diff --git a/automotive/evs/1.0/vts/functional/Android.bp b/automotive/evs/1.0/vts/functional/Android.bp
index 22ceff3..e86e9bc 100644
--- a/automotive/evs/1.0/vts/functional/Android.bp
+++ b/automotive/evs/1.0/vts/functional/Android.bp
@@ -19,7 +19,8 @@
 
     srcs: [
         "VtsEvsV1_0TargetTest.cpp",
-        "FrameHandler.cpp"
+        "FrameHandler.cpp",
+        "FormatConvert.cpp"
     ],
 
     defaults: [
diff --git a/automotive/evs/1.0/vts/functional/FormatConvert.cpp b/automotive/evs/1.0/vts/functional/FormatConvert.cpp
new file mode 100644
index 0000000..e5cc8ee
--- /dev/null
+++ b/automotive/evs/1.0/vts/functional/FormatConvert.cpp
@@ -0,0 +1,173 @@
+/*
+ * Copyright (C) 2017 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#define LOG_TAG "VtsHalEvsTest"
+
+#include "FormatConvert.h"
+
+#include <algorithm>    // std::min
+
+
+// Round up to the nearest multiple of the given alignment value
+template<unsigned alignment>
+int align(int value) {
+    static_assert((alignment && !(alignment & (alignment - 1))),
+                  "alignment must be a power of 2");
+
+    unsigned mask = alignment - 1;
+    return (value + mask) & ~mask;
+}
+
+
+// Limit the given value to the provided range.  :)
+static inline float clamp(float v, float min, float max) {
+    if (v < min) return min;
+    if (v > max) return max;
+    return v;
+}
+
+
+static uint32_t yuvToRgbx(const unsigned char Y, const unsigned char Uin, const unsigned char Vin) {
+    // Don't use this if you want to see the best performance.  :)
+    // Better to do this in a pixel shader if we really have to, but on actual
+    // embedded hardware we expect to be able to texture directly from the YUV data
+    float U = Uin - 128.0f;
+    float V = Vin - 128.0f;
+
+    float Rf = Y + 1.140f*V;
+    float Gf = Y - 0.395f*U - 0.581f*V;
+    float Bf = Y + 2.032f*U;
+    unsigned char R = (unsigned char)clamp(Rf, 0.0f, 255.0f);
+    unsigned char G = (unsigned char)clamp(Gf, 0.0f, 255.0f);
+    unsigned char B = (unsigned char)clamp(Bf, 0.0f, 255.0f);
+
+    return (R      ) |
+           (G <<  8) |
+           (B << 16) |
+           0xFF000000;  // Fill the alpha channel with ones
+}
+
+
+void copyNV21toRGB32(unsigned width, unsigned height,
+                     uint8_t* src,
+                     uint32_t* dst, unsigned dstStridePixels)
+{
+    // The NV21 format provides a Y array of 8bit values, followed by a 1/2 x 1/2 interleaved
+    // U/V array.  It assumes an even width and height for the overall image, and a horizontal
+    // stride that is an even multiple of 16 bytes for both the Y and UV arrays.
+    unsigned strideLum = align<16>(width);
+    unsigned sizeY = strideLum * height;
+    unsigned strideColor = strideLum;   // 1/2 the samples, but two interleaved channels
+    unsigned offsetUV = sizeY;
+
+    uint8_t* srcY = src;
+    uint8_t* srcUV = src+offsetUV;
+
+    for (unsigned r = 0; r < height; r++) {
+        // Note that we're walking the same UV row twice for even/odd luminance rows
+        uint8_t* rowY  = srcY  + r*strideLum;
+        uint8_t* rowUV = srcUV + (r/2 * strideColor);
+
+        uint32_t* rowDest = dst + r*dstStridePixels;
+
+        for (unsigned c = 0; c < width; c++) {
+            unsigned uCol = (c & ~1);   // uCol is always even and repeats 1:2 with Y values
+            unsigned vCol = uCol | 1;   // vCol is always odd
+            rowDest[c] = yuvToRgbx(rowY[c], rowUV[uCol], rowUV[vCol]);
+        }
+    }
+}
+
+
+void copyYV12toRGB32(unsigned width, unsigned height,
+                     uint8_t* src,
+                     uint32_t* dst, unsigned dstStridePixels)
+{
+    // The YV12 format provides a Y array of 8bit values, followed by a 1/2 x 1/2 U array, followed
+    // by another 1/2 x 1/2 V array.  It assumes an even width and height for the overall image,
+    // and a horizontal stride that is an even multiple of 16 bytes for each of the Y, U,
+    // and V arrays.
+    unsigned strideLum = align<16>(width);
+    unsigned sizeY = strideLum * height;
+    unsigned strideColor = align<16>(strideLum/2);
+    unsigned sizeColor = strideColor * height/2;
+    unsigned offsetU = sizeY;
+    unsigned offsetV = sizeY + sizeColor;
+
+    uint8_t* srcY = src;
+    uint8_t* srcU = src+offsetU;
+    uint8_t* srcV = src+offsetV;
+
+    for (unsigned r = 0; r < height; r++) {
+        // Note that we're walking the same U and V rows twice for even/odd luminance rows
+        uint8_t* rowY = srcY + r*strideLum;
+        uint8_t* rowU = srcU + (r/2 * strideColor);
+        uint8_t* rowV = srcV + (r/2 * strideColor);
+
+        uint32_t* rowDest = dst + r*dstStridePixels;
+
+        for (unsigned c = 0; c < width; c++) {
+            rowDest[c] = yuvToRgbx(rowY[c], rowU[c], rowV[c]);
