diff --git a/dev/make-root/rntuple_float_types.C b/dev/make-root/rntuple_float_types.C
new file mode 100644
index 0000000..9f523e4
--- /dev/null
+++ b/dev/make-root/rntuple_float_types.C
@@ -0,0 +1,136 @@
+/* https://root.cern/doc/master/ntpl001__staff_8C.html */
+R__LOAD_LIBRARY(ROOTNTuple)
+#include <ROOT/RField.hxx>
+#include <ROOT/RNTuple.hxx>
+#include <ROOT/RNTupleModel.hxx>
+#include <ROOT/RNTupleWriter.hxx>
+#include <ROOT/RNTupleUtil.hxx>
+#include <ROOT/RRawFile.hxx>
+
+#include <cassert>
+#include <variant>
+
+using RNTupleModel = ROOT::Experimental::RNTupleModel;
+using RNTupleWriter = ROOT::Experimental::RNTupleWriter;
+template <typename T>
+using RField = ROOT::Experimental::RField<T>;
+using EColumnType = ROOT::Experimental::EColumnType;
+
+void rntuple_float_types() {
+  std::string rootFileName{"test_float_types_rntuple_v1-0-0-0.root"};
+  auto model = RNTupleModel::Create();
+
+  auto trunc10 = std::make_unique<RField<float>>("trunc10");
+  trunc10->SetColumnRepresentatives({{EColumnType::kReal32Trunc}});
+  trunc10->SetTruncated(10);
+  model->AddField(std::move(trunc10));
+
+  auto trunc16 = std::make_unique<RField<float>>("trunc16");
+  trunc16->SetColumnRepresentatives({{EColumnType::kReal32Trunc}});
+  trunc16->SetTruncated(16);
+  model->AddField(std::move(trunc16));
+
+  auto trunc24 = std::make_unique<RField<float>>("trunc24");
+  trunc24->SetColumnRepresentatives({{EColumnType::kReal32Trunc}});
+  trunc24->SetTruncated(24);
+  model->AddField(std::move(trunc24));
+
+  auto trunc31 = std::make_unique<RField<float>>("trunc31");
+  trunc31->SetColumnRepresentatives({{EColumnType::kReal32Trunc}});
+  trunc31->SetTruncated(31);
+  model->AddField(std::move(trunc31));
+
+  auto quant1 = std::make_unique<RField<float>>("quant1");
+  quant1->SetColumnRepresentatives({{EColumnType::kReal32Quant}});
+  quant1->SetQuantized(-2.f, 3.f, 1);
+  model->AddField(std::move(quant1));
+
+  auto quant8 = std::make_unique<RField<float>>("quant8");
+  quant8->SetColumnRepresentatives({{EColumnType::kReal32Quant}});
+  quant8->SetQuantized(-2.f, 3.f, 8);
+  model->AddField(std::move(quant8));
+
+  auto quant16 = std::make_unique<RField<float>>("quant16");
+  quant16->SetColumnRepresentatives({{EColumnType::kReal32Quant}});
+  quant16->SetQuantized(-2.f, 3.f, 16);
+  model->AddField(std::move(quant16));
+
+  auto quant20 = std::make_unique<RField<float>>("quant20");
+  quant20->SetColumnRepresentatives({{EColumnType::kReal32Quant}});
+  quant20->SetQuantized(-2.f, 3.f, 20);
+  model->AddField(std::move(quant20));
+
+  auto quant24 = std::make_unique<RField<float>>("quant24");
+  quant24->SetColumnRepresentatives({{EColumnType::kReal32Quant}});
+  quant24->SetQuantized(-2.f, 3.f, 24);
+  model->AddField(std::move(quant24));
+
+  auto quant25 = std::make_unique<RField<float>>("quant25");
+  quant25->SetColumnRepresentatives({{EColumnType::kReal32Quant}});
+  quant25->SetQuantized(-2.f, 3.f, 25);
+  model->AddField(std::move(quant25));
+
+  auto quant32 = std::make_unique<RField<float>>("quant32");
+  quant32->SetColumnRepresentatives({{EColumnType::kReal32Quant}});
+  quant32->SetQuantized(-2.f, 3.f, 32);
+  model->AddField(std::move(quant32));
+
+  auto ntuple = RNTupleWriter::Recreate(std::move(model), "ntuple", rootFileName);
+  auto e = ntuple->CreateEntry();
+
+  *e->GetPtr<float>("trunc10") = 1.23456789f; // 00111111100111100000011001010010
+  *e->GetPtr<float>("trunc16") = 1.23456789f;
+  *e->GetPtr<float>("trunc24") = 1.23456789f;
+  *e->GetPtr<float>("trunc31") = 1.23456789f;
+  *e->GetPtr<float>("quant1") = 1.23456789f;
+  *e->GetPtr<float>("quant8") = 1.23456789f;
+  *e->GetPtr<float>("quant16") = 1.23456789f;
+  *e->GetPtr<float>("quant20") = 1.23456789f;
+  *e->GetPtr<float>("quant24") = 1.23456789f;
+  *e->GetPtr<float>("quant25") = 1.23456789f;
+  *e->GetPtr<float>("quant32") = 1.23456789f;
+
+  ntuple->Fill(*e);
+
+  *e->GetPtr<float>("trunc10") = 1.4660155e13f; // 01010101010101010101010101010101
+  *e->GetPtr<float>("trunc16") = 1.4660155e13f;
+  *e->GetPtr<float>("trunc24") = 1.4660155e13f;
