diff --git a/Source/FloatHelper.Test.cs b/Source/FloatHelper.Test.cs
index 15a27c5..5e1bc5b 100644
--- a/Source/FloatHelper.Test.cs
+++ b/Source/FloatHelper.Test.cs
@@ -143,25 +143,80 @@ namespace Nuclex.Support {
       );
     }
 
-    // http://www.altdevblogaday.com/2012/02/22/comparing-floating-point-numbers-2012-edition/
-    // Make both positive
-    // If both are negative -> fine
-    // If both are positive -> fine
-    // If different -> Measure both distances to zero in ulps and sum them
     /// <summary>
-    ///   Verifies that the negative floating point zero is within one ulp of the positive
-    ///   floating point zero and vice versa
+    ///   Verifies that two denormalized floats can be compared in ulps
     /// </summary>
     [Test]
-    public void NegativeZeroFloatEqualsPositiveZero() {
+    public void DenormalizedFloatsCanBeCompared() {
       float zero = 0.0f;
       float zeroPlusOneUlp = FloatHelper.ReinterpretAsFloat(
         FloatHelper.ReinterpretAsInt(zero) + 1
       );
       float zeroMinusOneUlp = -zeroPlusOneUlp;
 
-      bool test = FloatHelper.AreAlmostEqual(zeroMinusOneUlp, zeroPlusOneUlp, 1);
+      // Across zero
+      Assert.IsFalse(FloatHelper.AreAlmostEqual(zeroMinusOneUlp, zeroPlusOneUlp, 1));
+      Assert.IsTrue(FloatHelper.AreAlmostEqual(zeroPlusOneUlp, zeroMinusOneUlp, 2));
 
+      // Against zero
+      Assert.IsFalse(FloatHelper.AreAlmostEqual(zero, zeroPlusOneUlp, 0));
+      Assert.IsTrue(FloatHelper.AreAlmostEqual(zero, zeroPlusOneUlp, 1));
+      Assert.IsFalse(FloatHelper.AreAlmostEqual(zero, zeroMinusOneUlp, 0));
+      Assert.IsTrue(FloatHelper.AreAlmostEqual(zero, zeroMinusOneUlp, 1));
+    }
+
+    /// <summary>
+    ///   Verifies that the negative floating point zero is within one ulp of the positive
+    ///   floating point zero and vice versa
+    /// </summary>
+    [Test]
+    public void NegativeZeroFloatEqualsPositiveZero() {
+      Assert.IsTrue(
+        FloatHelper.AreAlmostEqual(
+          FloatHelper.NegativeZeroFloat, FloatHelper.PositiveZeroFloat, 0
+        )
+      );
+      Assert.IsTrue(
+        FloatHelper.AreAlmostEqual(
+          FloatHelper.PositiveZeroFloat, FloatHelper.NegativeZeroFloat, 0
+        )
+      );
+    }
+
+    /// <summary>Verifies that floats can be compared across the zero boundary</summary>
+    [Test]
+    public void FloatsCanBeComparedAcrossZeroInUlps() {
+      float tenUlps = float.Epsilon * 10.0f;
+
+      Assert.IsTrue(FloatHelper.AreAlmostEqual(-tenUlps, tenUlps, 20));
+      Assert.IsTrue(FloatHelper.AreAlmostEqual(tenUlps, -tenUlps, 20));
+      Assert.IsFalse(FloatHelper.AreAlmostEqual(-tenUlps, tenUlps, 19));
+
+      Assert.IsTrue(FloatHelper.AreAlmostEqual(-tenUlps, 0, 10));
+      Assert.IsTrue(FloatHelper.AreAlmostEqual(0, -tenUlps, 10));
+      Assert.IsFalse(FloatHelper.AreAlmostEqual(-tenUlps, 0, 9));
+
+      Assert.IsTrue(FloatHelper.AreAlmostEqual(0, tenUlps, 10));
+      Assert.IsTrue(FloatHelper.AreAlmostEqual(tenUlps, 0, 10));
+      Assert.IsFalse(FloatHelper.AreAlmostEqual(0, tenUlps, 9));
+    }
+
+    /// <summary>
+    ///   Verifies that two denormalized doubles can be compared in ulps
+    /// </summary>
+    [Test]
+    public void DenormalizedDoublesCanBeCompared() {
+      double zero = 0.0;
+      double zeroPlusOneUlp = FloatHelper.ReinterpretAsDouble(
+        FloatHelper.ReinterpretAsLong(zero) + 1
+      );
+      double zeroMinusOneUlp = -zeroPlusOneUlp;
+
+      // Across zero
+      Assert.IsFalse(FloatHelper.AreAlmostEqual(zeroMinusOneUlp, zeroPlusOneUlp, 1));
+      Assert.IsTrue(FloatHelper.AreAlmostEqual(zeroPlusOneUlp, zeroMinusOneUlp, 2));
+
+      // Against zero
       Assert.IsFalse(FloatHelper.AreAlmostEqual(zero, zeroPlusOneUlp, 0));
       Assert.IsTrue(FloatHelper.AreAlmostEqual(zero, zeroPlusOneUlp, 1));
       Assert.IsFalse(FloatHelper.AreAlmostEqual(zero, zeroMinusOneUlp, 0));
@@ -174,18 +229,34 @@ namespace Nuclex.Support {
     /// </summary>
     [Test]
     public void NegativeZeroDoubleEqualsPositiveZero() {
-      double zero = 0.0;
-      double zeroPlusOneUlp = FloatHelper.ReinterpretAsDouble(
-        FloatHelper.ReinterpretAsLong(zero) + 1
+      Assert.IsTrue(
+        FloatHelper.AreAlmostEqual(
+          FloatHelper.NegativeZeroDouble, FloatHelper.NegativeZeroDouble, 0
+        )
       );
-      double zeroMinusOneUlp = -zeroPlusOneUlp;
+      Assert.IsTrue(
+        FloatHelper.AreAlmostEqual(
+          FloatHelper.NegativeZeroDouble, FloatHelper.NegativeZeroDouble, 0
+        )
+      );
+    }
 
