snr_analysis_corr.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "88069857-9801-42e3-b106-287eced2aa62",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from numpy.fft import rfft,irfft\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "import pfb\n",
    "import conjugate_gradient as cg\n",
    "\n",
    "from datetime import datetime as dt\n",
    "import pickle\n",
    "import time"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "32ed67e1-cb19-4266-a8ce-33b29e74f1dd",
   "metadata": {},
   "source": [
    "So as not to overcomplicate life just yet, this notebook will only compare Wiener with no Wiener, we'll leave the CG optimization to the next notebook. We're also only doing 4bit quantization here. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "323bfaaf-f347-4796-a306-80105ec47fbd",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of up-channelized rows you'll get: 1017\n"
     ]
    }
   ],
   "source": [
    "# Constants\n",
    "N_EPOCHS       = 16\n",
    "LEN_EPOCH      = 1<<26 # 1<<26  # 1<<28 samples ~1.07 second's worth of data at 250 MSPS\n",
    "DELTA_4BIT     = 0.353  # Optimal delta for 15-level quantization\n",
    "NFRAME         = 2048   # 1<<11\n",
    "NTAP           = 4      # 1<<2\n",
    "UPCHAN_FACTOR  = 32     # 1<<5, higher frequency resolution for upchannelization\n",
    "TD_NOISE_SIGMA = 10     # Amount of noise added to time-domain signal\n",
    "PRNG_SEED      = 1747 #4 # 42#133599340\n",
    "WIENER_THRESH  = 0.1\n",
    "DROP_NROWS     = 4 # 10 # Number of up-channelized rows to drop to make sure we don't count the transient; make this number even\n",
    "\n",
    "CONSTS = {\n",
    "    \"N_EPOCHS\":N_EPOCHS,\n",
    "    \"LEN_EPOCH\":LEN_EPOCH,\n",
    "    \"DELTA_4BIT\":DELTA_4BIT,\n",
    "    \"NFRAME\":NFRAME,\n",
    "    \"NTAP\":NTAP,\n",
    "    \"UPCHAN_FACTOR\":UPCHAN_FACTOR,\n",
    "    \"TD_NOISE_SIGMA\":TD_NOISE_SIGMA,\n",
    "    \"PRNG_SEED\":PRNG_SEED,\n",
    "    \"WIENER_THRESH\":WIENER_THRESH,\n",
    "    \"DROP_NROWS\":DROP_NROWS\n",
    "}\n",
    "\n",
    "print(\"Number of up-channelized rows you'll get:\", (1<<int(.5+np.log2(LEN_EPOCH) - np.log2(NFRAME*UPCHAN_FACTOR))) - NTAP+1 - DROP_NROWS)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "fe21c240-245d-444b-920a-38a3c4986286",
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_rfft_spec(sig, nframe):\n",
    "    spec = np.reshape(sig, (len(sig)//nframe,nframe))\n",
    "    return rfft(spec, axis=1)\n",
    "\n",
    "def rechannelize(sig, quantize=False, usepfb=True, isupchan=False, \n",
    "                upchan_factor=4, wiener_thresh=0.0, drop_nrows=0, \n",
    "                only_upchan=False):\n",
    "    \"\"\"Channelize and optionally quantize and/or up-channelize a signal. \n",
    "    \n",
    "    Uses global constants (only capitalized variables). \n",
    "    \n",
    "    Parameters\n",
    "    ----------\n",
    "    sig : np.ndarray\n",
    "        1d numpy real, time-domain signal\n",
    "    quantize : bool\n",
    "        If true, quantize the signal to four bits real + four bits imaginary once channelized\n",
    "    usepfb : bool\n",
    "        If True, channelize with a PFB. If False, channelize with STFTs. \n",
    "    isupchan : bool\n",
    "        If True, up-channelize the signal by a factor of upchan_factor\n",
    "    upchan_factor : int\n",
    "        If upchan is True, the signal will be upchannelized by a factor of \n",
    "        If it's not a power of two this method might break. Just make it a power of two. \n",
    "    wiener_thresh : float\n",
    "        Wiener threshold. Set to 0 for no Wiener filtering. This is only relevant if \n",
    "        upchan is True. \n",
    "    only_upchan : bool\n",
    "        If this is True, only apply the higher-frequency channelization. Over-rides \n",
    "        quantization and inverse.\n",
    "\n",
    "    returns\n",
    "    -------\n",
    "    snr : np.ndarray\n",
    "        Signal to noise ratio in each channel, in dB\n",
    "    \"\"\"\n",
    "    if usepfb is True:\n",
    "        channelize = lambda x: pfb.forward_pfb(x, nchan=NFRAME//2+1, ntap=NTAP)\n",
    "        upchan     = lambda x: pfb.forward_pfb(x, nchan=(NFRAME * upchan_factor)//2+1, ntap=NTAP)\n",
    "        inverse    = lambda x: pfb.inverse_pfb(x, nchan=NFRAME//2+1, ntap=NTAP, \n",
    "                                               wiener_thresh=wiener_thresh)\n",
    "    else:\n",
    "        channelize = lambda x: get_rfft_spec(x, NFRAME)\n",
    "        upchan     = lambda x: get_rfft_spec(x, NFRAME * upchan_factor)\n",
    "        inverse    = lambda x: irfft(x, axis=1).flatten()\n",
    "    if only_upchan is True:\n",
    "        spec = upchan(sig)\n",
    "    else:\n",
    "        spec = channelize(sig)\n",
    "        if quantize is True:\n",
    "            std_spec = (np.std(np.real(spec)) + np.std(np.imag(spec)))/2 # for normalization\n",
    "            spec = pfb.quantize_8_bit(spec, delta=std_spec * DELTA_4BIT)\n",
    "        if isupchan is True:\n",
    "            spec = upchan(np.real(inverse(spec)))\n",
    "    if DROP_NROWS//2 > 0:\n",
    "        spec = spec[DROP_NROWS//2:-DROP_NROWS//2,:]\n",
    "    return spec\n",
    "\n",
    "def get_snr_corr(sig1, sig2, verbose_plot=False, **kwargs):\n",
    "    \"\"\"Estimate the Signal to Noise Ratio in each channel by correlating sig1 with sig2.\n",
    "    \n",
    "    Uses global constants (only capitalized variables). \n",
    "\n",
    "    Parameters\n",
    "    ----------\n",
    "    sig1 : np.ndarray\n",
    "        1d numpy real, time-domain signal with noise, to be correlated with sig2\n",
    "    sig2 : np.ndarray\n",
    "        1d numpy real, time-domain signal with noise, to be correlated with sig1\n",
    "    verbose_plot : bool\n",
    "        Defaults to False. If True, plot stuff. \n",
    "    **kwargs : dict\n",
    "        Arguments passed to rechannelize()\n",
    "\n",
    "    returns\n",
    "    -------\n",
    "    snr : np.ndarray\n",
    "        Signal to noise ratio in each channel, in dB\n",
    "    \"\"\"\n",
    "    spec1 = rechannelize(sig1, **kwargs)\n",
    "    spec2 = rechannelize(sig2, **kwargs)\n",
    "    corr = (spec1 * np.conj(spec2)).mean(axis=0)\n",
    "    autocorr1 = (abs(spec1)**2).mean(axis=0)\n",
    "    autocorr2 = (abs(spec2)**2).mean(axis=0)\n",
    "    s = np.real(corr)             # Signal\n",
    "    n = autocorr1 - np.real(corr) # Noise\n",
    "    n2= autocorr2 - np.real(corr)\n",
    "    # print(\"s\",s)\n",
    "    # print(\"n\",n)\n",
    "    snr = 10 * np.log10(s/n)      # Signal to Noise Ratio\n",
    "    info_string = \"{}{}{}{}\\nSNR = {:.2f}\".format(\"PFB\" if kwargs.get(\"usepfb\",False) else \"FFT\",\n",
    "        \", quantized\" if kwargs.get(\"quantize\",False) else \"\",\n",
    "        f\", upchannelized by {kwargs.get('upchan_factor','<default>')}\" if kwargs.get(\"isupchan\",False) else \"\",\n",
    "        f\", Wienered at {kwargs.get('wiener_thresh',False)}\" if kwargs.get(\"wiener_thresh\",False) else \"\",\n",
    "        snr.mean())\n",
    "    if verbose_plot is True:\n",
    "        plt.figure(figsize=(6,2))\n",
    "        plt.title(info_string,fontsize=9)\n",
    "        plt.plot(s,label=\"signal\")\n",
    "        plt.plot(n,label=f\"noise min={min(n)}\")\n",
    "        plt.plot(n2,label=f\"n2 min={min(n2)}\")\n",
    "        # print(f\"noise mean = {n.mean()}, std = {n.std()}\")\n",
    "        plt.legend()\n",
    "        plt.tight_layout()\n",
    "        plt.show()\n",
    "    else:\n",
    "        print(info_string)\n",
    "    return snr"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "25652d40-9b8a-4297-9474-b39dc9f71d5b",
   "metadata": {},
   "source": [
    "### Bench"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "2264466d-5165-43f6-a648-cec128ec3c11",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "156 ms ± 1.5 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)\n",
      "158 ms ± 2.71 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)\n",
      "698 ms ± 37.9 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)\n",
      "160 ms ± 2.27 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)\n"
     ]
    }
   ],
   "source": [
    "# Make Sig\n",
    "signal = np.random.randn(LEN_EPOCH // 8)\n",
    "%timeit signal = np.random.randn(LEN_EPOCH // 8)\n",
    "\n",
    "# Forward PFB (x2)\n",
    "spec = pfb.forward_pfb(signal, nchan=NFRAME//2+1, ntap=NTAP)\n",
    "%timeit pfb.forward_pfb(signal, nchan=NFRAME//2+1, ntap=NTAP)\n",
    "\n",
    "# Inverse PFB (x1)\n",
    "%timeit pfb.inverse_pfb(spec, nchan=NFRAME//2+1, ntap=NTAP, wiener_thresh=0.1)\n",