+        }
+    }
+}
+
+
+void copyYUYVtoRGB32(unsigned width, unsigned height,
+                     uint8_t* src, unsigned srcStridePixels,
+                     uint32_t* dst, unsigned dstStridePixels)
+{
+    uint32_t* srcWords = (uint32_t*)src;
+
+    const int srcRowPadding32 = srcStridePixels/2 - width/2;  // 2 bytes per pixel, 4 bytes per word
+    const int dstRowPadding32 = dstStridePixels   - width;    // 4 bytes per pixel, 4 bytes per word
+
+    for (unsigned r = 0; r < height; r++) {
+        for (unsigned c = 0; c < width/2; c++) {
+            // Note:  we're walking two pixels at a time here (even/odd)
+            uint32_t srcPixel = *srcWords++;
+
+            uint8_t Y1 = (srcPixel)       & 0xFF;
+            uint8_t U  = (srcPixel >> 8)  & 0xFF;
+            uint8_t Y2 = (srcPixel >> 16) & 0xFF;
+            uint8_t V  = (srcPixel >> 24) & 0xFF;
+
+            // On the RGB output, we're writing one pixel at a time
+            *(dst+0) = yuvToRgbx(Y1, U, V);
+            *(dst+1) = yuvToRgbx(Y2, U, V);
+            dst += 2;
+        }
+
+        // Skip over any extra data or end of row alignment padding
+        srcWords += srcRowPadding32;
+        dst += dstRowPadding32;
+    }
+}
+
+
+void copyMatchedInterleavedFormats(unsigned width, unsigned height,
+                                   void* src, unsigned srcStridePixels,
+                                   void* dst, unsigned dstStridePixels,
+                                   unsigned pixelSize) {
+    for (unsigned row = 0; row < height; row++) {
+        // Copy the entire row of pixel data
+        memcpy(dst, src, width * pixelSize);
+
+        // Advance to the next row (keeping in mind that stride here is in units of pixels)
+        src = (uint8_t*)src + srcStridePixels * pixelSize;
+        dst = (uint8_t*)dst + dstStridePixels * pixelSize;
+    }
+}
diff --git a/automotive/evs/1.0/vts/functional/FormatConvert.h b/automotive/evs/1.0/vts/functional/FormatConvert.h
new file mode 100644
index 0000000..3ff1eec
--- /dev/null
+++ b/automotive/evs/1.0/vts/functional/FormatConvert.h
@@ -0,0 +1,60 @@
+/*
+ * Copyright (C) 2017 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef EVS_VTS_FORMATCONVERT_H
+#define EVS_VTS_FORMATCONVERT_H
+
+#include <queue>
+#include <stdint.h>
+
+
+// Given an image buffer in NV21 format (HAL_PIXEL_FORMAT_YCRCB_420_SP), output 32bit RGBx values.
+// The NV21 format provides a Y array of 8bit values, followed by a 1/2 x 1/2 interleaved
+// U/V array.  It assumes an even width and height for the overall image, and a horizontal
+// stride that is an even multiple of 16 bytes for both the Y and UV arrays.
+void copyNV21toRGB32(unsigned width, unsigned height,
+                     uint8_t* src,
+                     uint32_t* dst, unsigned dstStridePixels);
+
+
+// Given an image buffer in YV12 format (HAL_PIXEL_FORMAT_YV12), output 32bit RGBx values.
+// The YV12 format provides a Y array of 8bit values, followed by a 1/2 x 1/2 U array, followed
+// by another 1/2 x 1/2 V array.  It assumes an even width and height for the overall image,
+// and a horizontal stride that is an even multiple of 16 bytes for each of the Y, U,
+// and V arrays.
+void copyYV12toRGB32(unsigned width, unsigned height,
+                     uint8_t* src,
+                     uint32_t* dst, unsigned dstStridePixels);
+
+
+// Given an image buffer in YUYV format (HAL_PIXEL_FORMAT_YCBCR_422_I), output 32bit RGBx values.
+// The NV21 format provides a Y array of 8bit values, followed by a 1/2 x 1/2 interleaved
+// U/V array.  It assumes an even width and height for the overall image, and a horizontal
+// stride that is an even multiple of 16 bytes for both the Y and UV arrays.
+void copyYUYVtoRGB32(unsigned width, unsigned height,
+                     uint8_t* src, unsigned srcStrideBytes,
+                     uint32_t* dst, unsigned dstStrideBytes);
+
+
+// Given an simple rectangular image buffer with an integer number of bytes per pixel,
+// copy the pixel values into a new rectangular buffer (potentially with a different stride).
+// This is typically used to copy RGBx data into an RGBx output buffer.
+void copyMatchedInterleavedFormats(unsigned width, unsigned height,
+                                   void* src, unsigned srcStridePixels,
+                                   void* dst, unsigned dstStridePixels,
+                                   unsigned pixelSize);
+
+#endif // EVS_VTS_FORMATCONVERT_H
diff --git a/automotive/evs/1.0/vts/functional/FrameHandler.cpp b/automotive/evs/1.0/vts/functional/FrameHandler.cpp
index 58c2f26..a69f72b 100644
--- a/automotive/evs/1.0/vts/functional/FrameHandler.cpp
+++ b/automotive/evs/1.0/vts/functional/FrameHandler.cpp
@@ -17,6 +17,7 @@
 #define LOG_TAG "VtsHalEvsTest"
 
