在gnuplot绘图时，能否直接从原始数据中减去平滑后的数据？

Absolutely—you don’t need external tools to do this! Gnuplot has built-in functionality to handle this operation directly. Here’s how to modify your existing command to plot the difference between your original sensor data and its acsplines-smoothed version:

Approach 1: Use a Temporary File for Smoothed Data #

This is the most reliable method, compatible with most gnuplot versions. First, we’ll generate the smoothed data and save it to a temporary file, then calculate the difference by matching x-values between the original and smoothed datasets:

# First, capture the smoothed data to a temporary file
set table 'temp_smooth.dat'
plot ["17.12.2020 08:00:00":"18.12.2020 20:00:00"] 'data3-17-28.csv1' using 4:5 smooth acsplines
unset table

# Configure time formatting for your x-axis (since your x is a timestamp)
set xdata time
set timefmt "%d.%m.%Y %H:%M:%S"
set format x "%d.%m %H:%M"  # Adjust this to your preferred display format

# Plot original, smoothed, and their difference
plot ["17.12.2020 08:00:00":"18.12.2020 20:00:00"] \
     'data3-17-28.csv1' using 4:5 title 'Sensor 3' with lines lc rgb '#0066ff', \
     'temp_smooth.dat' using 1:2 title 'Smoothed' with lines lc rgb '#ff3333', \
     '< join data3-17-28.csv1 temp_smooth.dat' using 1:($2 - $4) title 'Original - Smoothed' with lines lc rgb '#00cc66'

The join command merges the original and smoothed files by their x-values (assuming both are sorted chronologically, which your timestamp data should be). The $2 - $4 calculates the difference between the original y-value (column 2 of the merged output) and the smoothed y-value (column 4).

Approach 2: Use an In-Memory Data Block (Gnuplot 5.0+) #

If you’re using gnuplot 5.0 or later, you can avoid creating a physical temporary file by storing the smoothed data in an in-memory data block:

# Save smoothed data to an in-memory block named $SMOOTHED
set table $SMOOTHED
plot ["17.12.2020 08:00:00":"18.12.2020 20:00:00"] 'data3-17-28.csv1' using 4:5 smooth acsplines
unset table

# Configure time formatting
set xdata time
set timefmt "%d.%m.%Y %H:%M:%S"
set format x "%d.%m %H:%M"

# Plot the datasets and their difference
plot ["17.12.2020 08:00:00":"18.12.2020 20:00:00"] \
     'data3-17-28.csv1' using 4:5 title 'Sensor 3' with lines, \
     $SMOOTHED using 1:2 title 'Smoothed' with lines, \
     '< join data3-17-28.csv1 <(echo "$SMOOTHED")' using 1:($2 - $4) title 'Original - Smoothed' with lines

This works similarly to the first method, but uses a data block instead of a file, keeping your workspace clean.

Key Notes: #

• Ensure your original data is sorted by the x-axis (timestamp) for join to work correctly.
• If you don’t need to plot the original and smoothed datasets separately, you can omit those lines and only plot the difference.

内容的提问来源于stack exchange，提问作者Sergei