    "\n",
    "# Quantize (x1)\n",
    "std_spec = (np.std(np.real(spec)) + np.std(np.imag(spec)))/2 # for normalization\n",
    "spec_q = pfb.quantize_8_bit(spec, delta=std_spec * DELTA_4BIT)\n",
    "%timeit pfb.quantize_8_bit(spec, delta=std_spec * DELTA_4BIT)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b3a9ac44-c510-4ec9-a55a-7db96c5ef825",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "35c8e9e7-a12e-40e8-a795-1e0b63b62578",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "135a98c3-2b2b-4828-ae27-77140111fef0",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "38a80ea2-8aa4-46cb-9064-315a141553cd",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "c837366a-e00b-4984-ba23-8d6fec4ba8db",
   "metadata": {},
   "outputs": [],
   "source": [
    "# # Generate two timestreams, same underlying signal, different realizations\n",
    "# signal = np.random.randn(LEN_EPOCH * N_EPOCHS)\n",
    "# sig1 = signal + np.random.randn(LEN_EPOCH * N_EPOCHS) * TD_NOISE_SIGMA\n",
    "# sig2 = signal + np.random.randn(LEN_EPOCH * N_EPOCHS) * TD_NOISE_SIGMA"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2f4c0373-8af5-4f29-870a-ff60ad9c72b8",
   "metadata": {},
   "source": [
    "### FFT SNR analysis"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4fd2e88a-a10e-48ea-8f43-86b7251442cb",
   "metadata": {},
   "source": [
    "Without Quantization noise"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "7d09c88e-c658-4522-b871-724fae9eb81c",
   "metadata": {},
   "outputs": [],
   "source": [
    "# snr = get_snr_corr(sig1, sig2, quantize=False, usepfb=False, isupchan=False).mean()\n",
    "# snr = get_snr_corr(sig1, sig2, quantize=True, usepfb=False, isupchan=False).mean()\n",
    "# snr = get_snr_corr(sig1, sig2, quantize=False, usepfb=True, isupchan=False).mean()\n",
    "# snr = get_snr_corr(sig1, sig2, quantize=True, usepfb=True, isupchan=False).mean()\n",
    "\n",
    "# print()\n",
    "# snr = get_snr_corr(sig1, sig2, quantize=False, usepfb=False, isupchan=True, wiener_thresh=0.0).mean()\n",
    "# snr = get_snr_corr(sig1, sig2, quantize=True, usepfb=False, isupchan=True, wiener_thresh=0.0).mean()\n",
    "# snr = get_snr_corr(sig1, sig2, quantize=False, usepfb=True, isupchan=True, wiener_thresh=0.0).mean()\n",
    "# snr = get_snr_corr(sig1, sig2, quantize=True, usepfb=True, isupchan=True, upchan_factor=32, wiener_thresh=0.0).mean()\n",
    "\n",
    "# print()\n",
    "# snr = get_snr_corr(sig1, sig2, quantize=False, usepfb=False, isupchan=True, wiener_thresh=0.1).mean()\n",
    "# snr = get_snr_corr(sig1, sig2, quantize=True, usepfb=False, isupchan=True, wiener_thresh=0.1).mean()\n",
    "# snr = get_snr_corr(sig1, sig2, quantize=False, usepfb=True, isupchan=True, wiener_thresh=0.1).mean()\n",
    "# snr = get_snr_corr(sig1, sig2, quantize=True, usepfb=True, isupchan=True, upchan_factor=32, wiener_thresh=0.1).mean()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "89edd959-97e0-4929-9ffc-02f63f973564",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdin",
     "output_type": "stream",
     "text": [
      "Are you sure? [Enter] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Generating signal... took 2.619 seconds\n",
      "1/128 mean power FP precision 53245.9\t(4.3 s)\n",
      "1/128 mean power wiener filtered 50894.8\t(22.1 s)\n",
      "1/128 mean power no filter 45714.9\t(21.6 s)\n",
      "1/128 mean power CG 1% 49060.3\t(121.0 s)\n",
      "1/128 mean power CG 3% 49244.7\t(188.0 s)\n",
      "1/128 mean power CG 5% 49364.9\t(246.0 s)\n",
      "1/128 mean power CG 10% 49527.1\t(287.3 s)\n",
      "Generating signal... took 2.767 seconds\n",
      "2/128 mean power FP precision 53298.2\t(4.6 s)\n",
      "2/128 mean power wiener filtered 50750.7\t(19.4 s)\n",
      "2/128 mean power no filter 47103.4\t(18.8 s)\n",
      "2/128 mean power CG 1% 48997.3\t(111.2 s)\n",
      "2/128 mean power CG 3% 49163.2\t(180.7 s)\n",
      "2/128 mean power CG 5% 49264.4\t(247.3 s)\n",
      "2/128 mean power CG 10% 49334.9\t(287.1 s)\n",
      "Generating signal... took 2.692 seconds\n",
      "3/128 mean power FP precision 52020.7\t(4.5 s)\n",
      "3/128 mean power wiener filtered 49717.5\t(19.2 s)\n",
      "3/128 mean power no filter 52271.1\t(19.0 s)\n",
      "3/128 mean power CG 1% 47982.8\t(113.5 s)\n",
      "3/128 mean power CG 3% 48078.4\t(180.0 s)\n",
      "3/128 mean power CG 5% 48251.6\t(247.0 s)\n",
      "3/128 mean power CG 10% 48260.3\t(288.4 s)\n",
      "Generating signal... took 2.781 seconds\n",
      "4/128 mean power FP precision 51200.0\t(4.6 s)\n",
      "4/128 mean power wiener filtered 49062.4\t(19.4 s)\n",
      "4/128 mean power no filter 39151.0\t(18.8 s)\n",
      "4/128 mean power CG 1% 47356.1\t(115.5 s)\n",
      "4/128 mean power CG 3% 47532.7\t(180.2 s)\n",
      "4/128 mean power CG 5% 47624.8\t(247.7 s)\n",
      "4/128 mean power CG 10% 47703.7\t(288.1 s)\n",
      "Generating signal... took 2.703 seconds\n",
      "5/128 mean power FP precision 51277.3\t(4.5 s)\n",
      "5/128 mean power wiener filtered 49127.6\t(19.3 s)\n",
      "5/128 mean power no filter 54916.4\t(19.0 s)\n",
      "5/128 mean power CG 1% 47415.1\t(115.9 s)\n",
      "5/128 mean power CG 3% 47513.2\t(184.6 s)\n",
      "5/128 mean power CG 5% 47642.2\t(247.9 s)\n",
      "5/128 mean power CG 10% 47682.8\t(289.7 s)\n",
      "Generating signal... took 2.733 seconds\n",
      "6/128 mean power FP precision 51449.1\t(4.6 s)\n",
      "6/128 mean power wiener filtered 49425.7\t(19.3 s)\n",
      "6/128 mean power no filter 52079.4\t(18.9 s)\n",
      "6/128 mean power CG 1% 47600.1\t(114.0 s)\n",
      "6/128 mean power CG 3% 47752.5\t(179.5 s)\n",
      "6/128 mean power CG 5% 47842.9\t(246.4 s)\n",
      "6/128 mean power CG 10% 47850.5\t(291.1 s)\n",
      "Generating signal... took 2.715 seconds\n",
      "7/128 mean power FP precision 51089.9\t(4.5 s)\n",
      "7/128 mean power wiener filtered 48792.5\t(19.3 s)\n",
      "7/128 mean power no filter 54915.6\t(19.0 s)\n",
      "7/128 mean power CG 1% 47167.3\t(111.7 s)\n",
      "7/128 mean power CG 3% 47148.6\t(178.4 s)\n",
      "7/128 mean power CG 5% 47335.7\t(246.8 s)\n",
      "7/128 mean power CG 10% 47344.4\t(289.3 s)\n",
      "Generating signal... took 2.662 seconds\n",
      "8/128 mean power FP precision 50700.0\t(4.5 s)\n",
      "8/128 mean power wiener filtered 48374.3\t(19.3 s)\n",
      "8/128 mean power no filter 44431.8\t(18.9 s)\n",
      "8/128 mean power CG 1% 46669.9\t(113.3 s)\n",
      "8/128 mean power CG 3% 46856.8\t(179.2 s)\n",
      "8/128 mean power CG 5% 46877.5\t(247.7 s)\n",
      "8/128 mean power CG 10% 46943.7\t(288.4 s)\n",
      "Generating signal... took 2.693 seconds\n",
      "9/128 mean power FP precision 52963.6\t(4.5 s)\n",
      "9/128 mean power wiener filtered 50534.6\t(19.2 s)\n",
      "9/128 mean power no filter 60540.2\t(18.8 s)\n",
      "9/128 mean power CG 1% 48700.9\t(112.5 s)\n",
      "9/128 mean power CG 3% 48819.5\t(179.5 s)\n",
      "9/128 mean power CG 5% 48947.4\t(248.8 s)\n",
      "9/128 mean power CG 10% 48989.4\t(292.3 s)\n",
      "Generating signal... took 2.701 seconds\n",
      "10/128 mean power FP precision 51818.7\t(4.5 s)\n",
      "10/128 mean power wiener filtered 49360.6\t(19.3 s)\n",
      "10/128 mean power no filter 51350.2\t(18.9 s)\n",
      "10/128 mean power CG 1% 47668.1\t(117.3 s)\n",
      "10/128 mean power CG 3% 47824.9\t(181.5 s)\n",
      "10/128 mean power CG 5% 47852.2\t(252.7 s)\n",
      "10/128 mean power CG 10% 48039.0\t(294.1 s)\n",
      "Generating signal... took 2.669 seconds\n",
      "11/128 mean power FP precision 52395.5\t(4.5 s)\n",
      "11/128 mean power wiener filtered 50103.5\t(19.5 s)\n",
      "11/128 mean power no filter 48775.4\t(19.1 s)\n",
      "11/128 mean power CG 1% 48287.9\t(117.8 s)\n",
      "11/128 mean power CG 3% 48499.6\t(185.3 s)\n",
      "11/128 mean power CG 5% 48589.0\t(250.0 s)\n",
      "11/128 mean power CG 10% 48728.1\t(291.1 s)\n",
      "Generating signal... took 2.735 seconds\n",
      "12/128 mean power FP precision 51952.7\t(4.5 s)\n",
      "12/128 mean power wiener filtered 49606.5\t(20.0 s)\n",
      "12/128 mean power no filter 53558.2\t(19.4 s)\n",
      "12/128 mean power CG 1% 47844.4\t(119.2 s)\n",
      "12/128 mean power CG 3% 48020.6\t(189.0 s)\n",
      "12/128 mean power CG 5% 48185.0\t(255.1 s)\n",
      "12/128 mean power CG 10% 48163.8\t(295.2 s)\n",
      "Generating signal... took 2.661 seconds\n",
      "13/128 mean power FP precision 52918.1\t(4.6 s)\n",
      "13/128 mean power wiener filtered 50669.6\t(20.2 s)\n",
      "13/128 mean power no filter 39043.6\t(19.6 s)\n",
      "13/128 mean power CG 1% 48850.6\t(119.0 s)\n",
