OpenStage: a low-cost motorized microscope stage with sub-micron positioning accuracy

Abstract

Recent progress in intracellular calcium sensors and other fluorophores has promoted the widespread adoption of functional optical imaging in the life sciences. Home-built multiphoton microscopes are easy to build, highly customizable, and cost effective. For many imaging applications a 3-axis motorized stage is critical, but commercially available motorization hardware (motorized translators, controller boxes, etc) are often very expensive. Furthermore, the firmware on commercial motor controllers cannot easily be altered and is not usually designed with a microscope stage in mind. Here we describe an open-source motorization solution that is simple to construct, yet far cheaper and more customizable than commercial offerings. The cost of the controller and motorization hardware are under $1000. Hardware costs are kept low by replacing linear actuators with high quality stepper motors. Electronics are assembled from commonly available hobby components, which are easy to work with. Here we describe assembly of the system and quantify the positioning accuracy of all three axes. We obtain positioning repeatability of the order of 1 μm in X/Y and 0.1 μm in Z. A hand-held control-pad allows the user to direct stage motion precisely over a wide range of speeds (10(-1) to 10(2) μm·s(-1)), rapidly store and return to different locations, and execute "jumps" of a fixed size. In addition, the system can be controlled from a PC serial port. Our "OpenStage" controller is sufficiently flexible that it could be used to drive other devices, such as micro-manipulators, with minimal modifications.

Figure 1. Complete stage setup.

This color-coded image shows the main components of our microscope stage. The gantry is constructed out of ThorLabs XT95 rails (yellow). The objective is mounted on a linear translator (red) at the gantry's center. A 24′′ by 24′′ breadboard forms the X/Y stage (1. purple, ThorLabs PBG11105). A raised sub-stage (2. purple, ThorLabs MB1224) brings the specimen up to the level of the objective which is about 13′′ above the surface of the air table. The specimen is mounted on an independently movable platform (cyan), allowing its position to be manipulated manually with respect to the rest of the stage. Our motorization hardware (motors, couplers, and flexible shafts) are colored green.

Figure 2. X/Y Drive system hardware.

A. Two different views of the X/Y drive system hardware for one axis. A ThorLabs PT1 linear translator with micrometer (top of image) is coupled to a stepper motor via a flexible shaft (colored blue). The flexible shaft (normally curved but here shown “broken”) is attached to the stepper motor shaft and the micrometer thimble via two cylindrical male/female adapters machined in-house (colored red). For scale: holes on the PT1 stage are located 1′′ apart. B. Image of the assembled X/Y translator pair. Two PT1 stages are mounted at right angles to one another and bolted to the air-table. Cylindrical couplers are colored red. Flexible shafts (colored blue) are 12′′ long (SDP/SI. part number A 7C12-12633). Motors (OrientalMotor.com, part number PK243M-02BA) are colored green. Pitch of the holes on the air-table is 1′′. Electrical cables have been removed from this image for clarity.

Figure 3. X/Y Stage.

A. The 24′′ by 24′′ X/Y stage rests on 4 pairs of PT1 translators. One pair is actuated by micrometers and driven by the motors (colored). The other three pairs have no micrometers and the internal tensioning are removed. Placing the driver translators near the middle of the stage proved essential for obtaining reproducible motions. B. The bread-board bolted to the 4 translator pairs. Screws (colored red) highlight the locations of of the translators.

Figure 4. Z Stage.

This color-coded drawing shows the assembly of the focusing unit (the region colored red in Fig. 1). The descending XT95 gantry rail is colored yellow. A Newport 461-X-M translator (cyan) is coupled to the rail and actuated by a Newport HR-13 micrometer (green). A custom machined aluminum block (red) couples the translator to a custom machined objective holder (purple). The objective shown (grey) is the Olympus XLUMPlanFLNW 20x used for the measurements in this report.

Figure 5. USB interface and DualShock layout.

The DualShock has a mini-USB socket which allows it to be connected to the Arduino Mega via an interface board known as a USB Host Shield. A. Wiring diagram for connecting a USB Host Shield to an Arduino Mega 2560. We include this here as instructions on-line are hard to find. B. The PlayStation3 DualShock controller. The buttons currently used by OpenStage are colored. Red: shoulder buttons (L1 and R1) select the Speed Mode. Blue: direction pad, executes fixed-step motions in X, Y or Z (for Z the user holds down the Triangle button whilst pushing up or down). Yellow: analog sticks provide motion in X, Y, (left stick) and Z (right stick). Green: right-hand buttons (Triangle, Circle, Cross, and Square) allow the user to store and go to four different set point locations.