 #include "FrameHandler.h"
+#include "FormatConvert.h"
 
 #include <stdio.h>
 #include <string.h>
@@ -25,14 +26,6 @@
 #include <cutils/native_handle.h>
 #include <ui/GraphicBuffer.h>
 
-#include <algorithm>    // std::min
-
-
-// For the moment, we're assuming that the underlying EVS driver we're working with
-// is providing 4 byte RGBx data.  This is fine for loopback testing, although
-// real hardware is expected to provide YUV data -- most likly formatted as YV12
-static const unsigned kBytesPerPixel = 4;   // assuming 4 byte RGBx pixels
-
 
 FrameHandler::FrameHandler(android::sp <IEvsCamera> pCamera, CameraDesc cameraInfo,
                            android::sp <IEvsDisplay> pDisplay,
@@ -58,14 +51,18 @@
 
 
 bool FrameHandler::startStream() {
+    // Tell the camera to start streaming
+    Return<EvsResult> result = mCamera->startVideoStream(this);
+    if (result != EvsResult::OK) {
+        return false;
+    }
+
     // Mark ourselves as running
     mLock.lock();
     mRunning = true;
     mLock.unlock();
 
-    // Tell the camera to start streaming
-    Return<EvsResult> result = mCamera->startVideoStream(this);
-    return (result == EvsResult::OK);
+    return true;
 }
 
 
@@ -82,7 +79,9 @@
 
     // Wait until the stream has actually stopped
     std::unique_lock<std::mutex> lock(mLock);
-    mSignal.wait(lock, [this](){ return !mRunning; });
+    if (mRunning) {
+        mSignal.wait(lock, [this]() { return !mRunning; });
+    }
 }
 
 
@@ -179,13 +178,13 @@
 
         switch (mReturnMode) {
         case eAutoReturn:
-            // Send the camera buffer back now that we're done with it
+            // Send the camera buffer back now that the client has seen it
             ALOGD("Calling doneWithFrame");
             // TODO:  Why is it that we get a HIDL crash if we pass back the cloned buffer?
             mCamera->doneWithFrame(bufferArg);
             break;
         case eNoAutoReturn:
-            // Hang onto the buffer handle for now -- we'll return it explicitly later
+            // Hang onto the buffer handle for now -- the client will return it explicitly later
             mHeldBuffers.push(bufferArg);
         }
 