+  *e->GetPtr<float>("trunc31") = 1.4660155e13f;
+  *e->GetPtr<float>("quant1") = 1.6666666f;  // 00111111110101010101010101010101
+  *e->GetPtr<float>("quant8") = 1.6666666f;
+  *e->GetPtr<float>("quant16") = 1.6666666f;
+  *e->GetPtr<float>("quant20") = 1.6666666f;
+  *e->GetPtr<float>("quant24") = 1.6666666f;
+  *e->GetPtr<float>("quant25") = 1.6666666f;
+  *e->GetPtr<float>("quant32") = 1.6666666f;
+
+  ntuple->Fill(*e);
+
+  *e->GetPtr<float>("trunc10") = -6.2875986e-22f;  // 10011100001111100000011111110000
+  *e->GetPtr<float>("trunc16") = -6.2875986e-22f;
+  *e->GetPtr<float>("trunc24") = -6.2875986e-22f;
+  *e->GetPtr<float>("trunc31") = -6.2875986e-22f;
+  *e->GetPtr<float>("quant1") = -6.2875986e-22f;
+  *e->GetPtr<float>("quant8") = -6.2875986e-22f;
+  *e->GetPtr<float>("quant16") = -6.2875986e-22f;
+  *e->GetPtr<float>("quant20") = -6.2875986e-22f;
+  *e->GetPtr<float>("quant24") = -6.2875986e-22f;
+  *e->GetPtr<float>("quant25") = -6.2875986e-22f;
+  *e->GetPtr<float>("quant32") = -6.2875986e-22f;
+
+  ntuple->Fill(*e);
+
+  *e->GetPtr<float>("trunc10") = -1.9060668f;  // 10111111111100111111100111111111
+  *e->GetPtr<float>("trunc16") = -1.9060668f;
+  *e->GetPtr<float>("trunc24") = -1.9060668f;
+  *e->GetPtr<float>("trunc31") = -1.9060668f;
+  *e->GetPtr<float>("quant1") = -1.9060668f;
+  *e->GetPtr<float>("quant8") = -1.9060668f;
+  *e->GetPtr<float>("quant16") = -1.9060668f;
+  *e->GetPtr<float>("quant20") = -1.9060668f;
+  *e->GetPtr<float>("quant24") = -1.9060668f;
+  *e->GetPtr<float>("quant25") = -1.9060668f;
+  *e->GetPtr<float>("quant32") = -1.9060668f;
+
+  ntuple->Fill(*e);
+}
diff --git a/dev/make-root/rntuple_multiple_representations.C b/dev/make-root/rntuple_multiple_representations.C
new file mode 100644
index 0000000..df72bf4
--- /dev/null
+++ b/dev/make-root/rntuple_multiple_representations.C
@@ -0,0 +1,43 @@
+/* https://root.cern/doc/master/ntpl001__staff_8C.html */
+R__LOAD_LIBRARY(ROOTNTuple)
+#include <ROOT/RNTuple.hxx>
+#include <ROOT/RFieldBase.hxx>
+#include <ROOT/RNTupleModel.hxx>
+#include <ROOT/RNTupleWriter.hxx>
+#include <ROOT/RNTupleUtil.hxx>
+#include <ROOT/RRawFile.hxx>
+
+#include <cassert>
+#include <variant>
+
+using RNTupleModel = ROOT::Experimental::RNTupleModel;
+using RNTupleWriter = ROOT::Experimental::RNTupleWriter;
+using RFieldBase = ROOT::Experimental::RFieldBase;
+using EColumnType = ROOT::Experimental::EColumnType;
+
+void rntuple_multiple_representations() {
+  std::string rootFileName{"test_multiple_representations_rntuple_v1-0-0-0.root"};
+  auto model = RNTupleModel::Create();
+
+  auto fldReal = RFieldBase::Create("real", "float").Unwrap();
+  fldReal->SetColumnRepresentatives({{EColumnType::kReal32}, {EColumnType::kReal16}});
+  model->AddField(std::move(fldReal));
+  auto ptrReal = model->GetDefaultEntry().GetPtr<float>("real");
+
+  auto ntuple = RNTupleWriter::Recreate(std::move(model), "ntuple", rootFileName);
+
+  *ptrReal = 1.0;
+  ntuple->Fill();
+  ntuple->CommitCluster();
+
+  ROOT::Experimental::Internal::RFieldRepresentationModifier::SetPrimaryColumnRepresentation(
+         const_cast<RFieldBase &>(ntuple->GetModel().GetConstField("real")), 1);
+  *ptrReal = 2.0;
+  ntuple->Fill();
+  ntuple->CommitCluster();
+
+  ROOT::Experimental::Internal::RFieldRepresentationModifier::SetPrimaryColumnRepresentation(
+         const_cast<RFieldBase &>(ntuple->GetModel().GetConstField("real")), 0);
+  *ptrReal = 3.0;
+  ntuple->Fill();
+}
diff --git a/src/skhep_testdata/data/test_float_types_rntuple_v1-0-0-0.root b/src/skhep_testdata/data/test_float_types_rntuple_v1-0-0-0.root
new file mode 100644
index 0000000..b59b82f
Binary files /dev/null and b/src/skhep_testdata/data/test_float_types_rntuple_v1-0-0-0.root differ
diff --git a/src/skhep_testdata/data/test_multiple_representations_rntuple_v1-0-0-0.root b/src/skhep_testdata/data/test_multiple_representations_rntuple_v1-0-0-0.root
new file mode 100644
index 0000000..5e6764a
Binary files /dev/null and b/src/skhep_testdata/data/test_multiple_representations_rntuple_v1-0-0-0.root differ