      "13/128 mean power CG 3% 48951.8\t(186.1 s)\n",
      "13/128 mean power CG 5% 49033.7\t(253.4 s)\n",
      "13/128 mean power CG 10% 49205.9\t(294.4 s)\n",
      "Generating signal... took 2.699 seconds\n",
      "14/128 mean power FP precision 51733.3\t(4.5 s)\n",
      "14/128 mean power wiener filtered 49417.5\t(20.2 s)\n",
      "14/128 mean power no filter 51881.1\t(19.3 s)\n",
      "14/128 mean power CG 1% 47710.7\t(120.5 s)\n",
      "14/128 mean power CG 3% 47767.4\t(187.3 s)\n",
      "14/128 mean power CG 5% 47880.3\t(256.0 s)\n",
      "14/128 mean power CG 10% 48006.5\t(297.4 s)\n",
      "Generating signal... took 2.700 seconds\n",
      "15/128 mean power FP precision 53232.7\t(4.5 s)\n",
      "15/128 mean power wiener filtered 50822.9\t(21.0 s)\n",
      "15/128 mean power no filter 47816.0\t(20.5 s)\n",
      "15/128 mean power CG 1% 49051.1\t(121.4 s)\n",
      "15/128 mean power CG 3% 49163.1\t(184.9 s)\n",
      "15/128 mean power CG 5% 49375.0\t(253.5 s)\n",
      "15/128 mean power CG 10% 49444.0\t(290.5 s)\n",
      "Generating signal... took 2.671 seconds\n",
      "16/128 mean power FP precision 51662.0\t(4.5 s)\n",
      "16/128 mean power wiener filtered 49385.7\t(19.5 s)\n",
      "16/128 mean power no filter 63631.7\t(19.6 s)\n",
      "16/128 mean power CG 1% 47680.6\t(116.5 s)\n",
      "16/128 mean power CG 3% 47752.8\t(187.0 s)\n",
      "16/128 mean power CG 5% 47838.5\t(254.1 s)\n",
      "16/128 mean power CG 10% 47959.5\t(293.0 s)\n",
      "Generating signal... took 2.723 seconds\n",
      "17/128 mean power FP precision 52597.9\t(4.5 s)\n",
      "17/128 mean power wiener filtered 50187.3\t(19.7 s)\n",
      "17/128 mean power no filter 44747.0\t(19.3 s)\n",
      "17/128 mean power CG 1% 48433.1\t(117.1 s)\n",
      "17/128 mean power CG 3% 48471.5\t(186.6 s)\n",
      "17/128 mean power CG 5% 48627.2\t(252.9 s)\n",
      "17/128 mean power CG 10% 48655.4\t(294.2 s)\n",
      "Generating signal... took 2.711 seconds\n",
      "18/128 mean power FP precision 52593.5\t(4.6 s)\n",
      "18/128 mean power wiener filtered 50112.2\t(20.3 s)\n",
      "18/128 mean power no filter 64481.1\t(19.2 s)\n",
      "18/128 mean power CG 1% 48366.0\t(118.1 s)\n",
      "18/128 mean power CG 3% 48520.3\t(185.8 s)\n",
      "18/128 mean power CG 5% 48655.2\t(255.2 s)\n",
      "18/128 mean power CG 10% 48687.0\t(294.4 s)\n",
      "Generating signal... took 2.795 seconds\n",
      "19/128 mean power FP precision 53032.0\t(4.6 s)\n",
      "19/128 mean power wiener filtered 50623.6\t(19.3 s)\n",
      "19/128 mean power no filter 54317.7\t(18.8 s)\n",
      "19/128 mean power CG 1% 48870.2\t(116.4 s)\n",
      "19/128 mean power CG 3% 49086.0\t(184.4 s)\n",
      "19/128 mean power CG 5% 49123.1\t(248.6 s)\n",
      "19/128 mean power CG 10% 49239.0\t(292.0 s)\n",
      "Generating signal... took 2.701 seconds\n",
      "20/128 mean power FP precision 52051.4\t(4.5 s)\n",
      "20/128 mean power wiener filtered 49431.5\t(19.4 s)\n",
      "20/128 mean power no filter 49862.1\t(19.3 s)\n",
      "20/128 mean power CG 1% 47798.7\t(117.1 s)\n",
      "20/128 mean power CG 3% 47954.1\t(184.3 s)\n",
      "20/128 mean power CG 5% 47981.2\t(249.2 s)\n",
      "20/128 mean power CG 10% 48091.6\t(295.7 s)\n",
      "Generating signal... took 2.695 seconds\n",
      "21/128 mean power FP precision 53516.2\t(4.5 s)\n",
      "21/128 mean power wiener filtered 51075.5\t(19.6 s)\n",
      "21/128 mean power no filter 55847.0\t(19.2 s)\n",
      "21/128 mean power CG 1% 49321.7\t(120.7 s)\n",
      "21/128 mean power CG 3% 49433.5\t(186.3 s)\n",
      "21/128 mean power CG 5% 49589.0\t(254.2 s)\n",
      "21/128 mean power CG 10% 49696.2\t(294.3 s)\n",
      "Generating signal... took 2.740 seconds\n",
      "22/128 mean power FP precision 52399.0\t(4.6 s)\n",
      "22/128 mean power wiener filtered 50013.6\t(20.3 s)\n",
      "22/128 mean power no filter 54739.2\t(19.8 s)\n",
      "22/128 mean power CG 1% 48334.2\t(121.0 s)\n",
      "22/128 mean power CG 3% 48396.5\t(185.8 s)\n",
      "22/128 mean power CG 5% 48467.5\t(254.1 s)\n",
      "22/128 mean power CG 10% 48523.4\t(294.6 s)\n",
      "Generating signal... took 2.731 seconds\n",
      "23/128 mean power FP precision 53041.7\t(4.6 s)\n",
      "23/128 mean power wiener filtered 50677.9\t(20.0 s)\n",
      "23/128 mean power no filter 46354.6\t(19.7 s)\n",
      "23/128 mean power CG 1% 48912.6\t(119.2 s)\n",
      "23/128 mean power CG 3% 49043.8\t(188.3 s)\n",
      "23/128 mean power CG 5% 49076.7\t(255.0 s)\n",
      "23/128 mean power CG 10% 49144.0\t(293.2 s)\n",
      "Generating signal... took 2.705 seconds\n",
      "24/128 mean power FP precision 52414.8\t(4.5 s)\n",
      "24/128 mean power wiener filtered 50113.3\t(20.7 s)\n",
      "24/128 mean power no filter 50002.6\t(19.8 s)\n",
      "24/128 mean power CG 1% 48301.8\t(119.5 s)\n",
      "24/128 mean power CG 3% 48463.7\t(188.6 s)\n",
      "24/128 mean power CG 5% 48577.7\t(256.1 s)\n",
      "24/128 mean power CG 10% 48583.7\t(293.2 s)\n",
      "Generating signal... took 2.720 seconds\n",
      "25/128 mean power FP precision 52079.0\t(4.5 s)\n",
      "25/128 mean power wiener filtered 49867.2\t(20.8 s)\n",
      "25/128 mean power no filter 41055.4\t(19.3 s)\n",
      "25/128 mean power CG 1% 48019.4\t(118.4 s)\n",
      "25/128 mean power CG 3% 48273.7\t(186.3 s)\n",
      "25/128 mean power CG 5% 48399.0\t(254.1 s)\n",
      "25/128 mean power CG 10% 48467.3\t(294.5 s)\n",
      "Generating signal... took 2.689 seconds\n",
      "26/128 mean power FP precision 51962.1\t(4.5 s)\n",
      "26/128 mean power wiener filtered 49643.4\t(20.6 s)\n",
      "26/128 mean power no filter 42542.2\t(21.5 s)\n",
      "26/128 mean power CG 1% 47914.2\t(120.3 s)\n",
      "26/128 mean power CG 3% 48003.9\t(187.8 s)\n",
      "26/128 mean power CG 5% 48134.2\t(254.7 s)\n",
      "26/128 mean power CG 10% 48134.4\t(294.7 s)\n",
      "Generating signal... took 2.733 seconds\n",
      "27/128 mean power FP precision 52610.0\t(4.5 s)\n",
      "27/128 mean power wiener filtered 50164.4\t(20.1 s)\n",
      "27/128 mean power no filter 38869.1\t(19.7 s)\n",
      "27/128 mean power CG 1% 48320.0\t(120.1 s)\n",
      "27/128 mean power CG 3% 48593.6\t(186.9 s)\n",
      "27/128 mean power CG 5% 48628.9\t(254.6 s)\n",
      "27/128 mean power CG 10% 48777.2\t(292.9 s)\n",
      "Generating signal... took 2.714 seconds\n",
      "28/128 mean power FP precision 51267.1\t(4.5 s)\n",
      "28/128 mean power wiener filtered 48899.0\t(19.3 s)\n",
      "28/128 mean power no filter 63295.6\t(19.3 s)\n",
      "28/128 mean power CG 1% 47148.4\t(118.1 s)\n",
      "28/128 mean power CG 3% 47277.0\t(187.0 s)\n",
      "28/128 mean power CG 5% 47351.4\t(254.3 s)\n",
      "28/128 mean power CG 10% 47414.0\t(296.9 s)\n",
      "Generating signal... took 2.687 seconds\n",
      "29/128 mean power FP precision 51368.0\t(4.6 s)\n",
      "29/128 mean power wiener filtered 49063.4\t(20.7 s)\n",
      "29/128 mean power no filter 59445.4\t(20.1 s)\n",
      "29/128 mean power CG 1% 47337.5\t(120.0 s)\n",
      "29/128 mean power CG 3% 47410.1\t(185.4 s)\n",
      "29/128 mean power CG 5% 47587.5\t(256.9 s)\n",
      "29/128 mean power CG 10% 47705.7\t(295.4 s)\n",
      "Generating signal... took 2.700 seconds\n",
      "30/128 mean power FP precision 51558.1\t(4.5 s)\n",
      "30/128 mean power wiener filtered 49103.9\t(20.3 s)\n",
      "30/128 mean power no filter 58406.9\t(19.3 s)\n",
      "30/128 mean power CG 1% 47411.4\t(121.5 s)\n",
      "30/128 mean power CG 3% 47585.5\t(186.1 s)\n",
      "30/128 mean power CG 5% 47645.5\t(254.8 s)\n",
      "30/128 mean power CG 10% 47818.6\t(294.3 s)\n",
      "Generating signal... took 2.732 seconds\n",
      "31/128 mean power FP precision 52602.2\t(4.6 s)\n",
      "31/128 mean power wiener filtered 50234.2\t(20.8 s)\n",
      "31/128 mean power no filter 57225.1\t(19.6 s)\n",
      "31/128 mean power CG 1% 48382.1\t(118.3 s)\n",
      "31/128 mean power CG 3% 48586.9\t(185.2 s)\n",
      "31/128 mean power CG 5% 48725.9\t(253.8 s)\n",
      "31/128 mean power CG 10% 48794.6\t(293.3 s)\n",
      "Generating signal... took 2.667 seconds\n",
      "32/128 mean power FP precision 51754.2\t(4.5 s)\n",
      "32/128 mean power wiener filtered 49458.0\t(19.8 s)\n",
      "32/128 mean power no filter 55979.4\t(19.7 s)\n",
      "32/128 mean power CG 1% 47694.8\t(118.1 s)\n",
      "32/128 mean power CG 3% 47926.3\t(185.4 s)\n",
      "32/128 mean power CG 5% 48001.0\t(252.2 s)\n",
      "32/128 mean power CG 10% 48011.9\t(292.7 s)\n",
      "Generating signal... took 2.670 seconds\n",
      "33/128 mean power FP precision 52328.0\t(4.5 s)\n",
      "33/128 mean power wiener filtered 50013.5\t(19.7 s)\n",
