*Not so light on processing* doesn't mean that it requires 64 bit float points calculations. They are different things. For instance, handling a million coefficients for polynomial fitting isn't about needing longer floating-point numbers. The real heavy lifting, where the bulk of computing power goes, is into modifying matrices in DSP design. Using 64-bit floats doesn't necessarily simplify or improve these matrix adjustments.It depends on complexity of algorithm you are using and if PEQ is the only thing. When already running some DSP on computer, why not to use it also for room eq convolution or for example 5.1 to stereo. You can be interested in using anything from VST plugin world ... for example I like Goodherz CanOpener VST plugin for headphone crossfeed. Not so light on processing ... All DSP in modern players is running in 32 or 64bit floating point resolution including VST plugins. So it is rather about the basic question: Do you enter the computer DSP world, or not?
Here's a savitzky golay filter I designed vs Matlab's savgol filter output. The filter I designed was for an embedded device, so the device is limited to single precision. Matlab uses double precision.
The differences are in worst case at -120db.
Last edited: