diff --git a/Transceiver52M/arm/Makefile.am b/Transceiver52M/arm/Makefile.am
new file mode 100644
index 0000000..6d34daa
--- /dev/null
+++ b/Transceiver52M/arm/Makefile.am
@@ -0,0 +1,21 @@
+if ARCH_ARM
+if ARCH_ARM_A15
+ARCH_FLAGS = -mfpu=neon-vfpv4
+else
+ARCH_FLAGS = -mfpu=neon
+endif
+
+AM_CFLAGS = -Wall $(ARCH_FLAGS) -std=gnu99 -I../common
+AM_CCASFLAGS = $(ARCH_FLAGS)
+
+noinst_LTLIBRARIES = libarch.la
+
+libarch_la_SOURCES = \
+	../common/convolve_base.c \
+	convert.c \
+	convert_neon.S \
+	convolve.c \
+	convolve_neon.S \
+	scale.c \
+	scale_neon.S
+endif
diff --git a/Transceiver52M/arm/convert.c b/Transceiver52M/arm/convert.c
new file mode 100644
index 0000000..a32eb55
--- /dev/null
+++ b/Transceiver52M/arm/convert.c
@@ -0,0 +1,96 @@
+/*
+ * NEON type conversions
+ * Copyright (C) 2012, 2013 Thomas Tsou <tom@tsou.cc>
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <malloc.h>
+#include <string.h>
+#include "convert.h"
+
+#ifdef HAVE_CONFIG_H
+#include "config.h"
+#endif
+
+void neon_convert_ps_si16_4n(short *, float *, float *, int);
+void neon_convert_si16_ps_4n(float *, short *, int);
+
+#ifndef HAVE_NEON
+static void convert_si16_ps(float *out, short *in, int len)
+{
+	for (int i = 0; i < len; i++)
+		out[i] = in[i];
+}
+
+static void convert_ps_si16(short *out, float *in, float scale, int len)
+{
+	for (int i = 0; i < len; i++)
+		out[i] = in[i] * scale;
+}
+#else
+/* 4*N 16-bit signed integer conversion with remainder */
+static void neon_convert_si16_ps(float *restrict out,
+				 short *restrict in,
+				 int len)
+{
+	int start = len / 4 * 4;
+
+	neon_convert_si16_ps_4n(out, in, len >> 2);
+
+	for (int i = 0; i < len % 4; i++)
+		out[start + i] = (float) in[start + i];
+}
+
+/* 4*N 16-bit signed integer conversion with remainder */
+static void neon_convert_ps_si16(short *restrict out,
+				 float *restrict in,
+				 float *restrict scale,
+				 int len)
+{
+	int start = len / 4 * 4;
+
+	neon_convert_ps_si16_4n(out, in, scale, len >> 2);
+
+	for (int i = 0; i < len % 4; i++)
+		out[start + i] = (short) (in[start + i] * (*scale));
+}
+#endif
+
+void convert_float_short(short *out, float *in, float scale, int len)
+{
+#ifdef HAVE_NEON
+        float q[4] = { scale, scale, scale, scale };
+
+	if (len % 4)
+		neon_convert_ps_si16(out, in, q, len);
+	else
+		neon_convert_ps_si16_4n(out, in, q, len >> 2);
+#else
+	convert_ps_si16(out, in, scale, len);
+#endif
+}
+
+void convert_short_float(float *out, short *in, int len)
+{
+#ifdef HAVE_NEON
+	if (len % 4)
+		neon_convert_si16_ps(out, in, len);
+	else
+		neon_convert_si16_ps_4n(out, in, len >> 2);
+#else
+	convert_si16_ps(out, in, len);
+#endif
+}
diff --git a/Transceiver52M/arm/convert_neon.S b/Transceiver52M/arm/convert_neon.S
new file mode 100644
index 0000000..842ed9f
--- /dev/null
+++ b/Transceiver52M/arm/convert_neon.S
@@ -0,0 +1,51 @@
+/*
+ * NEON type conversions
+ * Copyright (C) 2012, 2013 Thomas Tsou <tom@tsou.cc>
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+	.syntax unified
+	.text
+	.align 2
+	.global neon_convert_ps_si16_4n
+	.type neon_convert_ps_si16_4n, %function
+neon_convert_ps_si16_4n:
+	vld1.32          {q1}, [r2]
+.loop_fltint:
+	vld1.64       {d0-d1}, [r1]!