      "33/128 mean power no filter 56961.8\t(19.3 s)\n",
      "33/128 mean power CG 1% 48328.5\t(118.0 s)\n",
      "33/128 mean power CG 3% 48361.9\t(187.6 s)\n",
      "33/128 mean power CG 5% 48507.3\t(251.3 s)\n",
      "33/128 mean power CG 10% 48626.8\t(295.0 s)\n",
      "Generating signal... took 2.700 seconds\n",
      "34/128 mean power FP precision 51937.5\t(4.5 s)\n",
      "34/128 mean power wiener filtered 49659.2\t(19.6 s)\n",
      "34/128 mean power no filter 46019.5\t(19.3 s)\n",
      "34/128 mean power CG 1% 47903.4\t(117.6 s)\n",
      "34/128 mean power CG 3% 48049.4\t(182.9 s)\n",
      "34/128 mean power CG 5% 48113.7\t(252.7 s)\n",
      "34/128 mean power CG 10% 48258.3\t(293.0 s)\n",
      "Generating signal... took 2.739 seconds\n",
      "35/128 mean power FP precision 52612.1\t(4.6 s)\n",
      "35/128 mean power wiener filtered 50389.6\t(19.6 s)\n",
      "35/128 mean power no filter 54353.0\t(19.2 s)\n",
      "35/128 mean power CG 1% 48549.1\t(116.5 s)\n",
      "35/128 mean power CG 3% 48686.8\t(186.6 s)\n",
      "35/128 mean power CG 5% 48880.9\t(251.6 s)\n",
      "35/128 mean power CG 10% 48900.1\t(293.7 s)\n",
      "Generating signal... took 2.696 seconds\n",
      "36/128 mean power FP precision 51701.3\t(4.5 s)\n",
      "36/128 mean power wiener filtered 49325.8\t(19.7 s)\n",
      "36/128 mean power no filter 49386.7\t(19.1 s)\n",
      "36/128 mean power CG 1% 47606.2\t(118.7 s)\n",
      "36/128 mean power CG 3% 47697.7\t(185.4 s)\n",
      "36/128 mean power CG 5% 47787.9\t(254.9 s)\n",
      "36/128 mean power CG 10% 47936.2\t(294.3 s)\n",
      "Generating signal... took 2.696 seconds\n",
      "37/128 mean power FP precision 51057.5\t(4.5 s)\n",
      "37/128 mean power wiener filtered 48737.2\t(19.8 s)\n",
      "37/128 mean power no filter 47177.5\t(19.7 s)\n",
      "37/128 mean power CG 1% 47079.2\t(119.2 s)\n",
      "37/128 mean power CG 3% 47116.3\t(187.8 s)\n",
      "37/128 mean power CG 5% 47272.3\t(255.0 s)\n",
      "37/128 mean power CG 10% 47349.4\t(293.1 s)\n",
      "Generating signal... took 2.689 seconds\n",
      "38/128 mean power FP precision 52161.6\t(4.5 s)\n",
      "38/128 mean power wiener filtered 49823.3\t(20.8 s)\n",
      "38/128 mean power no filter 50884.1\t(21.1 s)\n",
      "38/128 mean power CG 1% 48068.0\t(120.5 s)\n",
      "38/128 mean power CG 3% 48202.2\t(187.9 s)\n",
      "38/128 mean power CG 5% 48262.9\t(251.4 s)\n",
      "38/128 mean power CG 10% 48338.5\t(295.0 s)\n",
      "Generating signal... took 2.669 seconds\n",
      "39/128 mean power FP precision 52918.3\t(4.5 s)\n",
      "39/128 mean power wiener filtered 50374.2\t(19.9 s)\n",
      "39/128 mean power no filter 51993.1\t(19.2 s)\n",
      "39/128 mean power CG 1% 48640.4\t(120.6 s)\n",
      "39/128 mean power CG 3% 48741.2\t(187.8 s)\n",
      "39/128 mean power CG 5% 48903.5\t(255.1 s)\n",
      "39/128 mean power CG 10% 49056.5\t(295.3 s)\n",
      "Generating signal... took 2.708 seconds\n",
      "40/128 mean power FP precision 53261.2\t(4.5 s)\n",
      "40/128 mean power wiener filtered 50965.3\t(20.0 s)\n",
      "40/128 mean power no filter 40491.4\t(19.5 s)\n",
      "40/128 mean power CG 1% 49118.4\t(119.0 s)\n",
      "40/128 mean power CG 3% 49290.1\t(189.2 s)\n",
      "40/128 mean power CG 5% 49389.3\t(257.0 s)\n",
      "40/128 mean power CG 10% 49396.8\t(294.2 s)\n",
      "Generating signal... took 2.697 seconds\n",
      "41/128 mean power FP precision 52584.7\t(4.6 s)\n",
      "41/128 mean power wiener filtered 50137.5\t(20.9 s)\n",
      "41/128 mean power no filter 53784.6\t(20.5 s)\n",
      "41/128 mean power CG 1% 48464.7\t(121.3 s)\n",
      "41/128 mean power CG 3% 48584.7\t(186.8 s)\n",
      "41/128 mean power CG 5% 48729.3\t(257.3 s)\n",
      "41/128 mean power CG 10% 48697.9\t(296.0 s)\n",
      "Generating signal... took 2.687 seconds\n",
      "42/128 mean power FP precision 51979.0\t(4.6 s)\n",
      "42/128 mean power wiener filtered 49471.1\t(20.9 s)\n",
      "42/128 mean power no filter 41450.5\t(20.1 s)\n",
      "42/128 mean power CG 1% 47758.8\t(120.0 s)\n",
      "42/128 mean power CG 3% 47899.1\t(187.1 s)\n",
      "42/128 mean power CG 5% 48101.0\t(254.9 s)\n",
      "42/128 mean power CG 10% 48137.6\t(294.8 s)\n",
      "Generating signal... took 2.685 seconds\n",
      "43/128 mean power FP precision 52759.0\t(4.5 s)\n",
      "43/128 mean power wiener filtered 50472.7\t(20.4 s)\n",
      "43/128 mean power no filter 40809.9\t(19.3 s)\n",
      "43/128 mean power CG 1% 48664.9\t(118.7 s)\n",
      "43/128 mean power CG 3% 48836.2\t(184.4 s)\n",
      "43/128 mean power CG 5% 48967.5\t(253.5 s)\n",
      "43/128 mean power CG 10% 49025.7\t(293.6 s)\n",
      "Generating signal... took 2.671 seconds\n",
      "44/128 mean power FP precision 52287.7\t(4.5 s)\n",
      "44/128 mean power wiener filtered 49886.8\t(19.8 s)\n",
      "44/128 mean power no filter 44683.4\t(19.5 s)\n",
      "44/128 mean power CG 1% 48149.7\t(119.0 s)\n",
      "44/128 mean power CG 3% 48284.7\t(185.3 s)\n",
      "44/128 mean power CG 5% 48437.2\t(254.7 s)\n",
      "44/128 mean power CG 10% 48586.8\t(295.0 s)\n",
      "Generating signal... took 2.670 seconds\n",
      "45/128 mean power FP precision 52017.5\t(4.5 s)\n",
      "45/128 mean power wiener filtered 49651.1\t(21.6 s)\n",
      "45/128 mean power no filter 40381.9\t(21.0 s)\n",
      "45/128 mean power CG 1% 47837.2\t(120.6 s)\n",
      "45/128 mean power CG 3% 47999.0\t(188.1 s)\n",
      "45/128 mean power CG 5% 48205.9\t(255.5 s)\n",
      "45/128 mean power CG 10% 48224.5\t(294.2 s)\n",
      "Generating signal... took 2.693 seconds\n",
      "46/128 mean power FP precision 52272.9\t(4.5 s)\n",
      "46/128 mean power wiener filtered 49881.8\t(20.0 s)\n",
      "46/128 mean power no filter 55184.6\t(19.5 s)\n",
      "46/128 mean power CG 1% 48166.0\t(121.2 s)\n",
      "46/128 mean power CG 3% 48315.0\t(188.1 s)\n",
      "46/128 mean power CG 5% 48392.3\t(252.5 s)\n",
      "46/128 mean power CG 10% 48481.4\t(292.5 s)\n",
      "Generating signal... took 2.695 seconds\n",
      "47/128 mean power FP precision 52377.6\t(4.5 s)\n",
      "47/128 mean power wiener filtered 50064.8\t(20.0 s)\n",
      "47/128 mean power no filter 46367.9\t(19.7 s)\n",
      "47/128 mean power CG 1% 48258.3\t(119.8 s)\n",
      "47/128 mean power CG 3% 48449.4\t(186.2 s)\n",
      "47/128 mean power CG 5% 48529.5\t(253.7 s)\n",
      "47/128 mean power CG 10% 48639.0\t(293.1 s)\n",
      "Generating signal... took 2.713 seconds\n",
      "48/128 mean power FP precision 51859.2\t(4.6 s)\n",
      "48/128 mean power wiener filtered 49645.0\t(20.2 s)\n",
      "48/128 mean power no filter 43835.0\t(19.8 s)\n",
      "48/128 mean power CG 1% 47934.6\t(120.7 s)\n",
      "48/128 mean power CG 3% 48085.3\t(187.0 s)\n",
      "48/128 mean power CG 5% 48203.9\t(255.9 s)\n",
      "48/128 mean power CG 10% 48236.2\t(294.5 s)\n",
      "Generating signal... took 2.697 seconds\n",
      "49/128 mean power FP precision 51504.8\t(4.5 s)\n",
      "49/128 mean power wiener filtered 49118.1\t(20.7 s)\n",
      "49/128 mean power no filter 46330.8\t(19.9 s)\n",
      "49/128 mean power CG 1% 47435.9\t(119.9 s)\n",
      "49/128 mean power CG 3% 47539.8\t(189.3 s)\n",
      "49/128 mean power CG 5% 47607.4\t(256.5 s)\n",
      "49/128 mean power CG 10% 47762.4\t(295.8 s)\n",
      "Generating signal... took 2.694 seconds\n",
      "50/128 mean power FP precision 51057.5\t(4.5 s)\n",
      "50/128 mean power wiener filtered 48607.7\t(20.6 s)\n",
      "50/128 mean power no filter 47955.8\t(20.2 s)\n",
      "50/128 mean power CG 1% 46807.2\t(121.6 s)\n",
      "50/128 mean power CG 3% 47056.4\t(187.2 s)\n",
      "50/128 mean power CG 5% 47098.9\t(256.3 s)\n",
      "50/128 mean power CG 10% 47177.5\t(296.1 s)\n",
      "Generating signal... took 2.671 seconds\n",
      "51/128 mean power FP precision 50715.4\t(4.5 s)\n",
      "51/128 mean power wiener filtered 48320.3\t(21.1 s)\n",
      "51/128 mean power no filter 74661.2\t(20.4 s)\n",
      "51/128 mean power CG 1% 46628.3\t(119.9 s)\n",
      "51/128 mean power CG 3% 46819.4\t(187.0 s)\n",
      "51/128 mean power CG 5% 46940.8\t(252.8 s)\n",
      "51/128 mean power CG 10% 47006.3\t(293.7 s)\n",
      "Generating signal... took 2.664 seconds\n",
      "52/128 mean power FP precision 52396.3\t(4.5 s)\n",
      "52/128 mean power wiener filtered 50148.7\t(20.6 s)\n",
      "52/128 mean power no filter 40084.9\t(20.7 s)\n",
      "52/128 mean power CG 1% 48389.6\t(120.7 s)\n",
      "52/128 mean power CG 3% 48486.1\t(185.9 s)\n",
      "52/128 mean power CG 5% 48566.7\t(253.6 s)\n",
      "52/128 mean power CG 10% 48649.2\t(294.3 s)\n",
