diff --git a/src/node_i18n.cc b/src/node_i18n.cc
index c537a247f55ff0..6c56c9dd166027 100644
--- a/src/node_i18n.cc
+++ b/src/node_i18n.cc
@@ -425,7 +425,11 @@ void ConverterObject::Create(const FunctionCallbackInfo<Value>& args) {
                         nullptr, nullptr, nullptr, &status);
   }
 
-  new ConverterObject(env, obj, conv, flags);
+  auto converter = new ConverterObject(env, obj, conv, flags);
+  size_t sublen = ucnv_getMinCharSize(conv);
+  std::string sub(sublen, '?');
+  converter->set_subst_chars(sub.c_str());
+
   args.GetReturnValue().Set(obj);
 }
 
@@ -473,6 +477,11 @@ void ConverterObject::Decode(const FunctionCallbackInfo<Value>& args) {
   size_t source_length = input.length();
 
   UChar* target = *result;
+  // printf("source c: %c\n", *source);
+  // printf("source d: %d\n", *source);
+  // printf("source s: %s\n", source);
+  // printf("source_length: %zu\n", source_length);
+  // printf("limit: %zu\n", limit);
   ucnv_toUnicode(converter->conv(),
                  &target,
                  target + limit,
@@ -485,6 +494,7 @@ void ConverterObject::Decode(const FunctionCallbackInfo<Value>& args) {
   if (U_SUCCESS(status)) {
     bool omit_initial_bom = false;
     if (limit > 0) {
+      // printf("result[0]: %d\n", result[0]);
       result.SetLength(target - &result[0]);
       if (result.length() > 0 &&
           converter->unicode() &&
diff --git a/test/parallel/test-whatwg-encoding-custom-textdecoder.js b/test/parallel/test-whatwg-encoding-custom-textdecoder.js
index 1fa65164c70678..a3a0cb58f15b86 100644
--- a/test/parallel/test-whatwg-encoding-custom-textdecoder.js
+++ b/test/parallel/test-whatwg-encoding-custom-textdecoder.js
@@ -199,3 +199,10 @@ if (common.hasIntl) {
   const str = decoder.decode(chunk);
   assert.strictEqual(str, 'foo\ufffd');
 }
+
+{
+  const decoder = new TextDecoder('Shift_JIS');
+  const chunk = new Uint8Array([-1]);
+  const str = decoder.decode(chunk);
+  assert.strictEqual(str, '\ufffd');
+}