@@ -228,25 +227,41 @@
         srcBuffer.width, srcBuffer.height, srcBuffer.format, 1, srcBuffer.usage,
         srcBuffer.stride);
 
-    // Lock our source buffer for reading
-    unsigned char* srcPixels = nullptr;
+    // Lock our source buffer for reading (current expectation are for this to be NV21 format)
+    uint8_t* srcPixels = nullptr;
     src->lock(GRALLOC_USAGE_SW_READ_OFTEN, (void**)&srcPixels);
 
-    // Lock our target buffer for writing
-    unsigned char* tgtPixels = nullptr;
+    // Lock our target buffer for writing (should be RGBA8888 format)
+    uint32_t* tgtPixels = nullptr;
     tgt->lock(GRALLOC_USAGE_SW_WRITE_OFTEN, (void**)&tgtPixels);
 
     if (srcPixels && tgtPixels) {
-        for (unsigned row = 0; row < height; row++) {
-            // Copy the entire row of pixel data
-            memcpy(tgtPixels, srcPixels, width * kBytesPerPixel);
-
-            // Advance to the next row (keeping in mind that stride here is in units of pixels)
-            tgtPixels += tgtBuffer.stride * kBytesPerPixel;
-            srcPixels += srcBuffer.stride * kBytesPerPixel;
+        if (tgtBuffer.format != HAL_PIXEL_FORMAT_RGBA_8888) {
+            // We always expect 32 bit RGB for the display output for now.  Is there a need for 565?
+            ALOGE("Diplay buffer is always expected to be 32bit RGBA");
+            success = false;
+        } else {
+            if (srcBuffer.format == HAL_PIXEL_FORMAT_YCRCB_420_SP) {   // 420SP == NV21
+                copyNV21toRGB32(width, height,
+                                srcPixels,
+                                tgtPixels, tgtBuffer.stride);
+            } else if (srcBuffer.format == HAL_PIXEL_FORMAT_YV12) { // YUV_420P == YV12
+                copyYV12toRGB32(width, height,
+                                srcPixels,
+                                tgtPixels, tgtBuffer.stride);
+            } else if (srcBuffer.format == HAL_PIXEL_FORMAT_YCBCR_422_I) { // YUYV
+                copyYUYVtoRGB32(width, height,
+                                srcPixels, srcBuffer.stride,
+                                tgtPixels, tgtBuffer.stride);
+            } else if (srcBuffer.format == tgtBuffer.format) {  // 32bit RGBA
+                copyMatchedInterleavedFormats(width, height,
+                                              srcPixels, srcBuffer.stride,
+                                              tgtPixels, tgtBuffer.stride,
+                                              tgtBuffer.pixelSize);
+            }
         }
     } else {
-        ALOGE("Failed to copy buffer contents");
+        ALOGE("Failed to lock buffer contents for contents transfer");
         success = false;
     }
 
diff --git a/automotive/evs/1.0/vts/functional/VtsEvsV1_0TargetTest.cpp b/automotive/evs/1.0/vts/functional/VtsEvsV1_0TargetTest.cpp
index 50b6581..2e80afe 100644
--- a/automotive/evs/1.0/vts/functional/VtsEvsV1_0TargetTest.cpp
+++ b/automotive/evs/1.0/vts/functional/VtsEvsV1_0TargetTest.cpp
@@ -107,6 +107,8 @@
  * call to closeCamera.  Then repeats the test to ensure all cameras can be reopened.
  */
 TEST_F(EvsHidlTest, CameraOpenClean) {
+    ALOGI("Starting CameraOpenClean test");
+
     // Get the camera list
     loadCameraList();
 
@@ -137,6 +139,8 @@
  * the system to be tolerant of shutdown/restart race conditions.
  */
 TEST_F(EvsHidlTest, CameraOpenAggressive) {
+    ALOGI("Starting CameraOpenAggressive test");
+
     // Get the camera list
     loadCameraList();
 