      "Generating signal... took 2.681 seconds\n",
      "53/128 mean power FP precision 51854.8\t(4.5 s)\n",
      "53/128 mean power wiener filtered 49603.8\t(19.8 s)\n",
      "53/128 mean power no filter 39212.6\t(19.9 s)\n",
      "53/128 mean power CG 1% 47760.2\t(119.3 s)\n",
      "53/128 mean power CG 3% 47925.1\t(185.4 s)\n",
      "53/128 mean power CG 5% 48075.8\t(255.9 s)\n",
      "53/128 mean power CG 10% 48096.8\t(295.0 s)\n",
      "Generating signal... took 2.775 seconds\n",
      "54/128 mean power FP precision 52116.2\t(4.8 s)\n",
      "54/128 mean power wiener filtered 49789.8\t(21.4 s)\n",
      "54/128 mean power no filter 49742.5\t(20.6 s)\n",
      "54/128 mean power CG 1% 48065.0\t(122.3 s)\n",
      "54/128 mean power CG 3% 48206.6\t(189.6 s)\n",
      "54/128 mean power CG 5% 48355.2\t(254.6 s)\n",
      "54/128 mean power CG 10% 48472.3\t(295.1 s)\n",
      "Generating signal... took 2.793 seconds\n",
      "55/128 mean power FP precision 53705.1\t(4.8 s)\n",
      "55/128 mean power wiener filtered 51188.1\t(20.6 s)\n",
      "55/128 mean power no filter 37633.8\t(20.3 s)\n",
      "55/128 mean power CG 1% 49414.0\t(122.0 s)\n",
      "55/128 mean power CG 3% 49530.9\t(194.6 s)\n",
      "55/128 mean power CG 5% 49669.8\t(261.4 s)\n",
      "55/128 mean power CG 10% 49811.2\t(329.6 s)\n",
      "Generating signal... took 2.850 seconds\n",
      "56/128 mean power FP precision 51672.0\t(4.9 s)\n",
      "56/128 mean power wiener filtered 49228.4\t(30.1 s)\n",
      "56/128 mean power no filter 52133.5\t(33.4 s)\n",
      "56/128 mean power CG 1% 47468.4\t(136.0 s)\n",
      "56/128 mean power CG 3% 47625.1\t(184.3 s)\n",
      "56/128 mean power CG 5% 47719.0\t(251.4 s)\n"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "Cell \u001b[0;32mIn[7], line 80\u001b[0m\n\u001b[1;32m     78\u001b[0m conj_kwargs[\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mfrac_prior\u001b[39m\u001b[38;5;124m\"\u001b[39m] \u001b[38;5;241m=\u001b[39m frac_prior\n\u001b[1;32m     79\u001b[0m conj_kwargs[\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mnpersave\u001b[39m\u001b[38;5;124m\"\u001b[39m]   \u001b[38;5;241m=\u001b[39m npersave\n\u001b[0;32m---> 80\u001b[0m _, sig1_10 \u001b[38;5;241m=\u001b[39m \u001b[43mcg\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mconj_grad_with_prior\u001b[49m\u001b[43m(\u001b[49m\u001b[43mx\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43msig1\u001b[49m\u001b[43m,\u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mconj_kwargs\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m     81\u001b[0m _, sig2_10 \u001b[38;5;241m=\u001b[39m cg\u001b[38;5;241m.\u001b[39mconj_grad_with_prior(x\u001b[38;5;241m=\u001b[39msig2,\u001b[38;5;241m*\u001b[39m\u001b[38;5;241m*\u001b[39mconj_kwargs)\n\u001b[1;32m     82\u001b[0m spec1\u001b[38;5;241m=\u001b[39mrechannelize(sig1_10,\u001b[38;5;241m*\u001b[39m\u001b[38;5;241m*\u001b[39mkwargs_upchan_time_domain_sig)\n",
      "File \u001b[0;32m~/code/pfb-mod/conjugate_gradient.py:444\u001b[0m, in \u001b[0;36mconj_grad_with_prior\u001b[0;34m(x, frac_prior, delta, k, lblock, verbose, wiener_thresh, npersave, x0_wiener)\u001b[0m\n\u001b[1;32m    442\u001b[0m \u001b[38;5;66;03m# RMS conj gradient descent\u001b[39;00m\n\u001b[1;32m    443\u001b[0m saveas \u001b[38;5;241m=\u001b[39m \u001b[38;5;124mf\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mimg/RMSE_conjugate_gradient_descent_\u001b[39m\u001b[38;5;132;01m{\u001b[39;00m\u001b[38;5;241m100\u001b[39m\u001b[38;5;241m*\u001b[39mfrac_prior\u001b[38;5;132;01m:\u001b[39;00m\u001b[38;5;124m.1f\u001b[39m\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124mpercent.png\u001b[39m\u001b[38;5;124m\"\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m verbose \u001b[38;5;28;01melse\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m\n\u001b[0;32m--> 444\u001b[0m x_out \u001b[38;5;241m=\u001b[39m \u001b[43mconjugate_gradient_descent\u001b[49m\u001b[43m(\u001b[49m\u001b[43mB\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mu\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mx0\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mx0_wiener\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mrmin\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[38;5;241;43m0.0\u001b[39;49m\u001b[43m,\u001b[49m\n\u001b[1;32m    445\u001b[0m \u001b[43m        \u001b[49m\u001b[43mmax_iter\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mmax_iter\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mk\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mk\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mlblock\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mlblock\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mverbose\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mverbose\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mx_true\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mx\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\n\u001b[1;32m    446\u001b[0m \u001b[43m        \u001b[49m\u001b[43mtitle\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[38;5;124;43m\"\u001b[39;49m\u001b[38;5;124;43mRMSE smoothed gradient steps 10\u001b[39;49m\u001b[38;5;132;43;01m% d\u001b[39;49;00m\u001b[38;5;124;43mata salvaged\u001b[39;49m\u001b[38;5;124;43m\"\u001b[39;49m\u001b[43m,\u001b[49m\n\u001b[1;32m    447\u001b[0m \u001b[43m        \u001b[49m\u001b[43msaveas\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43msaveas\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    448\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m x0_wiener, x_out\n",
      "File \u001b[0;32m~/code/pfb-mod/conjugate_gradient.py:300\u001b[0m, in \u001b[0;36mconjugate_gradient_descent\u001b[0;34m(B, u, x0, rmin, max_iter, k, lblock, verbose, x_true, title, saveas)\u001b[0m\n\u001b[1;32m    297\u001b[0m     \u001b[38;5;28;01mbreak\u001b[39;00m\n\u001b[1;32m    299\u001b[0m \u001b[38;5;66;03m# Compute the action of B on p\u001b[39;00m\n\u001b[0;32m--> 300\u001b[0m Bp \u001b[38;5;241m=\u001b[39m \u001b[43mB\u001b[49m\u001b[43m(\u001b[49m\u001b[43mp\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    301\u001b[0m alpha \u001b[38;5;241m=\u001b[39m np\u001b[38;5;241m.\u001b[39mdot(r,r) \u001b[38;5;241m/\u001b[39m np\u001b[38;5;241m.\u001b[39mdot(p,Bp)\n\u001b[1;32m    302\u001b[0m x \u001b[38;5;241m=\u001b[39m x \u001b[38;5;241m+\u001b[39m alpha \u001b[38;5;241m*\u001b[39m p\n",
      "File \u001b[0;32m~/code/pfb-mod/conjugate_gradient.py:393\u001b[0m, in \u001b[0;36mconj_grad_with_prior.<locals>.<lambda>\u001b[0;34m(ts)\u001b[0m\n\u001b[1;32m    390\u001b[0m Q_inv \u001b[38;5;241m=\u001b[39m np\u001b[38;5;241m.\u001b[39mones(\u001b[38;5;28mlen\u001b[39m(x)) \u001b[38;5;66;03m# this is a prior, change to zeros if you want zero for infinite uncertainty\u001b[39;00m\n\u001b[1;32m    391\u001b[0m Q_inv[saved_idxs] \u001b[38;5;241m=\u001b[39m np\u001b[38;5;241m.\u001b[39mones(\u001b[38;5;28mlen\u001b[39m(saved_idxs)) \u001b[38;5;241m*\u001b[39m (\u001b[38;5;241m12\u001b[39m \u001b[38;5;241m/\u001b[39m delta\u001b[38;5;241m*\u001b[39m\u001b[38;5;241m*\u001b[39m\u001b[38;5;241m2\u001b[39m) \u001b[38;5;66;03m# 8 bits per real number (finer std because no complex) \u001b[39;00m\n\u001b[0;32m--> 393\u001b[0m B \u001b[38;5;241m=\u001b[39m \u001b[38;5;28;01mlambda\u001b[39;00m ts: \u001b[43mAT\u001b[49m\u001b[43m(\u001b[49m\u001b[43mN_inv\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43m \u001b[49m\u001b[43mA\u001b[49m\u001b[43m(\u001b[49m\u001b[43mts\u001b[49m\u001b[43m)\u001b[49m\u001b[43m)\u001b[49m \u001b[38;5;241m+\u001b[39m Q_inv \u001b[38;5;241m*\u001b[39m ts \u001b[38;5;66;03m# think ts===x\u001b[39;00m\n\u001b[1;32m    394\u001b[0m u \u001b[38;5;241m=\u001b[39m AT(N_inv \u001b[38;5;241m*\u001b[39m d) \u001b[38;5;241m+\u001b[39m Q_inv \u001b[38;5;241m*\u001b[39m prior \u001b[38;5;66;03m# this is same as mult prior by var=12/delta^2\u001b[39;00m\n\u001b[1;32m    396\u001b[0m \u001b[38;5;250m\u001b[39m\u001b[38;5;124;03m\"\"\"Optimize CHI squared using conjugate gradient method.\"\"\"\u001b[39;00m\n",
      "File \u001b[0;32m~/code/pfb-mod/matrix_operators.py:82\u001b[0m, in \u001b[0;36mAT\u001b[0;34m(x)\u001b[0m\n\u001b[1;32m     81\u001b[0m \u001b[38;5;28;01mdef\u001b[39;00m \u001b[38;5;21mAT\u001b[39m(x): \u001b[38;5;66;03m# the transpose of A\u001b[39;00m\n\u001b[0;32m---> 82\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m R(\u001b[43mA\u001b[49m\u001b[43m(\u001b[49m\u001b[43mR\u001b[49m\u001b[43m(\u001b[49m\u001b[43mx\u001b[49m\u001b[43m)\u001b[49m\u001b[43m)\u001b[49m)\n",
