From 6ab94e0b318884bbcb95e2ea3835f951502e1d99 Mon Sep 17 00:00:00 2001 From: jaseg Date: Wed, 14 Oct 2020 12:47:28 +0200 Subject: Move firmware into subdirectory --- .../arm_signal_converge_data.c | 269 ++++++++++++ .../arm_signal_converge_example_f32.c | 259 ++++++++++++ .../ARM/arm_signal_converge_example/math_helper.c | 466 +++++++++++++++++++++ .../ARM/arm_signal_converge_example/math_helper.h | 63 +++ 4 files changed, 1057 insertions(+) create mode 100644 fw/midi-dials/Drivers/CMSIS/DSP/Examples/ARM/arm_signal_converge_example/arm_signal_converge_data.c create mode 100644 fw/midi-dials/Drivers/CMSIS/DSP/Examples/ARM/arm_signal_converge_example/arm_signal_converge_example_f32.c create mode 100644 fw/midi-dials/Drivers/CMSIS/DSP/Examples/ARM/arm_signal_converge_example/math_helper.c create mode 100644 fw/midi-dials/Drivers/CMSIS/DSP/Examples/ARM/arm_signal_converge_example/math_helper.h (limited to 'fw/midi-dials/Drivers/CMSIS/DSP/Examples/ARM/arm_signal_converge_example') diff --git a/fw/midi-dials/Drivers/CMSIS/DSP/Examples/ARM/arm_signal_converge_example/arm_signal_converge_data.c b/fw/midi-dials/Drivers/CMSIS/DSP/Examples/ARM/arm_signal_converge_example/arm_signal_converge_data.c new file mode 100644 index 0000000..3a2337d --- /dev/null +++ b/fw/midi-dials/Drivers/CMSIS/DSP/Examples/ARM/arm_signal_converge_example/arm_signal_converge_data.c @@ -0,0 +1,269 @@ +/* ---------------------------------------------------------------------- +* Copyright (C) 2010-2012 ARM Limited. All rights reserved. +* +* $Date: 17. January 2013 +* $Revision: V1.4.0 +* +* Project: CMSIS DSP Library +* Title: arm_signal_converge_data.c +* +* Description: Test input data for Floating point LMS Norm FIR filter +* +* Target Processor: Cortex-M4/Cortex-M3/Cortex-M0 +* +* Redistribution and use in source and binary forms, with or without +* modification, are permitted provided that the following conditions +* are met: +* - Redistributions of source code must retain the above copyright +* notice, this list of conditions and the following disclaimer. +* - Redistributions in binary form must reproduce the above copyright +* notice, this list of conditions and the following disclaimer in +* the documentation and/or other materials provided with the +* distribution. +* - Neither the name of ARM LIMITED nor the names of its contributors +* may be used to endorse or promote products derived from this +* software without specific prior written permission. +* +* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS +* FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE +* COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, +* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, +* BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; +* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER +* CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT +* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN +* ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +* POSSIBILITY OF SUCH DAMAGE. +* -------------------------------------------------------------------- */ + +#include "arm_math.h" + +/* ---------------------------------------------------------------------- +** Test input data for Floating point LMS Norm FIR filter +** Generated by the MATLAB randn() function +** ------------------------------------------------------------------- */ + +float32_t testInput_f32[1536] = +{ +-0.432565, -1.665584, 0.125332, 0.287676, -1.146471, 1.190915, 1.189164, -0.037633, +0.327292, 0.174639, -0.186709, 0.725791, -0.588317, 2.183186, -0.136396, 0.113931, +1.066768, 0.059281, -0.095648, -0.832349, 0.294411, -1.336182, 0.714325, 1.623562, +-0.691776, 0.857997, 1.254001, -1.593730, -1.440964, 0.571148, -0.399886, 0.689997, +0.815622, 0.711908, 1.290250, 0.668601, 1.190838, -1.202457, -0.019790, -0.156717, +-1.604086, 0.257304, -1.056473, 1.415141, -0.805090, 0.528743, 0.219321, -0.921902, +-2.170674, -0.059188, -1.010634, 0.614463, 0.507741, 1.692430, 0.591283, -0.643595, +0.380337, -1.009116, -0.019511, -0.048221, 0.000043, -0.317859, 1.095004, -1.873990, +0.428183, 0.895638, 0.730957, 0.577857, 0.040314, 0.677089, 0.568900, -0.255645, +-0.377469, -0.295887, -1.475135, -0.234004, 0.118445, 0.314809, 1.443508, -0.350975, +0.623234, 0.799049, 0.940890, -0.992092, 0.212035, 0.237882, -1.007763, -0.742045, +1.082295, -0.131500, 0.389880, 0.087987, -0.635465, -0.559573, 0.443653, -0.949904, +0.781182, 0.568961, -0.821714, -0.265607, -1.187777, -2.202321, 0.986337, -0.518635, +0.327368, 0.234057, 0.021466, -1.003944, -0.947146, -0.374429, -1.185886, -1.055903, +1.472480, 0.055744, -1.217317, -0.041227, -1.128344, -1.349278, -0.261102, 0.953465, +0.128644, 0.656468, -1.167819, -0.460605, -0.262440, -1.213152, -1.319437, 0.931218, +0.011245, -0.645146, 0.805729, 0.231626, -0.989760, 1.339586, 0.289502, 1.478917, +1.138028, -0.684139, -1.291936, -0.072926, -0.330599, -0.843628, 0.497770, 1.488490, +-0.546476, -0.846758, -0.246337, 0.663024, -0.854197, -1.201315, -0.119869, -0.065294, +0.485296, -0.595491, -0.149668, -0.434752, -0.079330, 1.535152, -0.606483, -1.347363, +0.469383, -0.903567, 0.035880, -0.627531, 0.535398, 0.552884, -0.203690, -2.054325, +0.132561, 1.592941, 1.018412, -1.580402, -0.078662, -0.681657, -1.024553, -1.234353, +0.288807, -0.429303, 0.055801, -0.367874, -0.464973, 0.370961, 0.728283, 2.112160, +-1.357298, -1.022610, 1.037834, -0.389800, -1.381266, 0.315543, 1.553243, 0.707894, +1.957385, 0.504542, 1.864529, -0.339812, -1.139779, -0.211123, 1.190245, -1.116209, +0.635274, -0.601412, 0.551185, -1.099840, 0.085991, -2.004563, -0.493088, 