+	vmul.f32           q0, q1
+	vcvt.s32.f32       q2, q0
+	vqmovn.s32         d0, q2
+	vst1.64          {d0}, [r0]!
+	subs               r3, #1
+	bne          .loop_fltint
+	bx                 lr
+	.size neon_convert_ps_si16_4n, .-neon_convert_ps_si16_4n
+	.text
+	.align 2
+	.global neon_convert_si16_ps_4n
+	.type neon_convert_si16_ps_4n, %function
+neon_convert_si16_ps_4n:
+.loop_intflt:
+	vld1.64          {d0}, [r1]!
+	vmovl.s16          q1, d0
+	vcvt.f32.s32       q0, q1
+	vst1.64          {q0}, [r0]!
+	subs               r2, #1
+	bne          .loop_intflt
+        bx                 lr
+	.size neon_convert_si16_ps_4n, .-neon_convert_si16_ps_4n
+	.section .note.GNU-stack,"",%progbits
diff --git a/Transceiver52M/arm/convolve.c b/Transceiver52M/arm/convolve.c
new file mode 100644
index 0000000..2b42090
--- /dev/null
+++ b/Transceiver52M/arm/convolve.c
@@ -0,0 +1,139 @@
+/*
+ * NEON Convolution
+ * Copyright (C) 2012, 2013 Thomas Tsou <tom@tsou.cc>
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <malloc.h>
+#include <string.h>
+#include <stdio.h>
+
+#ifdef HAVE_CONFIG_H
+#include "config.h"
+#endif
+
+/* Forward declarations from base implementation */
+int _base_convolve_real(float *x, int x_len,
+			float *h, int h_len,
+			float *y, int y_len,
+			int start, int len,
+			int step, int offset);
+
+int _base_convolve_complex(float *x, int x_len,
+			   float *h, int h_len,
+			   float *y, int y_len,
+			   int start, int len,
+			   int step, int offset);
+
+int bounds_check(int x_len, int h_len, int y_len,
+		 int start, int len, int step);
+
+#ifdef HAVE_NEON
+/* Calls into NEON assembler */
+void neon_conv_real4(float *x, float *h, float *y, int len);
+void neon_conv_real8(float *x, float *h, float *y, int len);
+void neon_conv_real12(float *x, float *h, float *y, int len);
+void neon_conv_real16(float *x, float *h, float *y, int len);
+void neon_conv_real20(float *x, float *h, float *y, int len);
+void mac_cx_neon4(float *x, float *h, float *y, int len);
+
+/* Complex-complex convolution */
+static void neon_conv_cmplx_4n(float *x, float *h, float *y, int h_len, int len)
+{
+	for (int i = 0; i < len; i++)
+		mac_cx_neon4(&x[2 * i], h, &y[2 * i], h_len >> 2);
+}
+#endif
+
+/* API: Aligned complex-real */
+int convolve_real(float *x, int x_len,
+		  float *h, int h_len,
+		  float *y, int y_len,
+		  int start, int len,
+		  int step, int offset)
+{
+	void (*conv_func)(float *, float *, float *, int) = NULL;
+
+	if (bounds_check(x_len, h_len, y_len, start, len, step) < 0)
+		return -1;
+
+	memset(y, 0, len * 2 * sizeof(float));
+
+#ifdef HAVE_NEON
+	if (step <= 4) {
+		switch (h_len) {
+		case 4:
+			conv_func = neon_conv_real4;
+			break;
+		case 8:
+			conv_func = neon_conv_real8;
+			break;
+		case 12:
+			conv_func = neon_conv_real12;
+			break;
+		case 16:
+			conv_func = neon_conv_real16;
+			break;
+		case 20:
+			conv_func = neon_conv_real20;
+			break;
+		}
+	}
+#endif
+	if (conv_func) {
+		conv_func(&x[2 * (-(h_len - 1) + start)],
+			  h, y, len);
+	} else {
+		_base_convolve_real(x, x_len,
+				    h, h_len,
+				    y, y_len,
+				    start, len, step, offset);
+	}
+
+	return len;
+}
+
+
+/* API: Aligned complex-complex */
+int convolve_complex(float *x, int x_len,
+		     float *h, int h_len,
+		     float *y, int y_len,
+		     int start, int len,
+		     int step, int offset)
+{
+	void (*conv_func)(float *, float *, float *, int, int) = NULL;
+
+	if (bounds_check(x_len, h_len, y_len, start, len, step) < 0)
+		return -1;
+
+	memset(y, 0, len * 2 * sizeof(float));
+
+#ifdef HAVE_NEON
+	if (step <= 4 && !(h_len % 4))
+		conv_func = neon_conv_cmplx_4n;
+#endif
+	if (conv_func) {
+		conv_func(&x[2 * (-(h_len - 1) + start)],
+			  h, y, h_len, len);
+	} else {
+		_base_convolve_complex(x, x_len,
+				       h, h_len,
+				       y, y_len,
+				       start, len, step, offset);
+	}
+
+	return len;
+}
diff --git a/Transceiver52M/arm/convolve_neon.S b/Transceiver52M/arm/convolve_neon.S
new file mode 100644
index 0000000..637d150
--- /dev/null
+++ b/Transceiver52M/arm/convolve_neon.S
@@ -0,0 +1,277 @@
+/*
+ * NEON Convolution
+ * Copyright (C) 2012, 2013 Thomas Tsou <tom@tsou.cc>
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifdef HAVE_CONFIG_H
+#include "config.h"
+#endif
+
+	.syntax unified
+	.text
+	.align 2
+	.global neon_conv_real4
+	.type neon_conv_real4, %function
+neon_conv_real4:
+	push            {r4, lr}
+	vpush       {q4-q7}
+	vld2.32     {q0-q1}, [r1]
+	ldr              r4, =8
+.neon_conv_loop4:
+	vld2.32     {q2-q3}, [r0], r4
+	vmul.f32         q4,   q2, q0
+	vmul.f32         q5,   q3, q0
+	vpadd.f32       d12,   d8, d9
+	vpadd.f32       d13,  d10, d11
+	vpadd.f32       d14,  d12, d13
+	vst1.64       {d14}, [r2]!
+	subs             r3,   r3, #1
+	bne          .neon_conv_loop4
+	vpop       {q4-q7}
+	pop             {r4, pc}
+	.size neon_conv_real4, .-neon_conv_real4
+	.align 2
+	.p2align 4,,15
+	.global neon_conv_real8
+	.type neon_conv_real8, %function
+neon_conv_real8:
+	push         {r4-r5, lr}
+	vpush        {q4-q7}
+	vld2.32     {q0-q1}, [r1]!
+	vld2.32     {q2-q3}, [r1]
+	add              r4, r0, #32
+	ldr              r5, =8
+.neon_conv_loop8:
+	vld2.32     {q4-q5}, [r0], r5
+	vld2.32     {q6-q7}, [r4], r5
+	vmul.f32         q8,   q4, q0
+	vmul.f32         q9,   q5, q0
+	vmul.f32        q10,   q6, q2
+	vmul.f32        q11,   q7, q2
+
+	vadd.f32        q12,   q8, q10
+	vadd.f32        q13,   q9, q11
+
+	vpadd.f32       d22,  d24, d25
+	vpadd.f32       d23,  d26, d27
+	vpadd.f32       d24,  d22, d23
+	vst1.64       {d24}, [r2]!
