diff --git a/src/HCADecodeService.cpp b/src/HCADecodeService.cpp
index 6685664..a78ff84 100644
--- a/src/HCADecodeService.cpp
+++ b/src/HCADecodeService.cpp
@@ -3,25 +3,20 @@
 #include "HCADecodeService.h"
 #include "clHCA.h"
 
-// Currently multithreaded decode is bugged, please use single thread for now
+// Currently multithreaded decode is bugged due to out of order decoding
 HCADecodeService::HCADecodeService(unsigned int numthreads)
-    : mainsem(numthreads),
+    : mainsem(1),
+      numthreads{1}, // Use single thread for now
       datasem{ 0 },
+      finsem{0},
       numchannels{0},
       workingrequest{nullptr},
       channels{nullptr},
       shutdown{false}
 {
-    if(numthreads > 0)
-    {
-        this->numthreads = numthreads;
-    }
-    else
-    {
-        this->numthreads = 1;
-    }
-    workingblocks = new int[numthreads];
-	for (unsigned int i = 0; i < numthreads; ++i)
+    workingblocks = new int[this->numthreads];
+    channels = new clHCA::stChannel[0x10 * this->numthreads];
+    for (unsigned int i = 0; i < this->numthreads; ++i)
 	{
 		workersem.emplace_back(0);
 		worker_threads.emplace_back(&HCADecodeService::Decode_Thread, this, i);
@@ -35,6 +30,7 @@ HCADecodeService::~HCADecodeService()
 	shutdown = true;
 	datasem.notify();
 	dispatchthread.join();
+    delete[] channels;
 }
 
 void HCADecodeService::cancel_decode(void* ptr)
@@ -59,6 +55,18 @@ void HCADecodeService::cancel_decode(void* ptr)
     }
 }
 
+void HCADecodeService::wait_for_finish()
+{
+    mutex.lock();
+    while(!filelist.empty() || !blocks.empty())
+    {
+        mutex.unlock();
+        finsem.wait();
+        mutex.lock();
+    }
+    mutex.unlock();
+}
+
 std::pair<void*, size_t> HCADecodeService::decode(const std::string& filename, DWORD samplenum)
 {
 	clHCA hca(0xBC731A85, 0x0002B875);
@@ -91,11 +99,11 @@ void HCADecodeService::Main_Thread()
 		workingfile = std::move(it->second);
         unsigned blocknum = it->first.second;
 		filelist.erase(it);
-		numchannels = workingfile.get_channelCount();
-		channels = new clHCA::stChannel[numchannels * numthreads];
+        numchannels = workingfile.get_channelCount();
 		workingfile.PrepDecode(channels, numthreads);
 		unsigned int blockCount = workingfile.get_blockCount();
-        for (unsigned int i = blocknum, j = 0; j < blockCount; ++i, ++j)
+        // initiate playback right away, and patch up "blip" cause by out of order decoding
+        for (unsigned int i = blocknum, j = 0; j < blockCount + blocknum; ++i, ++j)
 		{
             blocks.push_back(i % blockCount);
 		}
@@ -121,8 +129,8 @@ void HCADecodeService::Main_Thread()
         for (unsigned int i = 0; i < numthreads; ++i)
 		{
 			while (workingblocks[i] != -1); // busy wait until threads are finished
-		}
-		delete[] channels;
+        }
+        finsem.notify();
 	}
 	for (int i = 0; i < numthreads; ++i)
 	{
diff --git a/src/HCADecodeService.h b/src/HCADecodeService.h
index 94326cc..89f1cf7 100644
--- a/src/HCADecodeService.h
+++ b/src/HCADecodeService.h
@@ -14,7 +14,7 @@ class HCADecodeService
 	~HCADecodeService();
 	void cancel_decode(void* ptr);
     std::pair<void*, size_t> decode(const std::string& hcafilename, DWORD samplenum);
-	//void wait_for_finish();
+    void wait_for_finish();
 private:
 	void Decode_Thread(int id);
 	void Main_Thread();
@@ -27,7 +27,7 @@ class HCADecodeService
 	std::deque<unsigned int> blocks;
 	int* workingblocks;
 	std::deque<Semaphore> workersem;
-	Semaphore mainsem, datasem;
+    Semaphore mainsem, datasem, finsem;
 	std::mutex mutex;
 	clHCA::stChannel* channels;
 	bool shutdown;
diff --git a/src/clHCA.cpp b/src/clHCA.cpp
index 90d822b..1fad332 100644
--- a/src/clHCA.cpp
+++ b/src/clHCA.cpp
@@ -647,8 +647,9 @@ bool clHCA::Analyze(void*& wavptr, size_t& sz, const char* filenameHCA)
 void clHCA::AsyncDecode(stChannel* channelsOffset, unsigned int blocknum, unsigned char* outputwavptr)
 {
 	int seekhead = 0;
-	outputwavptr += 2 * blocknum * 8 * 128 * _channelCount + 44;
-	unsigned char* data = (unsigned char*)hcafileptr + (blocknum * _blockSize);
+    outputwavptr += 2 * blocknum * 8 * 128 * _channelCount + 44;
+    unsigned char* data = new unsigned char[_blockSize];
+    memcpy(data, (unsigned char*)hcafileptr + (blocknum * _blockSize), _blockSize);
 	//		if(((unsigned char *)data)[_blockSize-2]==0x5E)_asm int 3
 	_cipher.Mask(data, _blockSize);
 	clData d(data, _blockSize);
@@ -673,6 +674,7 @@ void clHCA::AsyncDecode(stChannel* channelsOffset, unsigned int blocknum, unsign
 			}
 		}
 	}
+    delete[] data;
 }
 
 //--------------------------------------------------
diff --git a/src/mainwindow.cpp b/src/mainwindow.cpp
index 42c7234..4ef0117 100644
--- a/src/mainwindow.cpp
+++ b/src/mainwindow.cpp
@@ -190,6 +190,7 @@ void MainWindow::setBGM(const QString& qStr)
     std::string name = qStr.toUtf8().constData();
     currSong = readablesong_to_filename[name];
     BASS_ChannelPause(mix_stream);
+    dec.wait_for_finish();
     BASS_Mixer_ChannelRemove(bgm->get_decode_channel());
     bgm->unload();
     bgm->load("res/" + currSong + "/bgm.hca");