      "File \u001b[0;32m~/code/pfb-mod/matrix_operators.py:24\u001b[0m, in \u001b[0;36mA\u001b[0;34m(x)\u001b[0m\n\u001b[1;32m     22\u001b[0m \u001b[38;5;250m\u001b[39m\u001b[38;5;124;03m\"\"\"Applies PFB, irfft's that, flatten.\"\"\"\u001b[39;00m\n\u001b[1;32m     23\u001b[0m \u001b[38;5;66;03m# Forward PFB the Signal\u001b[39;00m\n\u001b[0;32m---> 24\u001b[0m b \u001b[38;5;241m=\u001b[39m \u001b[43mpfb\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mforward_pfb\u001b[49m\u001b[43m(\u001b[49m\u001b[43mx\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m     25\u001b[0m \u001b[38;5;66;03m# Inverse Fourier Transform along axis=1\u001b[39;00m\n\u001b[1;32m     26\u001b[0m b \u001b[38;5;241m=\u001b[39m irfft(b)\n",
      "File \u001b[0;32m~/code/pfb-mod/pfb.py:56\u001b[0m, in \u001b[0;36mforward_pfb\u001b[0;34m(timestream, nchan, ntap, window)\u001b[0m\n\u001b[1;32m     53\u001b[0m     ts_sec \u001b[38;5;241m=\u001b[39m timestream[bi\u001b[38;5;241m*\u001b[39mlblock:(bi\u001b[38;5;241m+\u001b[39mntap)\u001b[38;5;241m*\u001b[39mlblock]\u001b[38;5;241m.\u001b[39mcopy()\n\u001b[1;32m     55\u001b[0m     \u001b[38;5;66;03m# perform a real DFT (with applied, chunked window)\u001b[39;00m\n\u001b[0;32m---> 56\u001b[0m     spec[bi] \u001b[38;5;241m=\u001b[39m rfft(s(ts_sec \u001b[38;5;241m*\u001b[39m w)) \n\u001b[1;32m     58\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m spec\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "input(\"Are you sure? [Enter]\")\n",
    "\n",
    "time_total_start = time.time()\n",
    "\n",
    "kwargs_wien = {\"quantize\":True, \"usepfb\":True, \"isupchan\":True, \"upchan_factor\":UPCHAN_FACTOR, \"wiener_thresh\":0.1}\n",
    "kwargs_nofilt = {\"quantize\":True, \"usepfb\":True, \"isupchan\":True, \"upchan_factor\":UPCHAN_FACTOR, \"wiener_thresh\":0.0}\n",
    "kwargs_upchan_time_domain_sig = {\"usepfb\":True, \"upchan_factor\":UPCHAN_FACTOR, \"only_upchan\":True}\n",
    "\n",
    "corrmean_wien   = []\n",
    "corrmean_nofilt = []\n",
    "corrmean_fp     = []\n",
    "corrmean_1perc  = []\n",
    "corrmean_3perc  = []\n",
    "corrmean_5perc  = []\n",
    "corrmean_10perc = []\n",
    "prng = np.random.Generator(np.random.PCG64(seed=PRNG_SEED))\n",
    "for epoch in range(N_EPOCHS):\n",
    "    timeA = time.time()\n",
    "    # Make a pseudo-random signal\n",
    "    print(\"Generating signal...\", end=\" \")\n",
    "    signal = prng.normal(0,1,LEN_EPOCH)\n",
    "    sig1 = signal + prng.normal(0,TD_NOISE_SIGMA,LEN_EPOCH) \n",
    "    sig2 = signal + prng.normal(0,TD_NOISE_SIGMA,LEN_EPOCH) \n",
    "    timeB = time.time()\n",
    "    print(f\"took {timeB-timeA:.3f} seconds\")\n",
    "\n",
    "    \n",
    "    # Infinite precision\n",
    "    spec1=rechannelize(sig1,**kwargs_upchan_time_domain_sig) # Just channelize original signal \n",
    "    spec2=rechannelize(sig2,**kwargs_upchan_time_domain_sig) # directly to higher resolution\n",
    "    corrmean_fp.append(np.mean(spec1 * np.conj(spec2)))\n",
    "    timeC = time.time()\n",
    "    print(f\"{epoch+1}/{N_EPOCHS} mean power FP precision {np.real(corrmean_fp[-1]):.1f}\\t({timeC-timeB:.1f} s)\")\n",
    "    \n",
    "    # Wiener filtered\n",
    "    spec1=rechannelize(sig1,**kwargs_wien)\n",
    "    spec2=rechannelize(sig2,**kwargs_wien)\n",
    "    corrmean_wien.append(np.mean(spec1 * np.conj(spec2)))\n",
    "    timeD = time.time()\n",
    "    print(f\"{epoch+1}/{N_EPOCHS} mean power wiener filtered {np.real(corrmean_wien[-1]):.1f}\\t({timeD-timeC:.1f} s)\")\n",
    "    \n",
    "    # No filter\n",
    "    spec1=rechannelize(sig1,**kwargs_nofilt)\n",
    "    spec2=rechannelize(sig2,**kwargs_nofilt)\n",
    "    corrmean_nofilt.append(np.mean(spec1 * np.conj(spec2)))\n",
    "    timeE = time.time()\n",
    "    print(f\"{epoch+1}/{N_EPOCHS} mean power no filter {np.real(corrmean_nofilt[-1]):.1f}\\t({timeE-timeD:.1f} s)\")\n",
    "    \n",
    "    # Optimize with CG\n",
    "    conj_kwargs = {\n",
    "        \"frac_prior\":0.01, \n",
    "        \"delta\":DELTA_4BIT, \n",
    "        \"k\":LEN_EPOCH//NFRAME, \n",
    "        \"lblock\":NFRAME, \n",
    "        \"verbose\":False, \n",
    "        \"wiener_thresh\":WIENER_THRESH,\n",
    "        \"npersave\":7\n",
    "    }\n",
    "    # _, _, sig1_10, sig1_x5, sig1_x3, sig1_x1 = cg.conj_grad_one_three_five_perc(\n",
    "    #     x=sig1,\n",
    "    #     delta=0.343,          # quantization delta in stds\n",
    "    #     k=LEN_EPOCH//NFRAME,     \n",
    "    #     lblock=NFRAME,\n",
    "    #     verbose=False,         # True for plots\n",
    "    #     wiener_thresh=0.1, npersave1=7, npersave3=5, npersave5=4, npersave10=3\n",
    "    # )\n",
    "    # _, _, sig2_10, sig2_x5, sig2_x3, sig2_x1 = cg.conj_grad_one_three_five_perc(\n",
    "    #     x=sig2,\n",
    "    #     delta=0.343,          # quantization delta in stds\n",
    "    #     k=LEN_EPOCH//NFRAME,     \n",
    "    #     lblock=NFRAME,\n",
    "    #     verbose=False,         # True for plots\n",
    "    #     wiener_thresh=0.1, npersave1=7, npersave3=5, npersave5=4, npersave10=3\n",
    "    # )\n",
    "    for corrmean_list,frac_prior,npersave in zip((corrmean_1perc, corrmean_3perc, corrmean_5perc, corrmean_10perc),(0.01, 0.03, 0.05, 0.1), (7, 5, 4, 3)):\n",
    "        timeF = time.time()\n",
    "        conj_kwargs[\"frac_prior\"] = frac_prior\n",
    "        conj_kwargs[\"npersave\"]   = npersave\n",
    "        _, sig1_10 = cg.conj_grad_with_prior(x=sig1,**conj_kwargs)\n",
    "        _, sig2_10 = cg.conj_grad_with_prior(x=sig2,**conj_kwargs)\n",
    "        spec1=rechannelize(sig1_10,**kwargs_upchan_time_domain_sig)\n",
    "        spec2=rechannelize(sig2_10,**kwargs_upchan_time_domain_sig)\n",
    "        corrmean_list.append(np.mean(spec1 * np.conj(spec2)))\n",
    "        timeG = time.time()\n",
    "        print(f\"{epoch+1}/{N_EPOCHS} mean power CG {int(0.5+100*frac_prior)}% {np.real(corrmean_list[-1]):.1f}\\t({timeG-timeF:.1f} s)\")\n",
    "\n",
    "\n",
    "time_total_end = time.time()\n",
    "time_total_elapsed = time_total_end - time_total_start\n",
    "print(f\"\\nDone!\\nTotal time elapsed: {int(time_total_elapsed/3600)} hours {int((time_total_elapsed%3600)/60)} minutes {time_total_elapsed%60:.3f} seconds\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "5c17cc42-f5f6-452e-af6e-45129a71e5c9",
   "metadata": {},
   "outputs": [],
   "source": [
    "dumpdict = {}\n",
    "for method,corrmean,kwargs in zip(\n",
    "    (\"wien\",\"nofilt\",\"fp\",\"cg_1perc\",\"cg_3perc\",\"cg_5perc\",\"cg_10perc\"), \n",
    "    (corrmean_wien, corrmean_nofilt, corrmean_fp, corrmean_1perc, corrmean_3perc, corrmean_5perc, corrmean_10perc), \n",
    "    (kwargs_wien, kwargs_nofilt, kwargs_upchan_time_domain_sig, kwargs_upchan_time_domain_sig, kwargs_upchan_time_domain_sig, kwargs_upchan_time_domain_sig, kwargs_upchan_time_domain_sig)):\n",
    "    dumpdict[method] = {\"kwargs\": kwargs, \"CONSTS\": CONSTS, \"corrmean\": corrmean}\n",
    "now = dt.now()\n",
    "with open(f'{now}_snr_measurement.pkl','wb') as f:\n",
    "    pickle.dump(dumpdict, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "5f671931-1da5-41a4-b191-ab6823727e35",
   "metadata": {},
   "outputs": [],
   "source": [
    "with open(f'seed_22_nepoch_2_2024-11-21 14:24:37.627006_snr_measurement.pkl','rb') as f:\n",
    "    dumpdict1 = pickle.load(f)\n",
    "\n",
    "with open(f'seed_22_nepoch_2_2024-11-21 14:26:46.030753_snr_measurement.pkl','rb') as f:\n",
    "    dumpdict2 = pickle.load(f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "fb574cf7-2f3d-44e9-b5f2-3a51440368f6",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "wien\n",
      "----\n",
      "Sample mean             : 4.98e+04 ± 1.01e+02 \t(0.203%)\n",
      "Sample STD              : 8.07e+02 ± 7.13e+01 \t(8.84%)\n",
      "SNR (sqrt x^2/std^2)    : 6.17e+01 ± 5.46e+00 \t(8.84%)\n",
      "SNR ratios \t\t: 98.35% ± 12.30\n",
      "\n",
      "nofilt\n",
      "------\n",
      "Sample mean             : 5.08e+04 ± 9.15e+02 \t(1.801%)\n",
      "Sample STD              : 7.32e+03 ± 6.47e+02 \t(8.84%)\n",
      "SNR (sqrt x^2/std^2)    : 6.94e+00 ± 6.26e-01 \t(9.02%)\n",
      "SNR ratios \t\t: 11.06% ± 1.40\n",
      "\n",
      "fp\n",
      "--\n",
      "Sample mean             : 5.21e+04 ± 1.04e+02 \t(0.199%)\n",
      "Sample STD              : 8.31e+02 ± 7.34e+01 \t(8.84%)\n",
      "SNR (sqrt x^2/std^2)    : 6.27e+01 ± 5.55e+00 \t(8.84%)\n",
      "SNR ratios \t\t: 100.00% ± 12.50\n",
      "\n",
      "cg_1perc\n",
      "--------\n",
      "Sample mean             : 4.80e+04 ± 9.71e+01 \t(0.202%)\n",
      "Sample STD              : 7.77e+02 ± 6.86e+01 \t(8.84%)\n",
      "SNR (sqrt x^2/std^2)    : 6.18e+01 ± 5.47e+00 \t(8.84%)\n",
      "SNR ratios \t\t: 98.55% ± 12.32\n",
      "\n",
      "cg_3perc\n",
      "--------\n",
