diff --git a/src/algorithms/toolbox.jl b/src/algorithms/toolbox.jl
index 8adb02378..48cdf58db 100644
--- a/src/algorithms/toolbox.jl
+++ b/src/algorithms/toolbox.jl
@@ -14,10 +14,15 @@ function infinite_temperature(ham::MPOHamiltonian)
     ]
 end
 
-"calculates the galerkin error"
+"""
+    calc_galerkin(state, envs)
+
+Calculate the galerkin error.
+"""
 function calc_galerkin(state::Union{InfiniteMPS,FiniteMPS,WindowMPS}, loc, envs)::Float64
-    out = ∂∂AC(loc, state, envs.opp, envs) * state.AC[loc]
-    out -= state.AL[loc] * state.AL[loc]' * out
+    AC´ = ∂∂AC(loc, state, envs.opp, envs) * state.AC[loc]
+    normalize!(AC´)
+    out = add!(AC´, state.AL[loc] * state.AL[loc]' * AC´, -1)
     return norm(out)
 end
 calc_galerkin(state::Union{InfiniteMPS,FiniteMPS,WindowMPS}, envs)::Float64 =
@@ -25,14 +30,15 @@ calc_galerkin(state::Union{InfiniteMPS,FiniteMPS,WindowMPS}, envs)::Float64 =
 function calc_galerkin(state::MPSMultiline, envs::PerMPOInfEnv)::Float64
     above = isnothing(envs.above) ? state : envs.above
 
-    return maximum(
-        [
-            norm(
-                leftnull(state.AC[row + 1, col])' *
-                (∂∂AC(row, col, state, envs.opp, envs) * above.AC[row, col]),
-            ) for (row, col) in product(1:size(state, 1), 1:size(state, 2))
-        ][:],
-    )
+    εs = zeros(Float64, size(state, 1), size(state, 2))
+    for (row, col) in product(1:size(state, 1), 1:size(state, 2))
+        AC´ = ∂∂AC(row, col, state, envs.opp, envs) * above.AC[row, col]
+        normalize!(AC´)
+        out = add!(AC´, state.AL[row + 1, col] * state.AL[row + 1, col]' * AC´, -1)
+        εs[row, col] = norm(out)
+    end
+
+    return maximum(εs[:])
 end
 
 "