Fiber Alignment with Zaber Devices

This repository contains code to perform 2-axis optical fiber alignment with Zaber stages and an optical power meter that is capable to outputing an analog signal as a power feedback. Algorithms described in the article are implemented in Python, along with classes allowing them to be easily used with Zaber stages via the Zaber Motion Library API.

Hardware Requirements

This code is designed to run on devices with Zaber controllers.

Notes:

• fiber_alignment_demo.py is written for two zaber axes controlled by a single Zaber controller with the power meter signal connected to an analog input on the same controller. The code can be adapted for setups with multiple controllers but some methods may not be compatible.
• The streamed_spiral_scan method in the FiberAlignment2D class is only compatible when the motion devices and power meter analog input are all connected to the same controller since it requires synchronized motion and inputs.

Dependencies / Software Requirements / Prerequisites

The script uses uv to manage virtual environment and dependencies:

pip install uv

The dependencies are listed in pyproject.toml.

Usage

fiber_alignment_2d.py is designed to be imported and reused in other programs rather than being run directly. fiber_alignment_demo.py is provided to demonstrate the usage of the classes in fiber_alignment_2d.py.

fiber_alignment_2d.py contains several methods for performing fiber alignment. The ideal choice will depend on the system and application it is being used for. Details on the purpose and benefits of each method are provided below in the FiberAlignment2D Class section.

Demo Script

Configuration / Parameters

Edit the script settings in fiber_alignment_demo.py to fit your setup before running the script.

SERIAL_PORT is the serial port that your Zaber device is connected to. For more information on how to identify the serial port, see Find the right serial port name.

Running the Script

To run the example code:

uv sync
uv run fiber_alignment_demo.py

FiberAlignment2D Class

Importing the class (Python):

from fiber_alignment_2d import DeviceAnalogInput, FiberAlignment2D, AlignmentResult, AlignmentSamples

Initialization (Python):

analog_input = DeviceAnalogInput(zaber_device, analog_input_number, gain, num_samples)
fiber_alignment = FiberAlignment2D(zaber_axis_1, zaber_axis_2, analog_input)

• The zaber_axis_1 and zaber_axis_2 parameters are the Axis class instances that the FiberAlignment2D class will move to maximize analog input signal during alignment.
• The analog_input parameter is an instance of the DeviceAnalogInput class which defines the analog input used for the power meter.
  • The zaber_device parameter is the Device class instance that contains the analog input.
  • The analog_input_number parameter specifies which analog input pin the power meter is connected to.
  • The gain parameter is a gain factor for converting the analog input voltage to power units. This parameter is optional and defaults to 1.
  • The num_samples parameter specifies how many samples to take and average when calling the get_signal method to reduce noise. This parameter is optional and defaults to 1.

First-Light Search Methods

These methods are used to scan for an initial signal strong enough to reliably to use a hill climb optimization routine.

raster_scan()

Performs a point-by-point raster scan of a square centered on the current position starting at one corner. The search_distance parameter defines the size of the square search area and the resolution parameter defines the step size between points.

This is useful for covering a large area if zaber_axis_2 is faster than zaber_axis_1 since the slow axis can remain stationary during each line scan.

This method stops when the signal exceeds first_light_threshold by default. If it is necessary to scan the entire area in order to map the signal in the space or find a global maximum, set the stop_at_threshold option to False which will allow the scan to complete and then move to the position with the highest signal.

spiral_scan()

This method is similar to raster_scan() but performs a square spiral starting at the current position and spirals outwards. This is typically more efficient than a raster scan if the initial position is already roughly aligned.

streamed_spiral_scan()

This methods performs a constant velocity circular spiral motion starting at the current position using streams and uses triggers to stop when a signal is detected. This method is faster than spiral_scan() since it is continously moving and monitoring the signal at high freqeuncy but can be less reliable if the power meter response is not fast enough or there is signal noise.

When trigger_threshold is exceeded, the current position is recorded and motion is stopped. After coming to a stop, the stages move back to the position recorded by the trigger, retakes the measurement, and compares it to first_light_threshold to ensure that the final signal is above the required threshold. The triggers can take up to a few milliseconds to record the position resulting in some error. trigger_threshold must be higher than first_light_threshold to account for this as well as other sources of error and noise.

The recorded position is read from the axis setting specified by the trigger_position_setting property in FiberAlignment2D. By default, encoder.pos will be used if the stages have built in encoders and pos will be used otherwise.

Reliability can be improved by reducing the speed and acceleration. To change the max speed or acceleration used during streamed motion use the set_stream_max_speed and set_stream_max_accel methods in the instance of FiberAlignment2D.

If the scan is reliably detecting first-light but the system is not returning to that position accurately enough to get a signal above first_light_threshold, an alternative to slowing down the streamed motion is to perform a spiral_scan() afterwards. This combination would make use of streamed_spiral_scan() to quickly get close to a position where first-light can be detected and then use the slower but more reliable spiral_scan() find the first-light signal again.

Hill Climb Optimization Methods

These methods optimize position to maximize the signal strength. Simple examples of a pattern search and gradient search are provided but more complex and efficient methods are also possible. These methods can also be extended to work with more axes simultaneously.

pattern_search()

The pattern search algorithm takes a positive and negative step with each axis and moves to the position with the highest signal. When no improvement is found, it halves the step size and repeats the process. The search iterates this process until the step size is min_step or max_iterations is reached.

gradient_search()

This search algorithm takes a step on each axis to calculate the gradient and takes a step in the direction of the gradient. When no improvement is found, it halves the step size and repeats the process. The search iterates this process until the step size is min_step or max_iterations is reached.

This method is more complex and sensitive to noise but can be faster than the pattern_search() because it requires fewer measurements per iteration and can move diagonally.

Troubleshooting Tips

• If you encounter an error when running one of the example scripts, check the Serial Port number (ex. COM5) using Zaber Launcher.