-      bool test = FloatHelper.AreAlmostEqual(zeroMinusOneUlp, zeroPlusOneUlp, 1);
+    /// <summary>Verifies that doubles can be compared across the zero boundary</summary>
+    [Test]
+    public void DoublesCanBeComparedAcrossZeroInUlps() {
+      double tenUlps = double.Epsilon * 10.0;
 
-      Assert.IsFalse(FloatHelper.AreAlmostEqual(zero, zeroPlusOneUlp, 0));
-      Assert.IsTrue(FloatHelper.AreAlmostEqual(zero, zeroPlusOneUlp, 1));
-      Assert.IsFalse(FloatHelper.AreAlmostEqual(zero, zeroMinusOneUlp, 0));
-      Assert.IsTrue(FloatHelper.AreAlmostEqual(zero, zeroMinusOneUlp, 1));
+      Assert.IsTrue(FloatHelper.AreAlmostEqual(-tenUlps, tenUlps, 20));
+      Assert.IsTrue(FloatHelper.AreAlmostEqual(tenUlps, -tenUlps, 20));
+      Assert.IsFalse(FloatHelper.AreAlmostEqual(-tenUlps, tenUlps, 19));
+
+      Assert.IsTrue(FloatHelper.AreAlmostEqual(-tenUlps, 0, 10));
+      Assert.IsTrue(FloatHelper.AreAlmostEqual(0, -tenUlps, 10));
+      Assert.IsFalse(FloatHelper.AreAlmostEqual(-tenUlps, 0, 9));
+
+      Assert.IsTrue(FloatHelper.AreAlmostEqual(0, tenUlps, 10));
+      Assert.IsTrue(FloatHelper.AreAlmostEqual(tenUlps, 0, 10));
+      Assert.IsFalse(FloatHelper.AreAlmostEqual(0, tenUlps, 9));
     }
 
   }
diff --git a/Source/FloatHelper.cs b/Source/FloatHelper.cs
index 0833753..1a09609 100644
--- a/Source/FloatHelper.cs
+++ b/Source/FloatHelper.cs
@@ -44,11 +44,11 @@ namespace Nuclex.Support {
   /// </remarks>
   public static class FloatHelper {
 
-    #region struct FloatIntUnion
+    #region struct FloatInt32Union
 
     /// <summary>Union of a floating point variable and an integer</summary>
     [StructLayout(LayoutKind.Explicit)]
-    private struct FloatIntUnion {
+    private struct FloatInt32Union {
 
       /// <summary>The union's value as a floating point variable</summary>
       [FieldOffset(0)]
@@ -64,13 +64,13 @@ namespace Nuclex.Support {
 
     }
 
-    #endregion // struct FloatIntUnion
+    #endregion // struct FloatInt32Union
 
-    #region struct DoubleLongUnion
+    #region struct DoubleInt64Union
 
     /// <summary>Union of a double precision floating point variable and a long</summary>
     [StructLayout(LayoutKind.Explicit)]
-    private struct DoubleLongUnion {
+    private struct DoubleInt64Union {
 
       /// <summary>The union's value as a double precision floating point variable</summary>
       [FieldOffset(0)]
@@ -86,7 +86,7 @@ namespace Nuclex.Support {
 
     }
 
-    #endregion // struct DoubleLongUnion
+    #endregion // struct DoubleInt64Union
 
