请求将IDL二进制数据读取器转换为Python代码

Here's the Python implementation matching your IDL code, including proper handling of signed 24-bit data and little-endian binary format:

import numpy as np
import os

def NC_vlf(file, t0m, t0s, dus):
    fs = 78125.0  # Sampling frequency in Hz (78.125 kHz)
    
    # Get file size and calculate number of 8-byte blocks
    file_size = os.path.getsize(file)
    np_blocks = file_size // 8  # Each block contains two 32-bit integers
    
    # Read binary data as little-endian 32-bit signed integers
    # Data is stored as WE, NS, WE, NS... so reshape to (2, np_blocks) where:
    # d[0] = all WE channel samples, d[1] = all NS channel samples
    data = np.fromfile(file, dtype='<i4')
    d = data.reshape((np_blocks, 2)).T
    
    # Define channel indices (adjust to [0,1] if NS comes first in your file)
    nswe = [1, 0]  # nswe[0] = NS channel index, nswe[1] = WE channel index
    
    # Calculate start and end indices for the selected interval
    start_time = t0m * 60 + t0s
    i0 = int(fs * start_time)
    i1 = i0 + int(np.ceil(fs * dus))
    
    # Ensure indices stay within array bounds
    i0 = max(0, i0)
    i1 = min(d.shape[1] - 1, i1)
    
    # Generate time axis (seconds since start of the hour)
    num_samples = i1 - i0 + 1
    t = np.arange(num_samples) / fs + t0s
    
    # Extract 24-bit data (discard lower 8 control bits) and convert to nanoTesla
    # Integer division by 256 shifts right 8 bits, preserving sign for signed 24-bit values
    B_NS_nT = (d[nswe[0], i0:i1+1] // 256) / 1e7
    B_WE_nT = (d[nswe[1], i0:i1+1] // 256) / 1e7
    
    return t, B_NS_nT, B_WE_nT

Key Details: #

• Binary Format: Uses <i4 dtype to read little-endian 32-bit signed integers, matching IDL's /Swap_If_Big_Endian behavior.
• 24-bit Data Handling: Integer division by 256 removes the 8-bit control word. Since we use signed integers, negative 24-bit values are correctly sign-extended.
• Channel Order: Adjust the nswe list if your file stores NS samples before WE samples.
• Large File Optimization: For files over 2GB, use np.memmap to avoid loading the entire file into memory:

  # Alternative for large files
  with np.memmap(file, dtype='<i4', mode='r') as data:
      d = data.reshape((np_blocks, 2)).T
      # Rest of the code remains identical
  

Usage Example: #

t, B_NS, B_WE = NC_vlf('CH1_2_20170103_060000', 42.0, 3.547, 0.008)

内容的提问来源于stack exchange，提问作者Richard Ferranti