Python中使用asammdf从Pandas DataFrame生成MDF文件时时间索引异常问题求助

I've run into this exact issue before—here's what's happening and how to fix it quickly:

The Root Cause #

When you call mdf_obj.append(df) directly, asammdf defaults to treating your DataFrame's integer index as timestamp values, with each index increment representing 1 second. Since your actual sample interval is 0.015 seconds, this means every data point gets scaled to 1 second in the MDF file—hence the 66.6x duration inflation you're seeing.

Solution 1: Use the time_column Parameter (Recommended) #

The MDF.append() method has a time_column argument that lets you specify which DataFrame column to use as the time axis (in seconds). You just need to convert your datetime TIME column to a Unix timestamp (float, in seconds) first:

import pandas as pd
import numpy as np
from asammdf import MDF, Signal

data = {
    "TIME": pd.date_range(start='2024-01-01', periods=200, freq='15ms'),
    "FLOAT_SIGNAL": np.random.rand(200),
}
df = pd.DataFrame(data)

# Convert datetime column to Unix timestamps (in seconds, with millisecond precision)
df["TIME_SEC"] = df["TIME"].apply(lambda dt: dt.timestamp())

mdf_obj = MDF()
# Tell asammdf to use our TIME_SEC column for timestamps
mdf_obj.append(df, time_column="TIME_SEC")
# Use raw string for Windows path to avoid escape issues
mdf_obj.save(r"C:\Users\user\Desktop\output.mdf", compression=2)

Solution 2: Manually Create Signal Objects #

If you need more control (or if your asammdf version doesn't support time_column), you can build Signal instances explicitly and attach the correct timestamp array:

import pandas as pd
import numpy as np
from asammdf import MDF, Signal

data = {
    "TIME": pd.date_range(start='2024-01-01', periods=200, freq='15ms'),
    "FLOAT_SIGNAL": np.random.rand(200),
}
df = pd.DataFrame(data)

# Convert datetime column to a numpy array of timestamps (seconds)
timestamps = df["TIME"].apply(lambda dt: dt.timestamp()).to_numpy()

# Create the signal with the correct time axis
float_signal = Signal(
    samples=df["FLOAT_SIGNAL"].to_numpy(),
    timestamps=timestamps,
    name="FLOAT_SIGNAL",
    unit="",  # Add your actual unit if applicable
    comment="Random floating-point signal"
)

mdf_obj = MDF()
mdf_obj.append(float_signal)
mdf_obj.save(r"C:\Users\user\Desktop\output.mdf", compression=2)

Key Notes #

• asammdf expects timestamps in seconds (floats are supported for sub-second precision, which is perfect for your 15ms intervals).
• The datetime-to-timestamp conversion preserves millisecond data (e.g., 2024-01-01 00:00:00.015 becomes 1704067200.015), so the MDF file will calculate the correct total duration based on the actual time between samples.
• Always use raw strings (r"path") or escaped backslashes ("C:\\Users\\...") for Windows file paths to avoid parsing errors.

内容的提问来源于stack exchange，提问作者jB777