API Reference

DR!(CSM::FreqArray)

Apply diagonal removal to CSM.

DR(CSM::FreqArray)

Apply diagonal removal to CSM and new array.

narrow2oct(x::FreqArray,n,nomial::Bool=true;psd=false)

Sum narrow band spectrum to 1/n octave bands given narrow band frequencies f in x.

Environment

The Environment struct is required for most methods and defines the geometrical setup, constants, and stores the relevant data together. The microphone array is assumed to be at the center of the coordinate system. The fields are

Arguments

Required:

• micgeom::Array{T,2}: (x,y,z) coordinates for microhone array.
• z0::Real: distance to source plane from microphone array.
• CSM::FreqArray: Cross-spectral matrix, size M x M x Nf, where M is the number of microphones and Nf is the number of frequency bins.

Optional:

• flim::Tuple=extrema(CSM.fc): Frequency limits (fmin,fmax).
• Nx::Integer=21: Number of computational gridpoint in x direction.
• Ny::Integer=21: Number of computational gridpoint in y direction.
• xlim::Tuple=(-1.,1.): Cartesian x-coordinate limits.
• ylim::Tuple=(-1.,1.): Cartesian y-coordinate limits.
• wv=ones(size(micgeom,2)): Microphones on/off.
• wc::Bool=false: Microphones coherence weighting.
• dr::Bool=false: Diagonal removal of CSM
• w=ones(M,Nf): Microphone weights. If wc=true it calculates coherence weighting.
• shear::Bool = false: Amiet phase correction
• ampcorr::Bool = shear: Amiet amplitude correction (only applies when shear = true)
• c::Real=343.: Speed of sound [m/s].
• Ma::Real=0.0: Mach number (sign determines flow direction)
• h::Real=0.0: Distance from array center to shear layer (Amiet correction) should be supplied when Ma != 0.
• multi_thread::Bool = false: Enable multi-threading via the ThreadsX package.

coherence_weights(CSM)

Calculate microphone weights according to mean mututal-coherence.

Amaral et al. “Slat Noise from an MD30P30N Airfoil at Extreme Angles of Attack,” AIAA Journal, vol. 56, no. 3, pp. 964–978, Mar. 2018.

enbw(fs,win)

Equivalent Noise Band Width. Use to convert power spectrum to power spectral density and reverse. PS = PSD*ENBW

octavebandlimits(fc,n)

Compute 1/n� octave band limits given center frequencies fc.

octavebands(n,flim=(16.,40000.),nomial::Bool=false)

Compute 1/n octave band center frequencies.

beamforming(Environment)

Calculate frequency-domain beamforming using cross-spectral matrix �csm and steering vector v stored in Environment struct

beamforming(csm,v)

Calculate frequency-domain beamforming using cross-spectral matrix �csm and steering vector v.� csm must be a square (Hermitian) matrix optionally with a �third (frequency) dimension. First dimension of v and csm must be equal.

csm(t;n=1024,noverlap=div(n,2),fs=1,win=DSP.hanning(n),scaling="spectrum")

Calculate cross-spectral matrix from time series t which is S x M dimensional, where S is the number of samples and Mthe number of microphones.

psf(Environment[,cent])

Calculate frequency-domain point spread function using the Environment struct to access steeringvectors. Optionally, supply the index where the psf is centered, default is (N/2)+1.

SPI(b::FreqArray,E::Environment,dxdy,limits)

Convenience function to normalize source integration of a beamforming result.

sourceintegration(x::Vector,env::Environment,limits::Vector{T}) where T
sourceintegration(x::Vector,env::Environment,limits::Vector{Vector{T}}) where T
sourceintegration(x::FreqArray,env::Environment,limits::Vector{T}) where T
sourceintegration(x::FreqArray,env::Environment,limits::Vector{Vector{T}}) where T

Source integration of x from limits given in cartesian coordinates [xmin,xmax,ymin,ymax].

src1 = [-0.1,0.1,-0.1,0.1]
src2 = [-0.2,0.0,-0.1,0.1]
limits = [src1,src2]
srcint = sourceintegration(x,env,limits)