      "Sample mean             : 4.82e+04 ± 9.73e+01 \t(0.202%)\n",
      "Sample STD              : 7.78e+02 ± 6.88e+01 \t(8.84%)\n",
      "SNR (sqrt x^2/std^2)    : 6.19e+01 ± 5.47e+00 \t(8.84%)\n",
      "SNR ratios \t\t: 98.63% ± 12.33\n",
      "\n",
      "cg_5perc\n",
      "--------\n",
      "Sample mean             : 4.83e+04 ± 9.73e+01 \t(0.202%)\n",
      "Sample STD              : 7.78e+02 ± 6.88e+01 \t(8.84%)\n",
      "SNR (sqrt x^2/std^2)    : 6.20e+01 ± 5.48e+00 \t(8.84%)\n",
      "SNR ratios \t\t: 98.85% ± 12.36\n",
      "\n",
      "cg_10perc\n",
      "---------\n",
      "Sample mean             : 4.83e+04 ± 9.68e+01 \t(0.200%)\n",
      "Sample STD              : 7.74e+02 ± 6.84e+01 \t(8.84%)\n",
      "SNR (sqrt x^2/std^2)    : 6.24e+01 ± 5.52e+00 \t(8.84%)\n",
      "SNR ratios \t\t: 99.50% ± 12.44\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 800x300 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def get_and_print_stats(method,data,compare_snr=None,isplot=True):\n",
    "    sample_mean = data.mean()\n",
    "    sample_std = np.sqrt( np.sum((sample_mean - data)**2) / (len(data)-1) )\n",
    "    SE_mean = sample_std / np.sqrt(len(data))\n",
    "    SE_std = sample_std / np.sqrt(2 * len(data))\n",
    "    snr = np.sqrt(sample_mean**2)/np.sqrt(sample_std**2)\n",
    "    snr_err = snr * np.sqrt((SE_mean/sample_mean)**2 + (SE_std/sample_std)**2) # assumes delta mu / mu and delta sig / sig are small\n",
    "    if isplot is True:\n",
    "        plt.plot(data,'.',label=method)\n",
    "        print(f\"\\\n",
    "Sample mean             : {np.real(sample_mean):.2e} ± {SE_mean:.2e} \\t({100*SE_mean/sample_mean:.3f}%)\\n\\\n",
    "Sample STD              : {sample_std:.2e} ± {SE_std:.2e} \\t({100*SE_std/sample_std:.2f}%)\\n\\\n",
    "SNR (sqrt x^2/std^2)    : {snr:.2e} ± {snr_err:.2e} \\t({100*snr_err/snr:.2f}%)\")\n",
    "    if compare_snr is not None:\n",
    "        snr0,snr0err = compare_snr[0],compare_snr[1]\n",
    "        snr_ratios_err = np.sqrt((snr/snr0)**2*(snr0err/snr0)**2 + (snr_err/snr0)**2)\n",
    "        print(f\"SNR ratios \\t\\t: {snr/compare_snr[0]*100:.2f}% ± {snr_ratios_err*100:.2f}\")\n",
    "    return snr, snr_err\n",
    "\n",
    "def print_data(dumpdict):\n",
    "    fp_snr,fp_snr_err = get_and_print_stats(\"fp\", np.real(dumpdict['fp']['corrmean']), isplot=False)\n",
    "    plt.figure(figsize=(8,3))\n",
    "    for method,datadict in dumpdict.items():\n",
    "        print(\"\\n\"+method)\n",
    "        print(\"-\"*len(method))\n",
    "        get_and_print_stats(method,np.real(datadict['corrmean']),compare_snr=(fp_snr,fp_snr_err))\n",
    "    plt.legend()\n",
    "    plt.tight_layout()\n",
    "    plt.show()\n",
    "\n",
    "dumpdict=dumpdict0\n",
    "print_data(dumpdict)\n",
    "# corrmean_wien = dumpdict['wien']['corrmean']\n",
    "# corrmean_wien = np.real(corrmean_wien)\n",
    "# plt.plot(corrmean_wien)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "5db810b1-6670-4adf-8744-89af0c1938a7",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "dict_keys(['kwargs', 'CONSTS', 'corrmean'])"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dumpdict0['wien'].keys()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "id": "bbf763ac-412b-4ffc-b205-98e298a240dd",
   "metadata": {},
   "outputs": [
    {
     "ename": "AssertionError",
     "evalue": "Error, kwargs different, cannot merge",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mAssertionError\u001b[0m                            Traceback (most recent call last)",
      "Cell \u001b[0;32mIn[61], line 27\u001b[0m\n\u001b[1;32m     25\u001b[0m dd \u001b[38;5;241m=\u001b[39m fuse_dump_dicts(dd, dumpdict3)\n\u001b[1;32m     26\u001b[0m dd \u001b[38;5;241m=\u001b[39m fuse_dump_dicts(dd, dumpdict4)\n\u001b[0;32m---> 27\u001b[0m dd \u001b[38;5;241m=\u001b[39m \u001b[43mfuse_dump_dicts\u001b[49m\u001b[43m(\u001b[49m\u001b[43mdd\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mdumpdict5\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m     29\u001b[0m print_data(dd)\n",
      "Cell \u001b[0;32mIn[61], line 9\u001b[0m, in \u001b[0;36mfuse_dump_dicts\u001b[0;34m(dumpdict0, dumpdict1)\u001b[0m\n\u001b[1;32m      7\u001b[0m \u001b[38;5;28;01massert\u001b[39;00m i[\u001b[38;5;241m0\u001b[39m]\u001b[38;5;241m==\u001b[39mj[\u001b[38;5;241m0\u001b[39m]\n\u001b[1;32m      8\u001b[0m key \u001b[38;5;241m=\u001b[39m i[\u001b[38;5;241m0\u001b[39m]\n\u001b[0;32m----> 9\u001b[0m \u001b[38;5;28;01massert\u001b[39;00m i[\u001b[38;5;241m1\u001b[39m][\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mkwargs\u001b[39m\u001b[38;5;124m'\u001b[39m]\u001b[38;5;241m==\u001b[39mj[\u001b[38;5;241m1\u001b[39m][\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mkwargs\u001b[39m\u001b[38;5;124m'\u001b[39m], \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mError, kwargs different, cannot merge\u001b[39m\u001b[38;5;124m\"\u001b[39m\n\u001b[1;32m     10\u001b[0m kwargs \u001b[38;5;241m=\u001b[39m i[\u001b[38;5;241m1\u001b[39m][\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mkwargs\u001b[39m\u001b[38;5;124m'\u001b[39m]\n\u001b[1;32m     11\u001b[0m \u001b[38;5;28;01massert\u001b[39;00m i[\u001b[38;5;241m1\u001b[39m][\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mCONSTS\u001b[39m\u001b[38;5;124m'\u001b[39m][\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mLEN_EPOCH\u001b[39m\u001b[38;5;124m'\u001b[39m]\u001b[38;5;241m==\u001b[39mj[\u001b[38;5;241m1\u001b[39m][\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mCONSTS\u001b[39m\u001b[38;5;124m'\u001b[39m][\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mLEN_EPOCH\u001b[39m\u001b[38;5;124m'\u001b[39m], \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mError, key CONSTS differ, cannot merge\u001b[39m\u001b[38;5;124m\"\u001b[39m\n",
      "\u001b[0;31mAssertionError\u001b[0m: Error, kwargs different, cannot merge"
     ]
    }
   ],
   "source": [
    "# fuse dump dicts\n",
    "def fuse_dump_dicts(dumpdict0, dumpdict1):\n",
    "    dd={}\n",
    "    for i,j in zip(dumpdict0.items(), dumpdict1.items()):\n",
    "        #print(i[0], i[1].keys())\n",
    "        # check that kwargs and consts are same\n",
    "        assert i[0]==j[0]\n",
    "        key = i[0]\n",
    "        assert i[1]['kwargs']==j[1]['kwargs'], \"Error, kwargs different, cannot merge\"\n",
    "        kwargs = i[1]['kwargs']\n",
    "        assert i[1]['CONSTS']['LEN_EPOCH']==j[1]['CONSTS']['LEN_EPOCH'], \"Error, key CONSTS differ, cannot merge\"\n",
    "        assert i[1]['CONSTS']['DELTA_4BIT']==j[1]['CONSTS']['DELTA_4BIT'], \"Error, key CONSTS differ, cannot merge\"\n",
    "        assert i[1]['CONSTS']['NFRAME']==j[1]['CONSTS']['NFRAME'], \"Error, key CONSTS differ, cannot merge\"\n",
    "        assert i[1]['CONSTS']['NTAP']==j[1]['CONSTS']['NTAP'], \"Error, key CONSTS differ, cannot merge\"\n",
    "        assert i[1]['CONSTS']['UPCHAN_FACTOR']==j[1]['CONSTS']['UPCHAN_FACTOR'], \"Error, key CONSTS differ, cannot merge\"\n",
    "        assert i[1]['CONSTS']['DROP_NROWS']==j[1]['CONSTS']['DROP_NROWS'], \"Error, key CONSTS differ, cannot merge\"\n",
    "        assert i[1]['CONSTS']['WIENER_THRESH']==j[1]['CONSTS']['WIENER_THRESH'], \"Error, key CONSTS differ, cannot merge\"\n",
    "        assert i[1]['CONSTS']['TD_NOISE_SIGMA']==j[1]['CONSTS']['TD_NOISE_SIGMA'], \"Error, key CONSTS differ, cannot merge\"\n",
    "        consts = i[1]['CONSTS']\n",
    "        dd[key] = {'kwargs':kwargs, 'CONSTS':consts, 'corrmean':i[1]['corrmean'] + j[1]['corrmean']}\n",
    "    return dd\n",
    "\n",
    "dd = fuse_dump_dicts(dumpdict0, dumpdict1)\n",
    "dd = fuse_dump_dicts(dd, dumpdict2)\n",
    "dd = fuse_dump_dicts(dd, dumpdict3)\n",
    "dd = fuse_dump_dicts(dd, dumpdict4)\n",
    "\n",
    "print_data(dd)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "bd5fe42e-f807-4073-8d6e-e6b4acf65cd4",
   "metadata": {},
   "outputs": [],
   "source": [
    "x=50982.51\n",
    "dx=1495.314\n",
    "std=51626.24\n",
    "stderr=1057.347\n",
    "f\"{x**2:.2e} , {2*x*dx:.2e} , {dx**2:.2e}, {x**2/std**2:.2e}\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "60318753-6b48-4558-bb8c-3c1e8078d242",
   "metadata": {},
   "outputs": [],
   "source": [
    "with open(f'2024-09-16 08:51:29.277051 snr_measurement.pkl','rb') as f:\n",
    "    kwargs = pickle.load(f)\n",
    "    consts = pickle.load(f)\n",
    "    data   = pickle.load(f)\n",
    "\n",
    "data = np.asarray(data)\n",
    "# plt.plot(np.real(arr))\n",
    "# plt.plot(np.imag(arr))\n",
    "sample_mean = data.mean()\n",
    "sample_std = np.sqrt(sum(abs(data - sample_mean)**2) / (consts['N_EPOCHS'] - 1))\n",
    "SE_mean = sample_std / np.sqrt(consts['N_EPOCHS']) # standard error in the mean\n",
