diff --git a/Audio/Chain/AudioChunk.h b/Audio/Chain/AudioChunk.h
index 5e9ac5f61..bc56986f2 100644
--- a/Audio/Chain/AudioChunk.h
+++ b/Audio/Chain/AudioChunk.h
@@ -84,6 +84,7 @@ enum {
 - (id)initWithProperties:(NSDictionary *)properties;
 
 - (void)assignSamples:(const void *)data frameCount:(size_t)count;
+- (void)assignData:(NSData *)data;
 
 - (NSData *)removeSamples:(size_t)frameCount;
 
diff --git a/Audio/Chain/AudioChunk.m b/Audio/Chain/AudioChunk.m
index e119e3597..679864b60 100644
--- a/Audio/Chain/AudioChunk.m
+++ b/Audio/Chain/AudioChunk.m
@@ -148,13 +148,19 @@ static const uint32_t AudioChannelConfigTable[] = {
 	}
 }
 
+- (void)assignData:(NSData *)data {
+	[chunkData appendData:data];
+}
+
 - (NSData *)removeSamples:(size_t)frameCount {
 	if(formatAssigned) {
-		const size_t bytesPerPacket = format.mBytesPerPacket;
-		const size_t byteCount = bytesPerPacket * frameCount;
-		NSData *ret = [chunkData subdataWithRange:NSMakeRange(0, byteCount)];
-		[chunkData replaceBytesInRange:NSMakeRange(0, byteCount) withBytes:NULL length:0];
-		return ret;
+		@autoreleasepool {
+			const size_t bytesPerPacket = format.mBytesPerPacket;
+			const size_t byteCount = bytesPerPacket * frameCount;
+			NSData *ret = [chunkData subdataWithRange:NSMakeRange(0, byteCount)];
+			[chunkData replaceBytesInRange:NSMakeRange(0, byteCount) withBytes:NULL length:0];
+			return ret;
+		}
 	}
 	return [NSData data];
 }
diff --git a/Audio/Chain/ChunkList.m b/Audio/Chain/ChunkList.m
index 4ccb88da4..04df44bcc 100644
--- a/Audio/Chain/ChunkList.m
+++ b/Audio/Chain/ChunkList.m
@@ -464,7 +464,7 @@ static void convert_be_to_le(uint8_t *buffer, size_t bitsPerSample, size_t bytes
 		AudioChunk *ret = [[AudioChunk alloc] init];
 		[ret setFormat:[chunk format]];
 		[ret setChannelConfig:[chunk channelConfig]];
-		[ret assignSamples:[removedData bytes] frameCount:maxFrameCount];
+		[ret assignData:removedData];
 		listDuration -= [ret duration];
 		inRemover = NO;
 		return ret;
diff --git a/Audio/Output/FSurroundFilter.mm b/Audio/Output/FSurroundFilter.mm
index 2b6f0e31b..24e02de82 100644
--- a/Audio/Output/FSurroundFilter.mm
+++ b/Audio/Output/FSurroundFilter.mm
@@ -126,6 +126,12 @@ struct freesurround_params {
 	freesurround_decoder *_decoder = (freesurround_decoder *)decoder;
 
 	float tempInput[4096 * 2];
+	uint32_t zeroCount = 0;
+
+	if(count > 4096) {
+		zeroCount = count - 4096;
+		count = 4096;
+	}
 
 	if(count < 4096) {
 		cblas_scopy(count * 2, samplesIn, 1, &tempInput[0], 1);
@@ -138,6 +144,9 @@ struct freesurround_params {
 	for(unsigned c = 0, num = channelCount; c < num; c++) {
 		unsigned idx = [AudioChunk channelIndexFromConfig:channelConfig forFlag:_params->chanmap[c]];
 		cblas_scopy(count, src + c, num, samplesOut + idx, num);
+		if(zeroCount) {
+			vDSP_vclr(samplesOut + idx + count, num, zeroCount);
+		}
 	}
 }
 
diff --git a/Audio/Output/OutputAVFoundation.h b/Audio/Output/OutputAVFoundation.h
index 7dc443e44..bbf9c64d4 100644
--- a/Audio/Output/OutputAVFoundation.h
+++ b/Audio/Output/OutputAVFoundation.h
@@ -114,6 +114,7 @@ using std::atomic_long;
 	FSurroundFilter *fsurround;
 
 	float *samplePtr;
+	float tempBuffer[512 * 32];
 	float inputBuffer[4096 * 32]; // 4096 samples times maximum supported channel count
 	float fsurroundBuffer[4096 * 6];
 	float hrtfBuffer[4096 * 2];
diff --git a/Audio/Output/OutputAVFoundation.m b/Audio/Output/OutputAVFoundation.m
index fd42b6ccc..9c7a9aca7 100644
--- a/Audio/Output/OutputAVFoundation.m
+++ b/Audio/Output/OutputAVFoundation.m
@@ -745,7 +745,10 @@ current_device_listener(AudioObjectID inObjectID, UInt32 inNumberAddresses, cons
 	int bytesRendered = 0;
 	do {
 		int maxToRender = MIN(4096 - inputRendered, 512);
-		int rendered = [self renderInput:maxToRender toBuffer:(float *)(((uint8_t *)inputBuffer) + bytesRendered)];
+		int rendered = [self renderInput:maxToRender toBuffer:&tempBuffer[0]];
+		if(rendered > 0) {
+			memcpy((((uint8_t *)inputBuffer) + bytesRendered), &tempBuffer[0], rendered * newFormat.mBytesPerPacket);
+		}
 		inputRendered += rendered;
 		bytesRendered += rendered * newFormat.mBytesPerPacket;
 		if(streamFormatChanged) {
@@ -886,6 +889,8 @@ current_device_listener(AudioObjectID inObjectID, UInt32 inNumberAddresses, cons
 		commandStop = NO;
 