Figure 6. X/Y positioning accuracy.

A. The X/Y stage was moved 50 times between three different pollen grains on a sample slide. The sampled pollen grains were about to apart. The points show the final position of the stage and the grey lines link these points, indicating the path taken by stage. At each of the three points (labeled B, C, and D in reference to the sub-plots with which they are associated) there are 50 data points, which at this scale can not be resolved. B-D. The positioning errors at each pollen grain. The colors indicate cycle number, with black being the first observation and white the 50th observation. The distribution of the colors indicates that the positioning errors are not random, and the stage drifts about over these 50 positioning cycles. E. Close-up image of the pollen grain at position C. This image is obtained by averaging the raw frames and so it is slightly blurred due to the positioning errors. F. Same as E, but frames were aligned before averaging and so slightly more detail is visible.

Figure 7. Validation of Z-positioner.

A. Single frame obtained by imaging a fluorescent slide which has been tilted by raising it on one side. The slide's tilt leads to a fluorescent vertical bar, since regions of the slide to the right and left of the bar are out of focus. B. Cross section of the bar's position as the objective is advanced in steps. C. Measured position of the Z-stage as a function of commanded position as the stepper motor is advanced in full steps (0.9°, which should result in motions of ). The slope of the regression line (red) is almost 1.0 (dashed line), indicating that the stage moves close to the commanded values. D1. Smooth motion of the bar over time under the control of the DualShock gamepad's analog input stick. Each row corresponds to a different frame. D2. Displacement of the objective as a function of time. Linear regression in red. D3. Residuals of regression from D2. The small periodic errors are due to non-linearities across the stepper motor's micro-stepping cycle.

Figure 8. Z-Stack behavior.

A. Position of the fluorescent bar as the objective is commanded to upwards in ten steps of . Three cycles of motion are shown. Each point represents data from a single frame. Grey points indicate frames when the Z-stage is moving. Colored points indicate frames when the Z-stage is stationary. Points at the same depth share the same color. Upward motion is indicated by more negative numbers. Positioning accuracy is unidirectional, since the objective always approaches each depth from the same direction. B. The location of the objective over 100 Z-stack cycles. Each point represents objective position from one cycle of one depth. There are 100 points for each depth. Motions are highly repeatable over time. Colored lines are linear regression fits. C. Correspondence between target and achieved position. D. Achieved position for the Z-depth. Note the data are bimodally distributed. E. Same data as D, but plotted as a function of stimulus repeat. Different symbols distinguish between data obtained in the first and second blocks (10 minute gap between blocks). F. Same as D, but for the depth.

Figure 9. Z-stage drift and absolute positioning accuracy.

A. Z-stage drift over a two hour time period (black trace). Each red point represents the return position after one down/up cycle, during which the objective is lowered then returned to the initial position. Blue background indicates time periods over which the room air conditioning is active. Green background indicates time period over which the air conditioning is deactivated, and the room warms. Upward motion is indicated by more negative numbers so, as expected, the objective rises when the room warms. Most of the gradual drift in objective position seems to be due to factors not directly related to motion of the drive system. B–E. Detail showing the motion epochs. Red indicates period during which objective motions are being executed. In three cases (arrowed) the objective undershoots on its return by about 0.05 to but rate of drift does not alter during the motion epoch. Thus, these larger amplitude motions are conducted with an accuracy similar to that of the Z-stack motions (Fig. 8).

Figure 10. Backlash and bidirectional repeatability.

A1. Configuration of the shaft for the data shown in A2 and A3. A2. Desired (black) and achieved (red) objective positions. Each point represented one half step. A3. Detail from A2. Backlash spike is arrowed. B1 to B3. Same as A1 to A3 but for a different shaft configuration. C1 and C2. Same as B2 and B3 but with backlash correction implemented. D1 to D4. Bidirectional repeatability for larger amplitude motions ( to ). Each panel shows the return position of the focuser. Data are divided according to the direction from which the focuser approached zero. Each data point is a different return cycle. The bars show mean (red line), 95% confidence interval for the mean (pink area), and 1 standard deviation (blue area). All points being at zero would indicate perfect performance.