
  3. Enter the build folder: cd rt-plugins-x.x.x/build/
  4. Run cmake: cmake ..
  5. Maybe edit the Makefile (you probably don't need to)
  6. Compile the code: make
  7. Install it: sudo make install

By default this installs the plugins in /usr/local/lib/ladspa.  If you want them to go somewhere else, edit the Makefile before installing.
You probably want to set the environment variable LADSPA_PATH to contain the path to the plugins, so they can be found; something like this in your ~/.bashrc (or some such):

export LADSPA_PATH=/usr/local/lib/ladspa:/usr/lib/ladspa

Unfortunately sine.so (installed by ladspa-sdk) has a bug that can throw a "double free or corruption" error, even if you don't actually use that plugin (I think it occurs when a LADSPA host scans for available plugins).  If you encounter this the best remedy is probably to just rm -f /usr/lib/ladspa/sine.so.

Denormal Handling

Denormals are vanishingly small floating point values which, if they occur, can cause floating point computations (e.g. DSP) to become grindingly slow.  A good technical description of the denormals problem in software DSP, and some solutions, can be found here; Carl Hetherington gives a useful discussion of denormals in LADSPA plugins.  He also has a plugin torture test that I've used to ensure my plugins' outputs don't generate denormals.  As of version 0.0.2 I use a (hopefully) platform-independent solution, so these plugins can be built for any architecture.


I used some of Steve Harris's wonderful swh-plugins as a code base, which in turn implements the digital biquads described in Robert Bristow-Johnson's "Cookbook formulae for audio EQ biquad filter coefficients" (you can find the corresponding transfer functions defined there).  I changed the interface to some of the filters and implemented some that were missing (shelving and all-pass filters).  Thanks Steve for making the code available.

Learn More

Most of what you need to understand how the filters work can be found in wikipedia:


Back to my home page