0.462048, +-0.321005, 1.236556, -0.631280, -2.325211, -1.231637, 1.055648, -0.113224, 0.379224, +0.944200, -2.120427, -0.644679, -0.704302, -1.018137, -0.182082, 1.521013, -0.038439, +1.227448, -0.696205, 0.007524, -0.782893, 0.586939, -0.251207, 0.480136, 0.668155, +-0.078321, 0.889173, 2.309287, 0.524639, -0.011787, 0.913141, 0.055941, -1.107070, +0.485498, -0.005005, -0.276218, 1.276452, 1.863401, -0.522559, 0.103424, -0.807649, +0.680439, -2.364590, 0.990115, 0.218899, 0.261662, 1.213444, -0.274667, -0.133134, +-1.270500, -1.663606, -0.703554, 0.280880, -0.541209, -1.333531, 1.072686, -0.712085, +-0.011286, -0.000817, -0.249436, 0.396575, -0.264013, -1.664011, -1.028975, 0.243095, +-1.256590, -0.347183, -0.941372, -1.174560, -1.021142, -0.401667, 0.173666, -0.116118, +1.064119, -0.245386, -1.517539, 0.009734, 0.071373, 0.316536, 0.499826, 1.278084, +-0.547816, 0.260808, -0.013177, -0.580264, 2.136308, -0.257617, -1.409528, 1.770101, +0.325546, -1.119040, 0.620350, 1.269782, -0.896043, 0.135175, -0.139040, -1.163395, +1.183720, -0.015430, 0.536219, -0.716429, -0.655559, 0.314363, 0.106814, 1.848216, +-0.275106, 2.212554, 1.508526, -1.945079, -1.680543, -0.573534, -0.185817, 0.008934, +0.836950, -0.722271, -0.721490, -0.201181, -0.020464, 0.278890, 1.058295, 0.621673, +-1.750615, 0.697348, 0.811486, 0.636345, 1.310080, 0.327098, -0.672993, -0.149327, +-2.449018, 0.473286, 0.116946, -0.591104, -0.654708, -1.080662, -0.047731, 0.379345, +-0.330361, -0.499898, -0.035979, -0.174760, -0.957265, 1.292548, 0.440910, 1.280941, +-0.497730, -1.118717, 0.807650, 0.041200, -0.756209, -0.089129, -2.008850, 1.083918, +-0.981191, -0.688489, 1.339479, -0.909243, -0.412858, -0.506163, 1.619748, 0.080901, +-1.081056, -1.124518, 1.735676, 1.937459, 1.635068, -1.255940, -0.213538, -0.198932, +0.307499, -0.572325, -0.977648, -0.446809, 1.082092, 2.372648, 0.229288, -0.266623, +0.701672, -0.487590, 1.862480, 1.106851, -1.227566, -0.669885, 1.340929, 0.388083, +0.393059, -1.707334, 0.227859, 0.685633, -0.636790, -1.002606, -0.185621, -1.054033, +-0.071539, 0.279198, 1.373275, 0.179841, -0.542017, 1.634191, 0.825215, 0.230761, +0.671634, -0.508078, 0.856352, 0.268503, 0.624975, -1.047338, 1.535670, 0.434426, +-1.917136, 0.469940, 1.274351, 0.638542, 1.380782, 1.319843, -0.909429, -2.305605, +1.788730, 0.390798, 0.020324, -0.405977, -1.534895, 0.221373, -1.374479, -0.839286, +-0.208643, 0.755913, 0.375734, -1.345413, 1.481876, 0.032736, 1.870453, -1.208991, +-0.782632, -0.767299, -0.107200, -0.977057, -0.963988, -2.379172, -0.838188, 0.257346, +-0.183834, -0.167615, -0.116989, 0.168488, -0.501206, -0.705076, 0.508165, -0.420922, +0.229133, -0.959497, -0.146043, 0.744538, -0.890496, 0.139062, -0.236144, -0.075459, +-0.358572, -2.077635, -0.143546, 1.393341, 0.651804, -0.377134, -0.661443, 0.248958, +-0.383516, -0.528480, 0.055388, 1.253769, -2.520004, 0.584856, -1.008064, 0.944285, +-2.423957, -0.223831, 0.058070, -0.424614, -0.202918, -1.513077, -1.126352, -0.815002, +0.366614, -0.586107, 1.537409, 0.140072, -1.862767, -0.454193, -0.652074, 0.103318, +-0.220632, -0.279043, -0.733662, -0.064534, -1.444004, 0.612340, -1.323503, -0.661577, +-0.146115, 0.248085, -0.076633, 1.738170, 1.621972, 0.626436, 0.091814, -0.807607, +-0.461337, -1.405969, -0.374530, -0.470911, 1.751296, 0.753225, 0.064989, -0.292764, +0.082823, 0.766191, 2.236850, 0.326887, 0.863304, 0.679387, 0.554758, 1.001630, +1.259365, 0.044151, -0.314138, 0.226708, 0.996692, 1.215912, -0.542702, 0.912228, +-0.172141, -0.335955, 0.541487, 0.932111, -0.570253, -1.498605, -0.050346, 0.553025, +0.083498, 1.577524, -0.330774, 0.795155, -0.784800, -1.263121, 0.666655, -1.392632, +-1.300562, -0.605022, -1.488565, 0.558543, -0.277354, -1.293685, -0.888435, -0.986520, +-0.071618, -2.414591, -0.694349, -1.391389, 0.329648, 0.598544, 0.147175, -0.101439, +-2.634981, 0.028053, -0.876310, -0.265477, -0.327578, -1.158247, 0.580053, 0.239756, +-0.350885, 0.892098, 1.578299, -1.108174, -0.025931, -1.110628, 0.750834, 0.500167, +-0.517261, -0.559209, -0.753371, 0.925813, -0.248520, -0.149835, -1.258415, 0.312620, +2.690277, 0.289696, -1.422803, 0.246786, -1.435773, 0.148573, -1.693073, 0.719188, +1.141773, 1.551936, 1.383630, -0.758092, 0.442663, 0.911098, -1.074086, 0.201762, +0.762863, -1.288187, -0.952962, 0.778175, -0.006331, 0.524487, 1.364272, 0.482039, +-0.787066, 0.751999, -0.166888, -0.816228, 2.094065, 0.080153, -0.937295, 0.635739, +1.682028, 0.593634, 0.790153, 0.105254, -0.158579, 0.870907, -0.194759, 0.075474, +-0.526635, -0.685484, -0.268388, -1.188346, 0.248579, 0.102452, -0.041007, -2.247582, +-0.510776, 0.249243, 0.369197, 0.179197, -0.037283, -1.603310, 0.339372, -0.131135, +0.485190, 0.598751, -0.086031, 0.325292, -0.335143, -0.322449, -0.382374, -0.953371, +0.233576, 1.235245, -0.578532, -0.501537, 0.722864, 0.039498, 1.541279, -1.701053, +-1.033741, -0.763708, 2.176426, 0.431612, -0.443765, 0.029996, -0.315671, 0.977846, +0.018295, 0.817963, 0.702341, -0.231271, -0.113690, 0.127941, -0.799410, -0.238612, +-0.089463, -1.023264, 0.937538, -1.131719, -0.710702, -1.169501, 1.065437, -0.680394, +-1.725773, 0.813200, 1.441867, 0.672272, 0.138665, -0.859534, -0.752251, 1.229615, +1.150754, -0.608025, 0.806158, 0.217133, -0.373461, -0.832030, 0.286866, -1.818892, +-1.573051, 2.015666, -0.071982, 2.628909, -0.243317, 0.173276, 0.923207, -0.178553, +-0.521705, 1.431962, -0.870117, 0.807542, -0.510635, 0.743514, 0.847898, -0.829901, +0.532994, 1.032848, -1.052024, 0.362114, -0.036787, -1.227636, -0.275099, -0.160435, +-1.083575, -1.954213, -0.909487, -0.005579, -1.723490, 1.263077, -0.600433, -2.063925, +0.110911, 1.487614, 0.053002, 0.161981, -0.026878, 0.173576, 0.882168, 0.182294, +0.755295, 0.508035, 0.131880, 0.280104, -0.982848, -0.944087, -0.013058, 0.354345, +-0.894709, 0.812111, 0.109537, 2.731644, 0.411079, -1.306862, 0.383806, 0.499504, +-0.510786, 0.234922, -0.597825, 0.020771, 0.419443, 1.191104, 0.771214, -2.644222, +0.285430, 0.826093, -0.008122, 0.858438, 0.774788, 1.305945, 1.231503, 0.958564, +-1.654548, -0.990396, 0.685236, -0.974870, -0.606726, 0.686794, 0.020049, 1.063801, +-1.341050, 0.479510, -1.633974, -1.442665, 0.293781, -0.140364, -1.130341, -0.292538, +-0.582536, -0.896348, 0.248601, -1.489663, 0.313509, -2.025084, 0.528990, 0.343471, +0.758193, -0.691940, 0.680179, -1.072541, 0.899772, -2.123092, 0.284712, -0.733323, +-0.773376, 0.151842, -0.336843, 0.970761, -0.107236, 1.013492, -0.475347, 0.068948, +0.398592, 1.116326, 0.620451, -0.287674, -1.371773, -0.685868, 0.331685, -0.997722, +0.291418, 1.107078, 0.244959, 0.164976, 0.406231, 1.215981, 1.448424, -1.025137, +0.205418, 0.588882, -0.264024, 2.495318, 0.855948, -0.850954, 0.811879, 0.700242, +0.759938, -1.712909, 1.537021, -1.609847, 1.109526, -1.109704, 0.385469, 0.965231, +0.818297, 0.037049, -0.926012, -0.111919, -0.803030, -1.665006, -0.901401, 0.588350, +0.554159, -0.415173, 0.061795, 0.457432, 0.199014, 0.257558, 2.080730, -2.277237, +0.339022, 0.289894, 0.662261, -0.580860, 0.887752, 0.171871, 0.848821, 0.963769, +1.321918, -0.064345, 1.317053, 0.228017, -1.429637, -0.149701, -0.504968, -1.729141, +-0.417472, -0.614969, 0.720777, 0.339364, 0.882845, 0.284245, -0.145541, -0.089646, +0.289161, 1.164831, 0.805729, -1.355643, 0.120893, -0.222178, 0.571732, -0.300140, +1.134277, -0.179356, -1.467067, 1.395346, 0.440836, 0.565384, -0.693623, 0.833869, +-2.237378, 1.097644, -0.001617, -1.614573, -1.228727, 0.207405, 0.220942, -1.006073, +-0.453067, 1.399453, -0.461964, 0.032716, 0.798783, 0.896816, 0.137892, -1.619146, +-1.646606, 0.428707, -0.737231, 0.564926, -1.384167, 0.460268, 0.629384, 0.379847, +-1.013330, -0.347243, 0.441912, -1.590240, -0.701417, -1.077601, 1.002220, 1.729481, +0.709032, -0.747897, 0.228862, -0.223497, -0.853275, 0.345627, 0.109764, -1.133039, +-0.683124, -0.277856, 0.654790, -1.248394, -0.597539, -0.481813, 0.983372, 1.762121, +1.427402, 0.911763, 0.326823, 0.069619, -1.499763, -0.418223, -0.021037, 0.228425, +-1.008196, -0.664622, 0.558177, -1.188542, -0.775481, 0.271042, 1.534976, -1.052283, +0.625559, -0.797626, -0.313522, -0.602210, 1.259060, 0.858484, -2.105292, -0.360937, +0.553557, -1.556384, -0.206666, -0.425568, 0.493778, -0.870908, 0.079828, -0.521619, +-1.413861, -0.384293, -0.457922, -0.291471, -0.301224, -1.588594, 1.094287, 1.324167, +-0.126480, -0.737164, 0.213719, -0.400529, 0.064938, -1.757996, 1.686748, 0.327400, +0.715967, 1.598648, -2.064741, -0.743632, 0.176185, 0.527839, -0.553153, 0.298280, +-1.226607, -0.189676, -0.301713, 0.956956, -0.533366, -0.901082, -0.892552, 0.278717, +-0.745807, 1.603464, 0.574270, 0.320655, -0.151383, 0.315762, 1.343703, -2.237832, +1.292906, -0.378459, 0.002521, 0.884641, 0.582450, -1.614244, -1.503666, 0.573586, +-0.910537, -1.631277, -0.359138, -0.397616, -1.161307, -1.109838, 0.290672, -1.910239, +1.314768, 0.665319, -0.275115, -0.023022, -0.907976, -1.043657, 0.373516, 0.901532, +1.278539, -0.128456, 0.612821, 1.956518, 2.266326, -0.373959, 2.238039, -0.159580, +-0.703281, 0.563477, -0.050296, 1.163593, 0.658808, -1.550089, -3.029118, 0.540578, +-1.008998, 0.908047, 1.582303, -0.979088, 1.007902, 0.158491, -0.586927, 1.574082, +-0.516649, 1.227800, 1.583876, -2.088950, 2.949545, 1.356125, 1.050068, -0.767170, +-0.257653, -1.371845, -1.267656, -0.894948, 0.589089, 1.842629, 1.347967, -0.491253, +-2.177568, 0.237000, -0.735411, -1.779419, 0.448030, 0.581214, 0.856607, -0.266263, +-0.417470, -0.205806, -0.174323, 0.217577, 1.684295, 0.119528, 0.650667, 2.080061, +-0.339225, 0.730113, 0.293969, -0.849109, -2.533858, -2.378941, -0.346276, -0.610937, +-0.408192, -1.415611, 0.227122, 0.207974, -0.719718, 0.757762, -1.643135, -1.056813, +-0.251662, -1.298441, 1.233255, 1.494625, 0.235938, -1.404359, 0.658791, -2.556613, +-0.534945, 3.202525, 0.439198, -1.149901, 0.886765, -0.283386, 1.035336, -0.364878, +1.341987, 1.008872, 0.213874, -0.299264, 0.255849, -0.190826, -0.079060, 0.699851, +-0.796540, -0.801284, -0.007599, -0.726810, -1.490902, 0.870335, -0.265675, -1.566695, +-0.394636, -0.143855, -2.334247, -1.357539, -1.815689, 1.108422, -0.142115, 1.112757, +0.559264, 0.478370, -0.679385, 0.284967, -1.332935, -0.723980, -0.663600, 0.198443, +-1.794868, -1.387673, 0.197768, 1.469328, 0.366493, -0.442775, -0.048563, 0.077709, +1.957910, -0.072848, 0.938810, -0.079608, -0.800959, 0.309424, 1.051826, -1.664211, +-1.090792, -0.191731, 0.463401, -0.924147, -0.649657, 0.622893, -1.335107, 1.047689, +0.863327, -0.642411, 0.660010, 1.294116, 0.314579, 0.859573, 0.128670, 0.016568, +-0.072801, -0.994310, -0.747358, -0.030814, 0.988355, -0.599017, 1.476644, -0.813801, +0.645040, -1.309919, -0.867425, -0.474233, 0.222417, 1.871323, 0.110001, -0.411341, +0.511242, -1.199117, -0.096361, 0.445817, -0.295825, -0.167996, 0.179543, 0.421118, +1.677678, 1.996949, 0.696964, -1.366382, 0.363045, -0.567044, -1.044154, 0.697139, +0.484026, -0.193751, -0.378095, -0.886374, -1.840197, -1.628195, -1.173789, -0.415411, +0.175088, 0.229433, -1.240889, 0.700004, 0.426877, 1.454803, -0.510186, -0.006657, +-0.525496, 0.717698, 1.088374, 0.500552, 2.771790, -0.160309, 0.429489, -1.966817, +-0.546019, -1.888395, -0.107952, -1.316144, -0.672632, -0.902365, -0.154798, 0.947242, +1.550375, 0.429040, -0.560795, 0.179304, -0.771509, -0.943390, -1.407569, -1.906131, +-0.065293, 0.672149, 0.206147, -0.008124, 0.020042, -0.558447, 1.886079, -0.219975, +-1.414395, -0.302811, -0.569574, -0.121495, -0.390171, -0.844287, -1.737757, -0.449520, +-1.547933, -0.095776, 0.907714, 2.369602, 0.519768, 0.410525, 1.052585, 0.428784, +1.295088, -0.186053, 0.130733, -0.657627, -0.759267, -0.595170, 0.812400, 0.069541, +-1.833687, 1.827363, 0.654075, -1.544769, -0.375109, 0.207688, -0.765615, -0.106355, +0.338769, 1.033461, -1.404822, -1.030570, -0.643372, 0.170787, 1.344839, 1.936273, +0.741336, 0.811980, -0.142808, -0.099858, -0.800131, 0.493249, 1.237574, 1.295951, +-0.278196, 0.217127, 0.630728, -0.548549, 0.229632, 0.355311, 0.521284, -0.615971, +1.345803, 0.974922, -2.377934, -1.092319, -0.325710, -2.012228, 1.567660, 0.233337, +0.646420, -1.129412, 0.197038, 1.696870, 0.726034, 0.792526, 0.603357, -0.058405, +-1.108666, 2.144229, -1.352821, 0.457021, 0.391175, 2.073013, -0.323318, 1.468132, +-0.502399, 0.209593, 0.754800, -0.948189, 0.613157, 1.760503, 0.088762, 2.595570, +-0.675470, 2.786804, -0.016827, 0.271651, -0.914102, -1.951371, -0.317418, 0.588333, +0.828996, -1.674851, -1.922293, -0.436662, 0.044974, 2.416609, -0.309892, 0.187583, +0.947699, -0.525703, -1.115605, -1.592320, 1.174844, 0.485144, 1.645480, -0.454233, +1.008768, 2.049403, 0.602020, 0.017860, -1.610426, 1.238752, 0.683587, -0.780716, +0.530979, 2.134498, 0.354361, 0.231700, 1.287980, -0.013488, -1.333345, -0.556343, +0.755597, -0.911854, 1.371684, 0.245580, 0.118845, 0.384690, -0.070152, -0.578309, +0.469308, 1.299687, 1.634798, -0.702809, 0.807253, -1.027451, 1.294496, 0.014930, +0.218705, 1.713188, -2.078805, 0.112917, -1.086491, -1.558311, 0.637406, -0.404576, +-0.403325, 0.084076, -0.435349, -0.562623, 0.878062, -0.814650, -0.258363, 0.493299, +-0.802694, -0.008329, 0.627571, 0.154382, 2.580735, -1.306246, 1.023526, 0.777795, +-0.833884, -0.586663, 0.065664, -0.012342, -0.076987, -1.558587, 1.702607, -0.468984, +0.094619, 0.287071, 0.919354, 0.510136, 0.245440, -1.400519, 0.969571, 1.593698, +-1.437917, -1.534230, -0.074710, 0.081459, -0.843240, -0.564640, -0.028207, -1.243702, +0.733039, 0.059580, 0.149144, 1.595857, -0.777250, 1.550277, 1.055002, -0.166654, +0.314484, 1.419571, 0.327348, 0.475653, 0.398754, -0.072770, 1.314784, 0.978279, +1.722114, -0.412302, 0.565133, 0.739851, 0.220138, 1.312807, 0.629152, -1.107987, +-0.447001, -0.725993, 0.354045, -0.506772, -2.103747, -0.664684, 1.450110, -0.329805, +2.701872, -1.634939, -0.536325, 0.547223, 1.492603, -0.455243, -0.496416, 1.235260, +0.040926, 0.748467, 1.230764, 0.304903, 1.077771, 0.765151, -1.319580, -0.509191, +0.555116, -1.957625, -0.760453, -2.443886, -0.659366, -0.114779, 0.300079, -0.583996, +-3.073745, 1.551042, -0.407369, 1.428095, -1.353242, 0.903970, 0.541671, -0.465020 +}; + + + +/* ---------------------------------------------------------------------- +** Coefficients for 32-tap filter for Floating point LMS FIR filter +* FIR high pass filter with cutoff freq 9.6kHz (transition 9.6KHz to 11.52KHz) +** ------------------------------------------------------------------- */ +float32_t lmsNormCoeff_f32[32] = { +-0.004240, 0.002301, 0.008860, -0.000000, -0.019782, -0.010543, 0.032881, 0.034736, +-0.037374, -0.069586, 0.022397, 0.102169, 0.014185, -0.115908, -0.061648, 0.101018, +0.101018, -0.061648, -0.115908, 0.014185, 0.102169, 0.022397, -0.069586, -0.037374, +0.034736, 0.032881, -0.010543, -0.019782, -0.000000, 0.008860, 0.002301, -0.004240 + +}; + +/* ---------------------------------------------------------------------- +** Coefficients for 32-tap filter for Floating point FIR filter +* FIR low pass filter with cutoff freq 24Hz (transition 24Hz to 240Hz) +** ------------------------------------------------------------------- */ +const float32_t FIRCoeff_f32[32] = { +0.004502, 0.005074, 0.006707, 0.009356, 0.012933, 0.017303, 0.022298, 0.027717, +0.033338, 0.038930, 0.044258, 0.049098, 0.053243, 0.056519, 0.058784, 0.059941, +0.059941, 0.058784, 0.056519, 0.053243, 0.049098, 0.044258, 0.038930, 0.033338, +0.027717, 0.022298, 0.017303, 0.012933, 0.009356, 0.006707, 0.005074, 0.004502 + +}; + diff --git a/fw/midi-dials/Drivers/CMSIS/DSP/Examples/ARM/arm_signal_converge_example/arm_signal_converge_example_f32.c b/fw/midi-dials/Drivers/CMSIS/DSP/Examples/ARM/arm_signal_converge_example/arm_signal_converge_example_f32.c new file mode 100644 index 0000000..d984e2f --- /dev/null +++ b/fw/midi-dials/Drivers/CMSIS/DSP/Examples/ARM/arm_signal_converge_example/arm_signal_converge_example_f32.c @@ -0,0 +1,259 @@ +/* ---------------------------------------------------------------------- +* Copyright (C) 2010-2012 ARM Limited. All rights reserved. +* +* $Date: 17. January 2013 +* $Revision: V1.4.0 +* +* Project: CMSIS DSP Library +* Title: arm_signal_converge_example_f32.c +* +* Description: Example code demonstrating convergence of an adaptive +* filter. +* +* Target Processor: Cortex-M4/Cortex-M3 +* +* Redistribution and use in source and binary forms, with or without +* modification, are permitted provided that the following conditions +* are met: +* - Redistributions of source code must retain the above copyright +* notice, this list of conditions and the following disclaimer. +* - Redistributions in binary form must reproduce the above copyright +* notice, this list of conditions and the following disclaimer in +* the documentation and/or other materials provided with the +* distribution. +* - Neither the name of ARM LIMITED nor the names of its contributors +* may be used to endorse or promote products derived from this +* software without specific prior written permission. +* +* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS +* FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE +* COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, +* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, +* BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; +* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER +* CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT +* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN +* ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +* POSSIBILITY OF SUCH DAMAGE. + * -------------------------------------------------------------------- */ + +/** + * @ingroup groupExamples + */ + +/** + * @defgroup SignalConvergence Signal Convergence Example + * + * \par Description: + * \par + * Demonstrates the ability of an adaptive filter to "learn" the transfer function of + * a FIR lowpass filter using the Normalized LMS Filter, Finite Impulse + * Response (FIR) Filter, and Basic Math Functions. + * + * \par Algorithm: + * \par + * The figure below illustrates the signal flow in this example. Uniformly distributed white + * noise is passed through an FIR lowpass filter. The output of the FIR filter serves as the + * reference input of the adaptive filter (normalized LMS filter). The white noise is input + * to the adaptive filter. The adaptive filter learns the transfer function of the FIR filter. + * The filter outputs two signals: (1) the output of the internal adaptive FIR filter, and + * (2) the error signal which is the difference between the adaptive filter and the reference + * output of the FIR filter. Over time as the adaptive filter learns the transfer function + * of the FIR filter, the first output approaches the reference output of the FIR filter, + * and the error signal approaches zero. + * \par + * The adaptive filter converges properly even if the input signal has a large dynamic + * range (i.e., varies from small to large values). The coefficients of the adaptive filter + * are initially zero, and then converge over 1536 samples. The internal function test_signal_converge() + * implements the stopping condition. The function checks if all of the values of the error signal have a + * magnitude below a threshold DELTA. + * + * \par Block Diagram: + * \par + * \image html SignalFlow.gif + * + * + * \par Variables Description: + * \par + * \li \c testInput_f32 points to the input data + * \li \c firStateF32 points to FIR state buffer + * \li \c lmsStateF32 points to Normalised Least mean square FIR filter state buffer + * \li \c FIRCoeff_f32 points to coefficient buffer + * \li \c lmsNormCoeff_f32 points to Normalised Least mean square FIR filter coefficient buffer + * \li \c wire1, wir2, wire3 temporary buffers + * \li \c errOutput, err_signal temporary error buffers + * + * \par CMSIS DSP Software Library Functions Used: + * \par + * - arm_lms_norm_init_f32() + * - arm_fir_init_f32() + * - arm_fir_f32() + * - arm_lms_norm_f32() + * - arm_scale_f32() + * - arm_abs_f32() + * - arm_sub_f32() + * - arm_min_f32() + * - arm_copy_f32() + * + * Refer + * \link arm_signal_converge_example_f32.c \endlink + * + */ + + +/** \example arm_signal_converge_example_f32.c + */ + +#include "arm_math.h" +#include "math_helper.h" + +/* ---------------------------------------------------------------------- +** Global defines for the simulation +* ------------------------------------------------------------------- */ + +#define TEST_LENGTH_SAMPLES 1536 +#define NUMTAPS 32 +#define BLOCKSIZE 32 +#define DELTA_ERROR 0.000001f +#define DELTA_COEFF 0.0001f +#define MU 0.5f + +#define NUMFRAMES (TEST_LENGTH_SAMPLES / BLOCKSIZE) + +/* ---------------------------------------------------------------------- +* Declare FIR state buffers and structure +* ------------------------------------------------------------------- */ + +float32_t firStateF32[NUMTAPS + BLOCKSIZE]; +arm_fir_instance_f32 LPF_instance; + +/* ---------------------------------------------------------------------- +* Declare LMSNorm state buffers and structure +* ------------------------------------------------------------------- */ + +float32_t lmsStateF32[NUMTAPS + BLOCKSIZE]; +float32_t errOutput[TEST_LENGTH_SAMPLES]; +arm_lms_norm_instance_f32 lmsNorm_instance; + + +/* ---------------------------------------------------------------------- +* Function Declarations for Signal Convergence Example +* ------------------------------------------------------------------- */ + +arm_status test_signal_converge_example( void ); + + +/* ---------------------------------------------------------------------- +* Internal functions +* ------------------------------------------------------------------- */ +arm_status test_signal_converge(float32_t* err_signal, + uint32_t blockSize); + +void getinput(float32_t* input, + uint32_t fr_cnt, + uint32_t blockSize); + +/* ---------------------------------------------------------------------- +* External Declarations for FIR F32 module Test +* ------------------------------------------------------------------- */ +extern float32_t testInput_f32[TEST_LENGTH_SAMPLES]; +extern float32_t lmsNormCoeff_f32[32]; +extern const float32_t FIRCoeff_f32[32]; +extern arm_lms_norm_instance_f32 lmsNorm_instance; + +/* ---------------------------------------------------------------------- +* Declare I/O buffers +* ------------------------------------------------------------------- */ + +float32_t wire1[BLOCKSIZE]; +float32_t wire2[BLOCKSIZE]; +float32_t wire3[BLOCKSIZE]; +float32_t err_signal[BLOCKSIZE]; + +/* ---------------------------------------------------------------------- +* Signal converge test +* ------------------------------------------------------------------- */ + +int32_t main(void) +{ + uint32_t i; + arm_status status; + uint32_t index; + float32_t minValue; + + /* Initialize the LMSNorm data structure */ + arm_lms_norm_init_f32(&lmsNorm_instance, NUMTAPS, lmsNormCoeff_f32, lmsStateF32, MU, BLOCKSIZE); + + /* Initialize the FIR data structure */ + arm_fir_init_f32(&LPF_instance, NUMTAPS, (float32_t *)FIRCoeff_f32, firStateF32, BLOCKSIZE); + + /* ---------------------------------------------------------------------- + * Loop over the frames of data and execute each of the processing + * functions in the system. + * ------------------------------------------------------------------- */ + + for(i=0; i < NUMFRAMES; i++) + { + /* Read the input data - uniformly distributed random noise - into wire1 */ + arm_copy_f32(testInput_f32 + (i * BLOCKSIZE), wire1, BLOCKSIZE); + + /* Execute the FIR processing function. Input wire1 and output wire2 */ + arm_fir_f32(&LPF_instance, wire1, wire2, BLOCKSIZE); + + /* Execute the LMS Norm processing function*/ + + arm_lms_norm_f32(&lmsNorm_instance, /* LMSNorm instance */ + wire1, /* Input signal */ + wire2, /* Reference Signal */ + wire3, /* Converged Signal */ + err_signal, /* Error Signal, this will become small as the signal converges */ + BLOCKSIZE); /* BlockSize */ + + /* apply overall gain */ + arm_scale_f32(wire3, 5, wire3, BLOCKSIZE); /* in-place buffer */ + } + + status = ARM_MATH_SUCCESS; + + /* ------------------------------------------------------------------------------- + * Test whether the error signal has reached towards 0. + * ----------------------------------------------------------------------------- */ + + arm_abs_f32(err_signal, err_signal, BLOCKSIZE); + arm_min_f32(err_signal, BLOCKSIZE, &minValue, &index); + + if (minValue > DELTA_ERROR) + { + status = ARM_MATH_TEST_FAILURE; + } + + /* ---------------------------------------------------------------------- + * Test whether the filter coefficients have converged. + * ------------------------------------------------------------------- */ + + arm_sub_f32((float32_t *)FIRCoeff_f32, lmsNormCoeff_f32, lmsNormCoeff_f32, NUMTAPS); + + arm_abs_f32(lmsNormCoeff_f32, lmsNormCoeff_f32, NUMTAPS); + arm_min_f32(lmsNormCoeff_f32, NUMTAPS, &minValue, &index); + + if (minValue > DELTA_COEFF) + { + status = ARM_MATH_TEST_FAILURE; + } + + /* ---------------------------------------------------------------------- + * Loop here if the signals did not pass the convergence check. + * This denotes a test failure + * ------------------------------------------------------------------- */ + + if ( status != ARM_MATH_SUCCESS) + { + while (1); + } + + while (1); /* main function does not return */ +} + + /** \endlink */ diff --git a/fw/midi-dials/Drivers/CMSIS/DSP/Examples/ARM/arm_signal_converge_example/math_helper.c b/fw/midi-dials/Drivers/CMSIS/DSP/Examples/ARM/arm_signal_converge_example/math_helper.c new file mode 100644 index 0000000..f615e6f --- /dev/null +++ b/fw/midi-dials/Drivers/CMSIS/DSP/Examples/ARM/arm_signal_converge_example/math_helper.c @@ -0,0 +1,466 @@ +/* ---------------------------------------------------------------------- +* Copyright (C) 2010-2012 ARM Limited. All rights reserved. +* +* $Date: 17. January 2013 +* $Revision: V1.4.0 b +* +* Project: CMSIS DSP Library +* +* Title: math_helper.c +* +* Description: Definition of all helper functions required. +* +* Target Processor: Cortex-M4/Cortex-M3 +* +* Redistribution and use in source and binary forms, with or without +* modification, are permitted provided that the following conditions +* are met: +* - Redistributions of source code must retain the above copyright +* notice, this list of conditions and the following disclaimer. +* - Redistributions in binary form must reproduce the above copyright +* notice, this list of conditions and the following disclaimer in +* the documentation and/or other materials provided with the +* distribution. +* - Neither the name of ARM LIMITED nor the names of its contributors +* may be used to endorse or promote products derived from this +* software without specific prior written permission. +* +* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS +* FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE +* COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, +* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, +* BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; +* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER +* CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT +* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN +* ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +* POSSIBILITY OF SUCH DAMAGE. +* -------------------------------------------------------------------- */ + +/* ---------------------------------------------------------------------- +* Include standard header files +* -------------------------------------------------------------------- */ +#include + +/* ---------------------------------------------------------------------- +* Include project header files +* -------------------------------------------------------------------- */ +#include "math_helper.h" + +/** + * @brief Caluclation of SNR + * @param[in] pRef Pointer to the reference buffer + * @param[in] pTest Pointer to the test buffer + * @param[in] buffSize total number of samples + * @return SNR + * The function Caluclates signal to noise ratio for the reference output + * and test output + */ + +float arm_snr_f32(float *pRef, float *pTest, uint32_t buffSize) +{ + float EnergySignal = 0.0, EnergyError = 0.0; + uint32_t i; + float SNR; + int temp; + int *test; + + for (i = 0; i < buffSize; i++) + { + /* Checking for a NAN value in pRef array */ + test = (int *)(&pRef[i]); + temp = *test; + + if (temp == 0x7FC00000) + { + return(0); + } + + /* Checking for a NAN value in pTest array */ + test = (int *)(&pTest[i]); + temp = *test; + + if (temp == 0x7FC00000) + { + return(0); + } + EnergySignal += pRef[i] * pRef[i]; + EnergyError += (pRef[i] - pTest[i]) * (pRef[i] - pTest[i]); + } + + /* Checking for a NAN value in EnergyError */ + test = (int *)(&EnergyError); + temp = *test; + + if (temp == 0x7FC00000) + { + return(0); + } + + + SNR = 10 * log10 (EnergySignal / EnergyError); + + return (SNR); + +} + + +/** + * @brief Provide guard bits for Input buffer + * @param[in,out] input_buf Pointer to input buffer + * @param[in] blockSize block Size + * @param[in] guard_bits guard bits + * @return none + * The function Provides the guard bits for the buffer + * to avoid overflow + */ + +void arm_provide_guard_bits_q15 (q15_t * input_buf, uint32_t blockSize, + uint32_t guard_bits) +{ + uint32_t i; + + for (i = 0; i < blockSize; i++) + { + input_buf[i] = input_buf[i] >> guard_bits; + } +} + +/** + * @brief Converts float to fixed in q12.20 format + * @param[in] pIn pointer to input buffer + * @param[out] pOut pointer to outputbuffer + * @param[in] numSamples number of samples in the input buffer + * @return none + * The function converts floating point values to fixed point(q12.20) values + */ + +void arm_float_to_q12_20(float *pIn, q31_t * pOut, uint32_t numSamples) +{ + uint32_t i; + + for (i = 0; i < numSamples; i++) + { + /* 1048576.0f corresponds to pow(2, 20) */ + pOut[i] = (q31_t) (pIn[i] * 1048576.0f); + + pOut[i] += pIn[i] > 0 ? 0.5 : -0.5; + + if (pIn[i] == (float) 1.0) + { + pOut[i] = 0x000FFFFF; + } + } +} + +/** + * @brief Compare MATLAB Reference Output and ARM Test output + * @param[in] pIn Pointer to Ref buffer + * @param[in] pOut Pointer to Test buffer + * @param[in] numSamples number of samples in the buffer + * @return maximum difference + */ + +uint32_t arm_compare_fixed_q15(q15_t *pIn, q15_t *pOut, uint32_t numSamples) +{ + uint32_t i; + int32_t diff, diffCrnt = 0; + uint32_t maxDiff = 0; + + for (i = 0; i < numSamples; i++) + { + diff = pIn[i] - pOut[i]; + diffCrnt = (diff > 0) ? diff : -diff; + + if (diffCrnt > maxDiff) + { + maxDiff = diffCrnt; + } + } + + return(maxDiff); +} + +/** + * @brief Compare MATLAB Reference Output and ARM Test output + * @param[in] pIn Pointer to Ref buffer + * @param[in] pOut Pointer to Test buffer + * @param[in] numSamples number of samples in the buffer + * @return maximum difference + */ + +uint32_t arm_compare_fixed_q31(q31_t *pIn, q31_t * pOut, uint32_t numSamples) +{ + uint32_t i; + int32_t diff, diffCrnt = 0; + uint32_t maxDiff = 0; + + for (i = 0; i < numSamples; i++) + { + diff = pIn[i] - pOut[i]; + diffCrnt = (diff > 0) ? diff : -diff; + + if (diffCrnt > maxDiff) + { + maxDiff = diffCrnt; + } + } + + return(maxDiff); +} + +/** + * @brief Provide guard bits for Input buffer + * @param[in,out] input_buf Pointer to input buffer + * @param[in] blockSize block Size + * @param[in] guard_bits guard bits + * @return none + * The function Provides the guard bits for the buffer + * to avoid overflow + */ + +void arm_provide_guard_bits_q31 (q31_t * input_buf, + uint32_t blockSize, + uint32_t guard_bits) +{ + uint32_t i; + + for (i = 0; i < blockSize; i++) + { + input_buf[i] = input_buf[i] >> guard_bits; + } +} + +/** + * @brief Provide guard bits for Input buffer + * @param[in,out] input_buf Pointer to input buffer + * @param[in] blockSize block Size + * @param[in] guard_bits guard bits + * @return none + * The function Provides the guard bits for the buffer + * to avoid overflow + */ + +void arm_provide_guard_bits_q7 (q7_t * input_buf, + uint32_t blockSize, + uint32_t guard_bits) +{ + uint32_t i; + + for (i = 0; i < blockSize; i++) + { + input_buf[i] = input_buf[i] >> guard_bits; + } +} + + + +/** + * @brief Caluclates number of guard bits + * @param[in] num_adds number of additions + * @return guard bits + * The function Caluclates the number of guard bits + * depending on the numtaps + */ + +uint32_t arm_calc_guard_bits (uint32_t num_adds) +{ + uint32_t i = 1, j = 0; + + if (num_adds == 1) + { + return (0); + } + + while (i < num_adds) + { + i = i * 2; + j++; + } + + return (j); +} + +/** + * @brief Apply guard bits to buffer + * @param[in,out] pIn pointer to input buffer + * @param[in] numSamples number of samples in the input buffer + * @param[in] guard_bits guard bits + * @return none + */ + +void arm_apply_guard_bits (float32_t *pIn, + uint32_t numSamples, + uint32_t guard_bits) +{ + uint32_t i; + + for (i = 0; i < numSamples; i++) + { + pIn[i] = pIn[i] * arm_calc_2pow(guard_bits); + } +} + +/** + * @brief Calculates pow(2, numShifts) + * @param[in] numShifts number of shifts + * @return pow(2, numShifts) + */ +uint32_t arm_calc_2pow(uint32_t numShifts) +{ + + uint32_t i, val = 1; + + for (i = 0; i < numShifts; i++) + { + val = val * 2; + } + + return(val); +} + + + +/** + * @brief Converts float to fixed q14 + * @param[in] pIn pointer to input buffer + * @param[out] pOut pointer to output buffer + * @param[in] numSamples number of samples in the buffer + * @return none + * The function converts floating point values to fixed point values + */ + +void arm_float_to_q14 (float *pIn, q15_t *pOut, uint32_t numSamples) +{ + uint32_t i; + + for (i = 0; i < numSamples; i++) + { + /* 16384.0f corresponds to pow(2, 14) */ + pOut[i] = (q15_t) (pIn[i] * 16384.0f); + + pOut[i] += pIn[i] > 0 ? 