Use Polynomials.jl instead of Polynomial.jl

REQUIRE

@@ -1,2 +1,2 @@
-Polynomial
+Polynomials
 Reexport

src/filter_design.jl

@@ -39,7 +39,7 @@
 # DAMAGE.
 
 module FilterDesign
-using Polynomial
+using Polynomials
 
 export ZPKFilter, TFFilter, BiquadFilter, SOSFilter, Butterworth, Lowpass, Highpass, Bandpass,
        Bandstop, analogfilter, digitalfilter, Filter
@@ -66,7 +66,7 @@ end
 similarzeros{T}(v::Array{T}, args...) = zeros(T, args...)
 
 # Get coefficients of a polynomial
-coeffs{T}(p::Poly{T}) = p.a[1+p.nzfirst:end]
+coeffs{T}(p::Poly{T}) = reverse(p.a)
 
 #
 # Filter types
@@ -87,11 +87,16 @@ immutable TFFilter{T} <: Filter
     a::Poly{T}
 
     function TFFilter(b::Poly{T}, a::Poly{T})
-        new(b/a[1], a/a[1])
+        new(b/a[end], a/a[end])
     end
 end
 TFFilter{T}(b::Poly{T}, a::Poly{T}) = TFFilter{T}(b, a)
-TFFilter{T}(b::Vector{T}, a::Vector{T}) = TFFilter{T}(Poly(b), Poly(a))
+
+# The DSP convention is lowest power first. The Polynomials.jl
+# convention is highest power first.
+TFFilter{T}(b::Vector{T}, a::Vector{T}) =
+    TFFilter{T}(Poly(b[end:-1:findfirst(b)]), Poly(a[end:-1:findfirst(a)]))
+
 function TFFilter{T,S}(b::Vector{T}, a::Vector{S})
     V = promote_type(T, S)
     TFFilter(convert(Vector{V}, b), convert(Vector{V}, a))
@@ -104,7 +109,7 @@ function convert(::Type{TFFilter}, f::ZPKFilter)
 end
 
 function convert{T}(::Type{ZPKFilter}, f::TFFilter{T})
-    k = real(f.b[1])
+    k = real(f.b[end])
     b = f.b / k
     z = convert(Vector{Complex{T}}, roots(b))
     p = convert(Vector{Complex{T}}, roots(f.a))
@@ -336,8 +341,8 @@ function transform_prototype(ftype::Bandpass, proto::TFFilter)
     bw = ftype.w2 - ftype.w1
     wo = sqrt(ftype.w1 * ftype.w2)
     tf = convert(TFFilter, proto)
-    b = tf.b.a
-    a = tf.a.a
+    b = coeffs(tf.b)
+    a = coeffs(tf.a)
     D = length(a) - 1
     N = length(b) - 1
     M = max(N, D)
@@ -378,8 +383,8 @@ function transform_prototype(ftype::Bandstop, proto::TFFilter)
     bw = ftype.w2 - ftype.w1
     wo = sqrt(ftype.w1 * ftype.w2)
     tf = convert(TFFilter, proto)
-    b = tf.b.a
-    a = tf.a.a
+    b = coeffs(tf.b)
+    a = coeffs(tf.a)
     D = length(a) - 1
     N = length(b) - 1
     M = max(N, D)

src/filter_response.jl

@@ -5,7 +5,7 @@ module FilterResponse
 
 export freqz, freqs
 
-using Polynomial
+using Polynomials
 using ..FilterDesign
 
 

test/filter_design.jl

@@ -1,4 +1,4 @@
-using DSP, Base.Test, Polynomial
+using DSP, Base.Test, Polynomials
 import Base.Sort.Lexicographic
 import DSP.FilterDesign: coeffs
 
@@ -81,7 +81,7 @@ accurate_a = coeffs(convert(TFFilter, accurate_f).a)
 
 f = convert(ZPKFilter, convert(TFFilter, Butterworth(20)))
 @test isempty(f.z)
-@test_approx_eq_eps sort(f.p, lt=lt) prototype 1e-7
+@test_approx_eq_eps sort(f.p, lt=lt) prototype 1e-6
 
 # Test that our answers are more accurate than MATLAB's
 # Output of [z, p, k] = buttap(20); [b, a] = zp2tf(z, p, k); tf2zpk(b, a)
@@ -221,7 +221,12 @@ for f in (digitalfilter(Lowpass(0.5), Butterworth(1)), digitalfilter(Lowpass(0.5
         f2 = convert(ftype1, f)
         for ftype2 in (ZPKFilter, TFFilter, BiquadFilter, SOSFilter)
             f3 = convert(ftype2, f)
-            zpkfilter_eq(f, convert(ZPKFilter, f3), eps())
+            try
+                zpkfilter_eq(f, convert(ZPKFilter, f3), sqrt(eps()))
+            catch e
+                println("Conversion from $ftype1 to $ftype2 failed:")
+                rethrow(e)
+            end
         end
     end
 end