+	subs             r3,   r3, #1
+	bne          .neon_conv_loop8
+	vpop       {q4-q7}
+	pop          {r4-r5, pc}
+	.size neon_conv_real8, .-neon_conv_real8
+	.align 2
+	.global neon_conv_real12
+	.type neon_conv_real12, %function
+neon_conv_real12:
+	push         {r4-r6, lr}
+	vpush       {q4-q7}
+	vld2.32     {q0-q1}, [r1]!
+	vld2.32     {q2-q3}, [r1]!
+	vld2.32     {q4-q5}, [r1]!
+	add              r4, r0, #32
+	add              r5, r0, #64
+	ldr              r6, =8
+.neon_conv_loop12:
+	vld2.32     {q6-q7}, [r0], r6
+	vld2.32     {q8-q9}, [r4], r6
+	vld2.32   {q10-q11}, [r5], r6
+#ifdef HAVE_NEON_FMA
+	vfma.f32         q1,   q6, q0
+	vfma.f32         q3,   q7, q0
+	vfma.f32         q1,   q8, q2
+	vfma.f32         q3,   q9, q2
+	vfma.f32         q1,  q10, q4
+	vfma.f32         q3,  q11, q4
+#else
+	vmul.f32        q12,   q6, q0
+	vmul.f32        q13,   q7, q0
+	vmul.f32        q14,   q8, q2
+	vmul.f32        q15,   q9, q2
+	vmul.f32         q1,  q10, q4
+	vmul.f32         q3,  q11, q4
+
+	vadd.f32         q5,  q12, q14
+	vadd.f32         q6,  q13, q15
+	vadd.f32         q1,   q5, q1
+	vadd.f32         q3,   q6, q3
+#endif
+	vpadd.f32        d2,   d2, d3
+	vpadd.f32        d3,   d6, d7
+	vpadd.f32        d6,   d2, d3
+	vst1.64        {d6}, [r2]!
+	subs             r3,   r3, #1
+	bne          .neon_conv_loop12
+	vpop         {q4-q7}
+	pop          {r4-r6, pc}
+	.size neon_conv_real12, .-neon_conv_real12
+	.align 2
+	.global neon_conv_real16
+	.type neon_conv_real16, %function
+neon_conv_real16:
+	push         {r4-r7, lr}
+	vpush        {q4-q7}
+	vld2.32     {q0-q1}, [r1]!
+	vld2.32     {q2-q3}, [r1]!
+	vld2.32     {q4-q5}, [r1]!
+	vld2.32     {q6-q7}, [r1]
+	add              r4, r0, #32
+	add              r5, r0, #64
+	add              r6, r0, #96
+	ldr              r7, =8
+.neon_conv_loop16:
+	vld2.32     {q8-q9}, [r0], r7
+	vld2.32   {q10-q11}, [r4], r7
+	vld2.32   {q12-q13}, [r5], r7
+	vld2.32   {q14-q15}, [r6], r7
+#ifdef HAVE_NEON_FMA
+	vmul.f32         q1,   q8, q0
+	vmul.f32         q3,   q9, q0
+	vfma.f32         q1,  q10, q2
+	vfma.f32         q3,  q11, q2
+	vfma.f32         q1,  q12, q4
+	vfma.f32         q3,  q13, q4
+	vfma.f32         q1,  q14, q6
+	vfma.f32         q3,  q15, q6
+#else
+	vmul.f32         q1,   q8, q0
+	vmul.f32         q3,   q9, q0
+	vmul.f32         q5,  q10, q2
+	vmul.f32         q7,  q11, q2
+	vmul.f32         q8,  q12, q4
+	vmul.f32         q9,  q13, q4
+	vmul.f32        q10,  q14, q6
+	vmul.f32        q11,  q15, q6
+
+	vadd.f32         q1,   q1, q5
+	vadd.f32         q3,   q3, q7
+	vadd.f32         q5,   q8, q10
+	vadd.f32         q7,   q9, q11
+	vadd.f32         q1,   q1, q5
+	vadd.f32         q3,   q3, q7
+#endif
+	vpadd.f32        d2,   d2, d3
+	vpadd.f32        d3,   d6, d7
+	vpadd.f32        d6,   d2, d3
+	vst1.64        {d6}, [r2]!