0.5 : -0.5; + + if (pIn[i] == (float) 2.0) + { + pOut[i] = 0x7FFF; + } + + } + +} + + +/** + * @brief Converts float to fixed q30 format + * @param[in] pIn pointer to input buffer + * @param[out] pOut pointer to output buffer + * @param[in] numSamples number of samples in the buffer + * @return none + * The function converts floating point values to fixed point values + */ + +void arm_float_to_q30 (float *pIn, q31_t * pOut, uint32_t numSamples) +{ + uint32_t i; + + for (i = 0; i < numSamples; i++) + { + /* 1073741824.0f corresponds to pow(2, 30) */ + pOut[i] = (q31_t) (pIn[i] * 1073741824.0f); + + pOut[i] += pIn[i] > 0 ? 0.5 : -0.5; + + if (pIn[i] == (float) 2.0) + { + pOut[i] = 0x7FFFFFFF; + } + } +} + +/** + * @brief Converts float to fixed q30 format + * @param[in] pIn pointer to input buffer + * @param[out] pOut pointer to output buffer + * @param[in] numSamples number of samples in the buffer + * @return none + * The function converts floating point values to fixed point values + */ + +void arm_float_to_q29 (float *pIn, q31_t *pOut, uint32_t numSamples) +{ + uint32_t i; + + for (i = 0; i < numSamples; i++) + { + /* 1073741824.0f corresponds to pow(2, 30) */ + pOut[i] = (q31_t) (pIn[i] * 536870912.0f); + + pOut[i] += pIn[i] > 0 ? 0.5 : -0.5; + + if (pIn[i] == (float) 4.0) + { + pOut[i] = 0x7FFFFFFF; + } + } +} + + +/** + * @brief Converts float to fixed q28 format + * @param[in] pIn pointer to input buffer + * @param[out] pOut pointer to output buffer + * @param[in] numSamples number of samples in the buffer + * @return none + * The function converts floating point values to fixed point values + */ + +void arm_float_to_q28 (float *pIn, q31_t *pOut, uint32_t numSamples) +{ + uint32_t i; + + for (i = 0; i < numSamples; i++) + { + /* 268435456.0f corresponds to pow(2, 28) */ + pOut[i] = (q31_t) (pIn[i] * 268435456.0f); + + pOut[i] += pIn[i] > 0 ? 0.5 : -0.5; + + if (pIn[i] == (float) 8.0) + { + pOut[i] = 0x7FFFFFFF; + } + } +} + +/** + * @brief Clip the float values to +/- 1 + * @param[in,out] pIn input buffer + * @param[in] numSamples number of samples in the buffer + * @return none + * The function converts floating point values to fixed point values + */ + +void arm_clip_f32 (float *pIn, uint32_t numSamples) +{ + uint32_t i; + + for (i = 0; i < numSamples; i++) + { + if (pIn[i] > 1.0f) + { + pIn[i] = 1.0; + } + else if ( pIn[i] < -1.0f) + { + pIn[i] = -1.0; + } + + } +} + + + + diff --git a/fw/midi-dials/Drivers/CMSIS/DSP/Examples/ARM/arm_signal_converge_example/math_helper.h b/fw/midi-dials/Drivers/CMSIS/DSP/Examples/ARM/arm_signal_converge_example/math_helper.h new file mode 100644 index 0000000..5a18734 --- /dev/null +++ b/fw/midi-dials/Drivers/CMSIS/DSP/Examples/ARM/arm_signal_converge_example/math_helper.h @@ -0,0 +1,63 @@ +/* ---------------------------------------------------------------------- +* Copyright (C) 2010-2013 ARM Limited. All rights reserved. +* +* $Date: 17. January 2013 +* $Revision: V1.4.0 +* +* Project: CMSIS DSP Library +* +* Title: math_helper.h +* +* Description: Prototypes of all helper functions required. +* +* Target Processor: Cortex-M4/Cortex-M3 +* +* Redistribution and use in source and binary forms, with or without +* modification, are permitted provided that the following conditions +* are met: +* - Redistributions of source code must retain the above copyright +* notice, this list of conditions and the following disclaimer. +* - Redistributions in binary form must reproduce the above copyright +* notice, this list of conditions and the following disclaimer in +* the documentation and/or other materials provided with the +* distribution. +* - Neither the name of ARM LIMITED nor the names of its contributors +* may be used to endorse or promote products derived from this +* software without specific prior written permission. +* +* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS +* FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE +* COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, +* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, +* BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; +* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER +* CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT +* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN +* ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +* POSSIBILITY OF SUCH DAMAGE. +* -------------------------------------------------------------------- */ + + +#include "arm_math.h" + +#ifndef MATH_HELPER_H +#define MATH_HELPER_H + +float arm_snr_f32(float *pRef, float *pTest, uint32_t buffSize); +void arm_float_to_q12_20(float *pIn, q31_t * pOut, uint32_t numSamples); +void arm_provide_guard_bits_q15(q15_t *input_buf, uint32_t blockSize, uint32_t guard_bits); +void arm_provide_guard_bits_q31(q31_t *input_buf, uint32_t blockSize, uint32_t guard_bits); +void arm_float_to_q14(float *pIn, q15_t *pOut, uint32_t numSamples); +void arm_float_to_q29(float *pIn, q31_t *pOut, uint32_t numSamples); +void arm_float_to_q28(float *pIn, q31_t *pOut, uint32_t numSamples); +void arm_float_to_q30(float *pIn, q31_t *pOut, uint32_t numSamples); +void arm_clip_f32(float *pIn, uint32_t numSamples); +uint32_t arm_calc_guard_bits(uint32_t num_adds); +void arm_apply_guard_bits (float32_t * pIn, uint32_t numSamples, uint32_t guard_bits); +uint32_t arm_compare_fixed_q15(q15_t *pIn, q15_t * pOut, uint32_t numSamples); +uint32_t arm_compare_fixed_q31(q31_t *pIn, q31_t *pOut, uint32_t numSamples); +uint32_t arm_calc_2pow(uint32_t guard_bits); +#endif + -- cgit