     /// <summary>A floating point value that holds a positive zero</summary>
     public const float PositiveZeroFloat = +0.0f;
@@ -145,11 +145,30 @@ namespace Nuclex.Support {
     ///   <para>
     ///     Implementation partially follows the code outlined here (link now defunct):
     ///     http://www.anttirt.net/2007/08/19/proper-floating-point-comparisons/
+    ///     And here:
+    ///     http://www.altdevblogaday.com/2012/02/22/comparing-floating-point-numbers-2012-edition/
     ///   </para>
     /// </remarks>
     public static bool AreAlmostEqual(float left, float right, int maxUlps) {
-      FloatIntUnion leftUnion = new FloatIntUnion();
-      FloatIntUnion rightUnion = new FloatIntUnion();
+      var leftUnion = new FloatInt32Union();
+      var rightUnion = new FloatInt32Union();
+
+      leftUnion.Float = left;
+      rightUnion.Float = right;
+
+      if(leftUnion.Int < 0) {
+        leftUnion.Int = unchecked((int)0x80000000 - leftUnion.Int);
+      }
+      if(rightUnion.Int < 0) {
+        rightUnion.Int = unchecked((int)0x80000000 - rightUnion.Int);
+      }
+
+      return Math.Abs(rightUnion.Int - leftUnion.Int) <= maxUlps;
+    }
+#if false
+    public static bool OldAreAlmostEqual(float left, float right, int maxUlps) {
+      FloatInt32Union leftUnion = new FloatInt32Union();
+      FloatInt32Union rightUnion = new FloatInt32Union();
 
       leftUnion.Float = left;
       rightUnion.Float = right;
@@ -165,6 +184,7 @@ namespace Nuclex.Support {
 
       return (Math.Abs(leftUnion.Int - rightUnion.Int) <= maxUlps);
     }
+#endif
 
     /// <summary>Compares two double precision floating point values for equality</summary>
     /// <param name="left">First double precision floating point value to be compared</param>
@@ -189,11 +209,30 @@ namespace Nuclex.Support {
     ///   <para>
     ///     Implementation partially follows the code outlined here:
     ///     http://www.anttirt.net/2007/08/19/proper-floating-point-comparisons/
+    ///     And here:
+    ///     http://www.altdevblogaday.com/2012/02/22/comparing-floating-point-numbers-2012-edition/
     ///   </para>
     /// </remarks>
     public static bool AreAlmostEqual(double left, double right, long maxUlps) {
-      DoubleLongUnion leftUnion = new DoubleLongUnion();
-      DoubleLongUnion rightUnion = new DoubleLongUnion();
+      var leftUnion = new DoubleInt64Union();
+      var rightUnion = new DoubleInt64Union();
+
+      leftUnion.Double = left;
+      rightUnion.Double = right;
+
+      if(leftUnion.Long < 0) {
+        leftUnion.Long = unchecked((long)0x8000000000000000 - leftUnion.Long);
+      }
+      if(rightUnion.Long < 0) {
+        rightUnion.Long = unchecked((long)0x8000000000000000 - rightUnion.Long);
+      }
+
+      return Math.Abs(rightUnion.Long - leftUnion.Long) <= maxUlps;
+    }
+#if false
+    public static bool OldAreAlmostEqual(double left, double right, long maxUlps) {
+      DoubleInt64Union leftUnion = new DoubleInt64Union();
+      DoubleInt64Union rightUnion = new DoubleInt64Union();
 
       leftUnion.Double = left;
       rightUnion.Double = right;
@@ -209,6 +248,7 @@ namespace Nuclex.Support {
 
       return (Math.Abs(leftUnion.Long - rightUnion.Long) <= maxUlps);
     }
+#endif
 
     /// <summary>
     ///   Reinterprets the memory contents of a floating point value as an integer value
@@ -220,7 +260,7 @@ namespace Nuclex.Support {
     ///   The memory contents of the floating point value interpreted as an integer
     /// </returns>
     public static int ReinterpretAsInt(this float value) {
-      FloatIntUnion union = new FloatIntUnion();
+      FloatInt32Union union = new FloatInt32Union();
       union.Float = value;
       return union.Int;
     }
@@ -237,7 +277,7 @@ namespace Nuclex.Support {
     ///   interpreted as an integer
     /// </returns>
     public static long ReinterpretAsLong(this double value) {
-      DoubleLongUnion union = new DoubleLongUnion();
+      DoubleInt64Union union = new DoubleInt64Union();
       union.Double = value;
       return union.Long;
     }
@@ -250,7 +290,7 @@ namespace Nuclex.Support {
     ///   The memory contents of the integer value interpreted as a floating point value
     /// </returns>
     public static float ReinterpretAsFloat(this int value) {
-      FloatIntUnion union = new FloatIntUnion();
+      FloatInt32Union union = new FloatInt32Union();
       union.Int = value;
       return union.Float;
     }
@@ -265,7 +305,7 @@ namespace Nuclex.Support {
     ///   floating point value
     /// </returns>
     public static double ReinterpretAsDouble(this long value) {
-      DoubleLongUnion union = new DoubleLongUnion();
+      DoubleInt64Union union = new DoubleInt64Union();
       union.Long = value;
       return union.Double;
     }