Optionally supply keyword int_thres to set a threshold on source integration. Default is set to Inf which sum all values within integration region.

steeringvectors(E::Environment)

Pre-compute steeringvectors for beamforming using an �Environment with the needed parameters.

csm_slow(t;n=1024,noverlap=div(n,2),fs=1,win=DSP.hanning(n),scaling="spectrum")

Calculate cross-spectral matrix from time series t which is S x M dimensional, where S is the number of samples and Mthe number of microphones. This version is slower than csm but allocates much less (in some cases), therefore csm_slow could be used for long time signals.

psf_col!(p,steeringvec,cent)

Calculate single frequency point spread function as a column vector.

point_to_region(src,dxdy)

Helper function to get integration limits from src coordinates and size of area dxdy

sources = [(0.1,0.1),(-0.1,0.1),(-0.1,-0.1),(0.1,-0.1)]
dxdy = (0.1,0.1)
limits = point_to_region(sources,dxdy)
srcint = sourceintegration(x,env,limits)

r,pc_pm = refraction_correction(E::Environment[,h1=1.5,h2=z0-h1])

Compute propagation time correction and amplitude correction for 2D planar jet.

References: - Glegg, S., & Devenport, W. (2017). Aeroacoustics of low mach number flows: Fundamentals, analysis, and measurement. Chap. 10.

r,pc_pm = refraction_correction(Δx,Δy,Δz,Ma,h1,h2)

Compute propagation time correction and amplitude correction for 2D planar jet.

References: - Glegg, S., & Devenport, W. (2017). Aeroacoustics of low mach number flows: Fundamentals, analysis, and measurement. Chap. 10.

cleanSC(E[;maxiter=50,ϕ=0.5,stopn=10,peak_removal=false,trust_region=nothing])

CLEAN-SC algorithm for source identification and quantification optionally setting maximum iterations maxiter (default 50), and loop gain ϕ (default 0.5). Additionally, a stopping criterion max_peak[i] > max_peak[i-stopn] can be set with parameter stopn. Subtraction of peak sources in the dirty map is supported by peak_removal=true/false to remove peak sources outside trust_region with limits given as [xmin,xmax,ymin,ymax] e.g. trust_region=[-0.75;1.5;-1.0;1.0].

References:

- Sijtsma, P. (2007). CLEAN based on spatial source coherence. International Journal of Aeroacoustics, 6(4), 357–374.

With inspiration from https://github.com/acoular/acoular/blob/66cba3cffb3bc72602c869f99347be76798f4ac1/acoular/fbeamform.py#L1496

damas(env::Environment, b [,f<:AbstractArray, ω::Real]; maxiter=10)

Performs exactly maxiter DAMAS iterations for all frequency bins in env.fn or frequencies in f. f must contain values that match exact a (sub)set of the values in env.fn. Successive Over Relaxation (SOR) can be set with relaxation parameter ω. Default is ω=1 corresponding to no relaxation.

References:

Brooks, T. F. et al. (2006). A deconvolution approach for the mapping of acoustic sources (DAMAS) determined from phased microphone arrays. J.Sound.Vib. 294(4), 856–879. https://doi.org/10.1016/j.jsv.2005.12.046

fista(env::Environment,b ,p [,f<:AbstractArray]; tol=1e-6, maxit=1000)

Performs deconvolution with FISTA with a nonnegativity constraint for all frequency bins in env.fn or frequencies in f. f must contain values that match exact a (sub)set of the values in env.fn.

Arguments

Required:

• env::Environment: Environment struct
• b::FreqArray: Beamforming array
• p::FreqArray: Point spread function array

Optional:

• tol: Convergence tolerance (default: 1e-6)
• maxit: Maximum number of iterations (default: 1000)

References:

• Beck, A., & Teboulle, M. (2009). A fast iterative shrinkage-thresholding algorithm for linear inverse problems. SIAM journal on imaging sciences, 2(1), 183-202.
• Lylloff, O., Fernández-Grande, E., Agerkvist, F., Hald, J., Tiana Roig, E., & Andersen, M. S. (2015). Improving the efficiency of deconvolution algorithms for sound source localization. The journal of the acoustical society of America, 138(1), 172-180