 		audioFormatDescription = NULL;
+		
+		enableFSurround = YES;
 
 		running = NO;
 		stopping = NO;
diff --git a/Audio/ThirdParty/fsurround/freesurround_decoder.cpp b/Audio/ThirdParty/fsurround/freesurround_decoder.cpp
index bd38036fa..361dfb59b 100755
--- a/Audio/ThirdParty/fsurround/freesurround_decoder.cpp
+++ b/Audio/ThirdParty/fsurround/freesurround_decoder.cpp
@@ -63,19 +63,19 @@ class decoder_impl {
 	public:
 	// instantiate the decoder with a given channel setup and processing block size (in samples)
 	decoder_impl(channel_setup setup, unsigned N)
-	: N(N), log2N(log2n(N)), Nover2((N + 1) / 2),
+	: N(N), log2N(log2n(N)),
 	  wnd(N), inbuf(3 * N), setup(setup), C((unsigned)chn_alloc[setup].size()),
 	  buffer_empty(true), lt(N), rt(N), dst(N), dstf(N),
 	  dftsetupF(vDSP_DFT_zrop_CreateSetupD(0, N, vDSP_DFT_FORWARD)),
 	  dftsetupB(vDSP_DFT_zrop_CreateSetupD(0, N, vDSP_DFT_INVERSE)) {
-		_dsp_complexalloc(&lf, Nover2);
-		_dsp_complexalloc(&rf, Nover2);
+		_dsp_complexalloc(&lf, N/2 + 1);
+		_dsp_complexalloc(&rf, N/2 + 1);
 
 		// allocate per-channel buffers
 		outbuf.resize((N + N / 2) * C);
 		signal.resize(C);
 		for(unsigned k = 0; k < C; k++)
-			_dsp_complexalloc(&signal[k], N);
+			_dsp_complexalloc(&signal[k], N/2 + 1);
 
 		// init the window function
 		for(unsigned k = 0; k < N; k++)
@@ -209,8 +209,8 @@ class decoder_impl {
 		vDSP_vmulD(&rt[0], 1, &wnd[0], 1, &rt[0], 1, N);
 
 		// map into spectral domain
-		vDSP_ctozD((DSPDoubleComplex *)(&lt[0]), 2, &lf, 1, Nover2);
-		vDSP_ctozD((DSPDoubleComplex *)(&rt[0]), 2, &rf, 1, Nover2);
+		vDSP_ctozD((DSPDoubleComplex *)(&lt[0]), 2, &lf, 1, N / 2);
+		vDSP_ctozD((DSPDoubleComplex *)(&rt[0]), 2, &rf, 1, N / 2);
 
 		vDSP_DFT_ExecuteD(dftsetupF, lf.realp, lf.imagp, lf.realp, lf.imagp);
 		vDSP_DFT_ExecuteD(dftsetupF, rf.realp, rf.imagp, rf.realp, rf.imagp);
@@ -218,6 +218,8 @@ class decoder_impl {
 		for(unsigned c = 0; c < C; c++) {
 			signal[c].realp[0] = 0;
 			signal[c].imagp[0] = 0;
+			signal[c].realp[N/2] = 0;
+			signal[c].imagp[N/2] = 0;
 		}
 
 		// compute multichannel output signal in the spectral domain
@@ -253,7 +255,7 @@ class decoder_impl {
 			// map position to channel volumes
 			for(unsigned c = 0; c < C - 1; c++) {
 				// look up channel map at respective position (with bilinear interpolation) and build the signal
-				vector<float *> &a = chn_alloc[setup][c];
+				const vector<float *> &a = chn_alloc[setup][c];
 				polar(amp_total * ((1 - x) * (1 - y) * a[q][p] + x * (1 - y) * a[q][p + 1] + (1 - x) * y * a[q + 1][p] + x * y * a[q + 1][p + 1]),
 				      phase_of[1 + (int)sign(chn_xsf[setup][c])], signal[c], f);
 			}
@@ -275,14 +277,14 @@ class decoder_impl {
 		}
 
 		// shift the last 2/3 to the first 2/3 of the output buffer
-		memcpy(&outbuf[0], &outbuf[C * N / 2], N * C * 4);
+		memmove(&outbuf[0], &outbuf[C * N / 2], N * C * 4);
 		// and clear the rest
 		memset(&outbuf[C * N], 0, C * 4 * N / 2);
 		// backtransform each channel and overlap-add
 		for(unsigned c = 0; c < C; c++) {
 			// back-transform into time domain
 			vDSP_DFT_ExecuteD(dftsetupB, signal[c].realp, signal[c].imagp, signal[c].realp, signal[c].imagp);
-			vDSP_ztocD(&signal[c], 1, (DSPDoubleComplex *)(&dst[0]), 2, Nover2);
+			vDSP_ztocD(&signal[c], 1, (DSPDoubleComplex *)(&dst[0]), 2, N / 2);
 			// add the result to the last 2/3 of the output buffer, windowed (and remultiplex)
 			vDSP_vmulD(&dst[0], 1, &wnd[0], 1, &dst[0], 1, N);
 			vDSP_vdpsp(&dst[0], 1, &dstf[0], 1, N);
@@ -334,7 +336,7 @@ class decoder_impl {
 
 	// constants
 	unsigned N, C; // number of samples per input/output block, number of output channels
-	unsigned log2N, Nover2; // derivations of the block size
+	unsigned log2N; // derivations of the block size
 	channel_setup setup; // the channel setup
 
 	// parameters