diff --git a/RecoVertex/PrimaryVertexProducer/python/OfflinePixel3DPrimaryVertices_cfi.py b/RecoVertex/PrimaryVertexProducer/python/OfflinePixel3DPrimaryVertices_cfi.py
index b6ad7e2ed8eec..39750e9dfd0f6 100644
--- a/RecoVertex/PrimaryVertexProducer/python/OfflinePixel3DPrimaryVertices_cfi.py
+++ b/RecoVertex/PrimaryVertexProducer/python/OfflinePixel3DPrimaryVertices_cfi.py
@@ -12,12 +12,28 @@
         maxEta = 100.,
     ),
 
-    TkClusParameters = dict(
-        TkDAClusParameters = dict(
-            dzCutOff = 4.0,
-            Tmin = 2.4,
-            vertexSize = 0.01,
-            uniquetrkweight = 0.9
+    TkClusParameters = cms.PSet(
+        algorithm   = cms.string("DA_vect"),
+        TkDAClusParameters = cms.PSet(
+            # the first 4 parameters are adjusted for pixel vertices
+            dzCutOff = cms.double(4.),        # outlier rejection after freeze-out (T<Tmin)
+            Tmin = cms.double(2.4),           # end of vertex splitting
+            vertexSize = cms.double(0.01),    # added in quadrature to track-z resolution
+            uniquetrkweight = cms.double(0.9),# require at least two tracks with this weight at T=Tpurge
+            # the rest is the same as the default defined in OfflinePrimaryVertices
+            coolingFactor = cms.double(0.6),  # moderate annealing speed
+            zrange = cms.double(4.),          # consider only clusters within 4 sigma*sqrt(T) of a track
+            delta_highT = cms.double(1.e-2),  # convergence requirement at high T
+            delta_lowT = cms.double(1.e-3),   # convergence requirement at low T
+            convergence_mode = cms.int32(0),  # 0 = two steps, 1 = dynamic with sqrt(T)
+            Tpurge = cms.double(2.0),         # cleaning
+            Tstop = cms.double(0.5),          # end of annealing
+            d0CutOff = cms.double(3.),        # downweight high IP tracks
+            zmerge = cms.double(1e-2),        # merge intermediat clusters separated by less than zmerge
+            uniquetrkminp = cms.double(0.0),  # minimal a priori track weight for counting unique tracks
+            runInBlocks = cms.bool(False),    # activate the DA running in blocks of z sorted tracks
+            block_size = cms.uint32(10000),   # block size in tracks
+            overlap_frac = cms.double(0.0)    # overlap between consecutive blocks (blocks_size*overlap_frac)
         )
     ),