+	subs             r3,   r3, #1
+	bne          .neon_conv_loop16
+	vpop       {q4-q7}
+	pop          {r4-r7, pc}
+	.size neon_conv_real16, .-neon_conv_real16
+	.align 2
+	.global neon_conv_real20
+	.type neon_conv_real20, %function
+neon_conv_real20:
+	push         {r4-r8, lr}
+	vpush        {q4-q7}
+	vld2.32     {q0-q1}, [r1]!
+	vld2.32     {q2-q3}, [r1]!
+	vld2.32     {q4-q5}, [r1]!
+	vld2.32     {q6-q7}, [r1]!
+	vld2.32     {q8-q9}, [r1]
+	add              r4, r0, #32
+	add              r5, r0, #64
+	add              r6, r0, #96
+	add              r7, r0, #128
+	ldr              r8, =8
+.neon_conv_loop20:
+	vld2.32   {q10-q11}, [r0], r8
+	vld2.32   {q12-q13}, [r4], r8
+	vld2.32   {q14-q15}, [r5], r8
+#ifdef HAVE_NEON_FMA
+	vmul.f32         q1,  q10, q0
+	vfma.f32         q1,  q12, q2
+	vfma.f32         q1,  q14, q4
+	vmul.f32         q3,  q11, q0
+	vfma.f32         q3,  q13, q2
+	vfma.f32         q3,  q15, q4
+
+	vld2.32   {q12-q13}, [r6], r8
+	vld2.32   {q14-q15}, [r7], r8
+
+	vfma.f32         q1,  q12, q6
+	vfma.f32         q3,  q13, q6
+	vfma.f32         q1,  q14, q8
+	vfma.f32         q3,  q15, q8
+#else
+	vmul.f32         q1,  q10, q0
+	vmul.f32         q3,  q12, q2
+	vmul.f32         q5,  q14, q4
+	vmul.f32         q7,  q11, q0
+	vmul.f32         q9,  q13, q2
+	vmul.f32        q10,  q15, q4
+	vadd.f32         q1,   q1, q3
+	vadd.f32         q3,   q7, q9
+	vadd.f32         q9,   q1, q5
+	vadd.f32        q10,   q3, q10
+
+	vld2.32   {q12-q13}, [r6], r8
+	vld2.32   {q14-q15}, [r7], r8
+
+	vmul.f32         q1,  q12, q6
+	vmul.f32         q3,  q13, q6
+	vmul.f32         q5,  q14, q8
+	vmul.f32         q7,  q15, q8
+	vadd.f32        q12,   q1, q9
+	vadd.f32        q14,   q3, q10
+	vadd.f32         q1,  q12, q5
+	vadd.f32         q3,  q14, q7
+#endif
+	vpadd.f32        d2,   d2, d3
+	vpadd.f32        d3,   d6, d7
+	vpadd.f32        d6,   d2, d3
+	vst1.64        {d6}, [r2]!
+	subs             r3,   r3, #1
+	bne          .neon_conv_loop20
+	vpop       {q4-q7}
+	pop          {r4-r8, pc}
+	.size neon_conv_real20, .-neon_conv_real20
+	.align 2
+	.global mac_cx_neon4
+	.type mac_cx_neon4, %function
+mac_cx_neon4:
+	push           {r4, lr}
+	ldr             r4, =32
+	veor           q14, q14
+	veor           q15, q15
+.neon_conv_loop_mac4:
+	vld2.32     {q0-q1}, [r0], r4
+	vld2.32     {q2-q3}, [r1]!
