Python Setup & Installation

1 What you’re installing, and why

Native Python from python.org, with VS Code as the editor. Three reasons, in order of how much they matter:

• Instrument drivers expect it. NI-DAQmx, NI-488.2, and Thorlabs Kinesis look for a system Python they can bind to. If you ever drive hardware from Python, this is the install that works.
• One Python, one package manager. Native Python plus pip is a single environment you can reason about. Adding a second package ecosystem alongside pip is the most common way a working setup quietly stops working.
• One editor for both halves of the job. VS Code runs notebooks and scripts, so moving from exploring data to automating a measurement doesn’t mean changing tools.

If you already have Anaconda, see If you already have Anaconda below rather than uninstalling it.

2 Install Python (Windows)

Use the Python Install Manager, the official tool from python.org for managing Python versions on Windows. The two downloads below are identical builds; python.org suggests the Store for most people.

Option A: install from the Microsoft Store

Search for “Python Install Manager” in the Microsoft Store and install it.

Option B: download from python.org

1. Go to python.org/downloads
2. Click “Download Python install manager”
3. Run the installer and follow the prompts

After installation, open PowerShell and install Python itself:

py install 3.13

Verify it:

py --version

To see the versions available:

py list --online

The py launcher is how you’ll invoke Python from here on — py script.py to run a script, py -m pip ... to install packages. It picks the right interpreter even with several versions installed.

3 Install Python (Mac)

1. Go to python.org/downloads
2. Download the macOS installer for the latest release (macOS 10.15 and later)
3. Run the installer and follow the prompts

Verify it by opening Terminal and running:

python3 --version

On Mac the command is python3 (and python3 -m pip) wherever this site shows py.

4 Install VS Code

Download it from code.visualstudio.com and run the installer.

Then open the Extensions panel (Ctrl+Shift+X) and install two extensions:

With those in place, VS Code opens both .py scripts and .ipynb notebooks. Workflows covers when to reach for each and how to work in the editor.

5 Install the analysis packages

These are the packages every kind of lab analysis uses.

Windows:

py -m pip install numpy scipy matplotlib jupyterlab uncertainties

Mac:

python3 -m pip install numpy scipy matplotlib jupyterlab uncertainties

That covers everything under Analyzing Data — curve fitting, error propagation, Fourier analysis. Add the hardware packages below only if your work needs them.

py -m venv .venv
.venv\Scripts\Activate.ps1
py -m pip install numpy matplotlib

Set-ExecutionPolicy -ExecutionPolicy RemoteSigned -Scope CurrentUser

6 Install hardware packages and drivers (Windows)

Talking to an instrument takes two pieces: a Python package (installed with pip) and a vendor driver (usually installed separately). They’re largely independent — install only the pairs for hardware you actually use.

py -m pip install pyvisa

import pyvisa
rm = pyvisa.ResourceManager()
print(rm.list_resources())   # lists all connected instruments

py -m pip install pyvisa-py pyserial

import pyvisa
rm = pyvisa.ResourceManager('@py')   # '@py' selects the pure-Python backend

py -m pip install nidaqmx

py -m nidaqmx installdriver

import nidaqmx
print(nidaqmx.system.System.local().driver_version)

py -m pip install pythonnet

C:\Program Files\Thorlabs\Kinesis\

7 Verify your setup

Run this to see what’s working. Each hardware block is wrapped in try/except, so it reports cleanly whether or not that piece is installed — you should expect errors for hardware you didn’t set up.

# Core analysis packages
import numpy as np
import matplotlib.pyplot as plt
from scipy.optimize import curve_fit
from uncertainties import ufloat
print("Core packages: OK")

# VISA (bench/GPIB instruments)
try:
    import pyvisa
    rm = pyvisa.ResourceManager()
    print(f"VISA resources found: {len(rm.list_resources())}")
except Exception as e:
    print(f"PyVISA: {e}")

# NI-DAQmx (data acquisition)
try:
    import nidaqmx
    print(f"NI-DAQmx version: {nidaqmx.system.System.local().driver_version}")
except Exception as e:
    print(f"NI-DAQmx: {e}")

# pythonnet (Thorlabs Kinesis)
try:
    import clr
    print("pythonnet: OK")
except Exception as e:
    print(f"pythonnet: {e}")

print("\nSetup complete.")

If the core packages import, you’re ready for everything in Analyzing Data.

8 If you already have Anaconda

Anaconda works, and we don’t recommend it. It bundles a second package manager (conda) alongside pip, and mixing the two is a reliable way to end up with an environment that breaks in confusing ways. If you’re starting fresh, install native Python as above.

If Anaconda is already on your machine and you’d rather not change, here’s what differs from the instructions on this site:

• py points at the wrong Python. Anaconda doesn’t install the py launcher. If you have it from an earlier install, it resolves to python.org interpreters rather than your conda environment — which is the more confusing failure. Use python and python -m pip from an Anaconda Prompt or an activated environment.
• Install into a named environment, not base. Create one with conda create -n lab python=3.13, then conda activate lab.
• conda install is fine for the analysis packages — NumPy, SciPy and Matplotlib are all in the default channels and conda resolves them together.
• The hardware packages aren’t in the default channels, but pyvisa, pyserial, nidaqmx-python and pythonnet are all on conda-forge (conda install -c conda-forge pyvisa nidaqmx-python). The conda builds tend to lag the PyPI releases, and you install the vendor driver separately either way — so this works, it just runs behind.
• Tell VS Code which interpreter to use — click the Python version in the status bar and pick the conda environment, not base.

9 Working without installing anything

Google Colab runs notebooks in your browser with nothing to install, and is a reasonable place to do analysis — particularly on a machine where you can’t install software.

Its default runtime is on Google’s servers, so it cannot reach instruments: no DAQ, no VISA, no motor control. Collect data on a lab computer, then upload the file to Colab to analyze it. Some courses use Colab as their main analysis environment; if yours does, follow that course’s instructions.

(Colab can be pointed at a local runtime on your own machine, which does reach local hardware — but that needs a working local Python install anyway, and if you have that, you may as well use VS Code.)

Back to Python for Lab Work