@@ -183,6 +187,8 @@
  * Test both clean shut down and "aggressive open" device stealing behavior.
  */
 TEST_F(EvsHidlTest, DisplayOpen) {
+    ALOGI("Starting DisplayOpen test");
+
     // Request exclusive access to the EVS display, then let it go
     {
         sp<IEvsDisplay> pDisplay = pEnumerator->openDisplay();
@@ -229,6 +235,8 @@
  * object itself or the owning enumerator.
  */
 TEST_F(EvsHidlTest, DisplayStates) {
+    ALOGI("Starting DisplayStates test");
+
     // Ensure the display starts in the expected state
     EXPECT_EQ((DisplayState)pEnumerator->getDisplayState(), DisplayState::NOT_OPEN);
 
@@ -270,15 +278,14 @@
     }
 
     // TODO:  This hack shouldn't be necessary.  b/36122635
-// NOTE:  Calling flushCommand here did not avoid the race.  Going back to sleep...  :(
-//    android::hardware::IPCThreadState::self()->flushCommands();
     sleep(1);
 
     // Now that the display pointer has gone out of scope, causing the IEvsDisplay interface
     // object to be destroyed, we should be back to the "not open" state.
     // NOTE:  If we want this to pass without the sleep above, we'd have to add the
     //        (now recommended) closeDisplay() call instead of relying on the smarter pointer
-    //        going out of scope.
+    //        going out of scope.  I've not done that because I want to verify that the deletion
+    //        of the object does actually clean up (eventually).
     EXPECT_EQ((DisplayState)pEnumerator->getDisplayState(), DisplayState::NOT_OPEN);
 }
 
@@ -288,6 +295,8 @@
  * Measure and qualify the stream start up time and streaming frame rate of each reported camera
  */
 TEST_F(EvsHidlTest, CameraStreamPerformance) {
+    ALOGI("Starting CameraStreamPerformance test");
+
     // Get the camera list
     loadCameraList();
 
@@ -304,7 +313,7 @@
         // Start the camera's video stream
         nsecs_t start = systemTime(SYSTEM_TIME_MONOTONIC);
         bool startResult = frameHandler->startStream();
-        EXPECT_EQ(startResult, true);
+        ASSERT_TRUE(startResult);
 
         // Ensure the first frame arrived within the expected time
         frameHandler->waitForFrameCount(1);
@@ -344,6 +353,8 @@
  * than one frame time.  The camera must cleanly skip frames until the client is ready again.
  */
 TEST_F(EvsHidlTest, CameraStreamBuffering) {
+    ALOGI("Starting CameraStreamBuffering test");
+
     // Arbitrary constant (should be > 1 and less than crazy)
     static const unsigned int kBuffersToHold = 6;
 
@@ -372,14 +383,14 @@
 
         // Start the camera's video stream
         bool startResult = frameHandler->startStream();
-        EXPECT_TRUE(startResult);
+        ASSERT_TRUE(startResult);
 
         // Check that the video stream stalls once we've gotten exactly the number of buffers
         // we requested since we told the frameHandler not to return them.
-        sleep(1);   // 1 second would be enough for at least 5 frames to be delivered worst case
+        sleep(2);   // 1 second should be enough for at least 5 frames to be delivered worst case
         unsigned framesReceived = 0;
         frameHandler->getFramesCounters(&framesReceived, nullptr);
-        EXPECT_EQ(kBuffersToHold, framesReceived);
+        ASSERT_EQ(kBuffersToHold, framesReceived) << "Stream didn't stall at expected buffer limit";
 
 
         // Give back one buffer
@@ -390,7 +401,7 @@
         // filled since we require 10fps minimum -- but give a 10% allowance just in case.
         usleep(110 * kMillisecondsToMicroseconds);
         frameHandler->getFramesCounters(&framesReceived, nullptr);
-        EXPECT_EQ(kBuffersToHold+1, framesReceived);
+        EXPECT_EQ(kBuffersToHold+1, framesReceived) << "Stream should've resumed";
 
         // Even when the camera pointer goes out of scope, the FrameHandler object will
         // keep the stream alive unless we tell it to shutdown.
@@ -411,6 +422,8 @@
  * which a human could observe to see the operation of the system on the physical display.
  */
 TEST_F(EvsHidlTest, CameraToDisplayRoundTrip) {
+    ALOGI("Starting CameraToDisplayRoundTrip test");
+
     // Get the camera list
     loadCameraList();
 
@@ -434,7 +447,7 @@
 
         // Start the camera's video stream
         bool startResult = frameHandler->startStream();
-        EXPECT_EQ(startResult, true);
+        ASSERT_TRUE(startResult);
 
         // Wait a while to let the data flow
         static const int kSecondsToWait = 5;