+
+	vmul.f32       q10,  q0, q2
+	vmul.f32       q11,  q1, q3
+	vmul.f32       q12,  q0, q3
+	vmul.f32       q13,  q2, q1
+	vsub.f32        q8, q10, q11
+	vadd.f32        q9, q12, q13
+
+	vadd.f32        q14, q8
+	vadd.f32        q15, q9
+	subs             r3, #1
+	bne       .neon_conv_loop_mac4
+
+	vld1.64          d0, [r2]
+	vpadd.f32       d28, d28, d29
+	vpadd.f32       d30, d30, d31
+	vpadd.f32        d1, d28, d30
+	vadd.f32         d1, d0
+	vst1.64          d1, [r2]
+        pop             {r4, pc}
+	.size mac_cx_neon4, .-mac_cx_neon4
+	.section .note.GNU-stack,"",%progbits
diff --git a/Transceiver52M/arm/scale.c b/Transceiver52M/arm/scale.c
new file mode 100644
index 0000000..2de13ff
--- /dev/null
+++ b/Transceiver52M/arm/scale.c
@@ -0,0 +1,56 @@
+/*
+ * NEON scaling
+ * Copyright (C) 2012,2013 Thomas Tsou <tom@tsou.cc>
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <malloc.h>
+#include <string.h>
+#include <scale.h>
+
+#ifdef HAVE_CONFIG_H
+#include "config.h"
+#endif
+
+void neon_scale_4n(float *, float *, float *, int);
+
+static void scale_ps(float *out, float *in, float *scale, int len)
+{
+	float ai, aq, bi, bq;
+
+	bi = scale[0];
+	bq = scale[1];
+
+	for (int i = 0; i < len; i++) {
+		ai = in[2 * i + 0];
+		aq = in[2 * i + 1];
+
+		out[2 * i + 0] = ai * bi - aq * bq;
+		out[2 * i + 1] = ai * bq + aq * bi;
+	}
+}
+
+void scale_complex(float *out, float *in, float* scale, int len)
+{
+#ifdef HAVE_NEON
+	if (len % 4)
+		scale_ps(out, in, scale, len);
+	else
+		neon_scale_4n(in, scale, out, len >> 2);
+#else
+	scale_ps(out, in, scale, len);
+#endif
+}
diff --git a/Transceiver52M/arm/scale_neon.S b/Transceiver52M/arm/scale_neon.S
new file mode 100644
index 0000000..a66fbe5
--- /dev/null
+++ b/Transceiver52M/arm/scale_neon.S
@@ -0,0 +1,50 @@
+/*
+ * ARM NEON Scaling
+ * Copyright (C) 2013 Thomas Tsou <tom@tsou.cc>
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+	.syntax unified
+	.text
+	.align 2
+	.global neon_scale_4n
+	.type neon_scale_4n, %function
+neon_scale_4n:
+	push            {r4, lr}
+	ldr              r4, =32
+
+	vld1.64          d0, [r1]
+	vmov.32          s4, s1
+	vmov.32          s1, s0
+	vmov.64          d1, d0
+	vmov.32          s5, s4
+	vmov.64          d3, d2
+.loop_mul_const:
+	vld2.32     {q2-q3}, [r0], r4
+
+	vmul.f32         q8, q0,  q2
+	vmul.f32         q9, q1,  q3
+	vmul.f32        q10, q0,  q3
+	vmul.f32        q11, q1,  q2
+	vsub.f32         q8, q8,  q9
+	vadd.f32         q9, q10, q11
+
+	vst2.32     {q8-q9}, [r2]!
+	subs             r3, #1
+	bne              .loop_mul_const
+	pop             {r4, pc}
+	.size neon_scale_4n, .-neon_scale_4n
+	.section .note.GNU-stack,"",%progbits