    "SE_std = sample_std / np.sqrt(2 * (consts['N_EPOCHS'] - 1))\n",
    "\n",
    "sample_mean, sample_std, SE_mean, SE_std\n",
    "\n",
    "print(f\"Sample mean            : {np.real(sample_mean):.2f}\\n\\\n",
    "Sample STD             : {sample_std:.2f}\\n\\\n",
    "Standard Error in mean : {SE_mean:.3f}\\n\\\n",
    "Standard Error in STD  : {SE_std:.3f}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "88348e68-e2c5-45ca-aa21-bb9167b38024",
   "metadata": {},
   "outputs": [],
   "source": [
    "consts"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "7124a5ad-5e1f-458d-9908-b7cc8a1ee46a",
   "metadata": {},
   "outputs": [],
   "source": [
    "powers = np.array(powers)\n",
    "n = len(powers)\n",
    "sample_mean = np.mean(powers)\n",
    "sample_stdev = np.sqrt(np.sum(abs(powers-sample_mean)**2) / (n-1))\n",
    "err_in_mean = sample_stdev / np.sqrt(n)\n",
    "\n",
    "print(f\"stdev = {sample_stdev:.1f}\\nmean = {np.real(sample_mean):.1f}\\nmean err = {err_in_mean:.3f}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "f42be19f-764a-47d8-b000-616d9d2b7564",
   "metadata": {},
   "outputs": [],
   "source": [
    "powers = np.array(powers)\n",
    "n = len(powers)\n",
    "sample_mean = np.mean(powers)\n",
    "sample_stdev = np.sqrt(np.sum(abs(powers-sample_mean)**2) / (n-1))\n",
    "err_in_mean = sample_stdev / np.sqrt(n)\n",
    "\n",
    "print(f\"stdev = {sample_stdev:.1f}\\nmean = {np.real(sample_mean):.1f}\\nmean err = {err_in_mean:.3f}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 118,
   "id": "f2812d68-5f17-4267-abc6-3e64b0898c17",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'2024-09-13 12:03:34.542569'"
      ]
     },
     "execution_count": 118,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from datetime import datetime as dt\n",
    "f\"{dt.now()}\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "id": "5beb91a2-b5ff-4658-8853-aea9bb4e6628",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pickle\n",
    "# serialize\n",
    "with open(\"data.pkl\",'wb') as f:\n",
    "    pickle.dump(kwargs, f) \n",
    "    pickle.dump(powers, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 116,
   "id": "e0ba9ad1-7cf1-41dd-8612-8ab7108b1983",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'quantize': True, 'usepfb': True, 'isupchan': True, 'upchan_factor': 32, 'wiener_thresh': 0.0, 'drop_nrows': 10}\n"
     ]
    },
    {
     "name": "stdin",
     "output_type": "stream",
     "text": [
      "[Enter] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[51468.60235381 +24.21375583j 51361.20546312 -39.13596832j\n",
      " 51416.72414044 -45.03663199j 51352.89460731 +20.03624506j\n",
      " 51242.15653027 +32.36121811j 51389.86389834  -0.14710532j\n",
      " 51333.56650997 -31.21759395j 51354.89791771-118.7208946j\n",
      " 51424.57108576 -29.71995028j 51309.50614489 -91.3380315j\n",
      " 51319.40044988 +13.08615023j 51421.94023697 +22.49406926j\n",
      " 51485.58761093  +2.35984471j 51447.7017655  -76.117301j\n",
      " 51454.11078157 +48.85819259j 51402.25099721 -68.30456146j\n",
      " 51371.74635442 +79.57525329j 51441.88104594 +56.74595166j\n",
      " 51544.24140276 -53.24474712j 51461.87805746+115.91339168j\n",
      " 51399.38980198 -48.7264323j  51363.11875782 +71.48814367j\n",
      " 51350.44692356 +15.6335975j  51453.49667809  -3.89558099j\n",
      " 51388.72557177 +24.10576404j 51507.71040142+103.27625128j\n",
      " 51389.95539954  +2.61220181j 51401.28237326 +13.9535463j\n",
      " 51354.85958408 +33.97144595j 51459.98978044 +94.73191225j\n",
      " 51450.44949158 +89.94387067j 51474.20103056 -24.99666245j]\n"
     ]
    }
   ],
   "source": [
    "with open('data.pkl','rb') as f:\n",
    "    print(pickle.load(f))\n",
    "    input(\"[Enter]\")\n",
    "    print(pickle.load(f))    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "id": "94aeef27-45cd-42ad-a147-b3f92ca2a1f3",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x78c7cc13f130>]"
      ]
     },
     "execution_count": 106,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "power = np.mean(corr,axis=0)\n",
    "plt.plot(np.real(power))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "id": "9818e38a-4d78-43ff-b6b5-d27c80e1183b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1021, 32769)"
      ]
     },
     "execution_count": 96,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "spec1.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "id": "542f7135-9595-4ce6-8afe-da771db1e82a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "15.0"
      ]
     },
     "execution_count": 98,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.log2(32768)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "id": "59806f81-943b-4575-a89c-7224d49e775f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2048"
      ]
     },
     "execution_count": 100,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "1<<(16-5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b75e1ed2-7e56-4e3a-9923-aae7fe6eb1f6",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "0a5c10b8-2ea0-4ba7-8703-28caba83be7a",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "9c136abe-cd1e-40f5-8909-ed959eb67fa5",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "id": "350bc614-10e9-47db-83fe-e9614f0a8043",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "SNR between 1s correlations without quantization noise is 30.77 dB\n"
     ]
    }
   ],
   "source": [
    "spec1 = get_rfft_spec(sig1, NFRAME)[:,1:] # don't care about DC\n",
    "spec2 = get_rfft_spec(sig2, NFRAME)[:,1:]\n",
    "corr = spec1 * np.conj(spec2)\n",
    "# Average the corrlation over one-second (LEN_EPOCH in num samples) intervals\n",
    "corr1s = corr.reshape((N_EPOCHS,LEN_EPOCH//2)).mean(axis=1)\n",
    "assert corr1s.shape == (N_EPOCHS,) # sanity check\n",
    "sig_1s_power_fft = np.real(corr1s.mean())\n",
    "noise_1s_power_fft = corr1s.std()\n",
    "snr_1s_fft = 10 * np.log10(sig_1s_power_fft / noise_1s_power_fft)\n",
    "print(f\"SNR between 1s correlations without quantization noise is {snr_1s_fft:.2f} dB\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1dfb4f93-7a31-41a0-9f5e-c6c153741455",
   "metadata": {},
   "source": [
    "Add quantization noise, see affect on SNR"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "ed153852-b99f-4bbb-b431-c1aff10e71df",
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'sig_1s_power_pfb' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "Cell \u001b[0;32mIn[20], line 10\u001b[0m\n\u001b[1;32m      8\u001b[0m sig_1s_power_quantized \u001b[38;5;241m=\u001b[39m np\u001b[38;5;241m.\u001b[39mreal(corr1s\u001b[38;5;241m.\u001b[39mmean())\n\u001b[1;32m      9\u001b[0m noise_1s_power_quantized \u001b[38;5;241m=\u001b[39m corr1s\u001b[38;5;241m.\u001b[39mstd()\n\u001b[0;32m---> 10\u001b[0m snr_1s_pfb \u001b[38;5;241m=\u001b[39m \u001b[38;5;241m10\u001b[39m \u001b[38;5;241m*\u001b[39m np\u001b[38;5;241m.\u001b[39mlog10(\u001b[43msig_1s_power_pfb\u001b[49m \u001b[38;5;241m/\u001b[39m noise_1s_power_pfb)\n\u001b[1;32m     11\u001b[0m \u001b[38;5;28mprint\u001b[39m(\u001b[38;5;124mf\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mSNR between 1s correlations with quantization noise is \u001b[39m\u001b[38;5;132;01m{\u001b[39;00msnr_1s_pfb\u001b[38;5;132;01m:\u001b[39;00m\u001b[38;5;124m.2f\u001b[39m\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m dB\u001b[39m\u001b[38;5;124m\"\u001b[39m)\n",
      "\u001b[0;31mNameError\u001b[0m: name 'sig_1s_power_pfb' is not defined"
     ]
    }
   ],
   "source": [
    "std_spec = np.std(np.concatenate([np.real(spec1),np.imag(spec1),np.real(spec2),np.imag(spec2)]))\n",
    "spec1_quant = pfb.quantize_8_bit(spec1, delta = std_spec * DELTA_4BIT)\n",
    "spec2_quant = pfb.quantize_8_bit(spec2, delta = std_spec * DELTA_4BIT)\n",
    "corr = spec1_quant * np.conj(spec2_quant)\n",
    "# Average the correlation over one-second intervals\n",
    "corr1s = corr.reshape((N_EPOCHS,LEN_EPOCH//2)).mean(axis=1)\n",
    "assert corr1s.shape == (N_EPOCHS,) # sanity check\n",
    "sig_1s_power_quantized = np.real(corr1s.mean())\n",
    "noise_1s_power_quantized = corr1s.std()\n",
    "snr_1s_pfb = 10 * np.log10(sig_1s_power_pfb / noise_1s_power_pfb)\n",
    "print(f\"SNR between 1s correlations with quantization noise is {snr_1s_pfb:.2f} dB\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b5036b07-7f4e-430e-9d3f-7cc02d25e551",
   "metadata": {},
   "source": [
    "### PFB SNR analysis"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a18c3e75-2943-4b0b-9b98-a07ced2f09f2",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "5a285de0-538f-420d-a226-821b5450773f",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}