Michael Juan
I'm Michael, I will be working on the litho-stepper
Last updated
I'm Michael, I will be working on the litho-stepper
Last updated
Update 0
Focus for this week is to create a plan to quantify the errors in the current litho stepper. Work on my tormach 440 cnc machine to have "in house" machining capability.
Preliminary Readings: ISO 230 Geometric accuracy of machines operating under no-load or quasi-static conditions, Foundations of Mechanical Accuracy, Precision Machine Design.
Create a project proposal
Created project proposal, received input from litho-stepper team and edited the proposal by added target values for mechanical accuracy of proposed nano-positioner.
Welded a steel coolant tank using TIG (tungsten inert gas) welding. Leak tested the coolant tank by filling it with water and waiting to see if the tank leaked. There were several pinhole leaks.
Created spreadsheet of potential tooling required for CNC milling nano-positioner parts.
Started on creating the documentation for measuring the mechanical accuracy of Stepper V2.
Problem 1
Coolant tank was not watertight. suspected cause was contamination from inadequate surface prep as well as skill issues with TIG (tungsten inert gas) welding. An attempt was made to braze the locations of leaks but a second leak test was not attempted due to time.
Proposed Solutions
Grind out areas that have leaks, weld the leak locations again.
Buy a coolant tank.
Problem 2
Did not finish documentation on test cases due to time overrun with coolant tank task.
Proposed Solution
Carry over task to the next week.
Problem 3
Unsure about who reviews my gitbook updates as well as github project tracker usage.
Proposed Solution
Ask at the next meeting.
Measurement:
Measure the mechanical accuracy of Stepper V2.
Machining:
Order tooling.
Order material.
Tram the head of the CNC machine.
PI tune CNC machine spindle motor.
Nano Positioner:
Create a plan for machining mechanical parts of the nanopositioner.
Work on CAM (computer aided manufacturing) for one loose tolerance part.
Carry Over Tasks:
Create and document plan to measure the mechanical accuracy
It is important to note that I am following axis orientation of the Stepper GUI, which is not consistent with industry norms.
Link to test procedure document:
Created a SOP for checking the parallelism of wafers.
The highest point and the lowest point were 7 microns apart. This could be due to residue from the cleaving the wafer, manufacturing tolerances, or small amounts of contamination that was not cleaned off through washing with acetone and isopropanol.
Did an informal test on contamination. Sharpie marks are around 2.5 micrometers, and finger smudges are measurable under 0.5 micrometer.
Tested the backlash on the x and y axis of stepper at 10 positions each.
The fixes for the X and Y axis by Carson resulted in less that 10 micron backlash.
The backlash on the Z axis was so bad that it over-traveled my indicator. (will test it after a proposed fix)
Tested the step accuracy of the stepper in 10 micron, 5 micron, 2.5 micron, and 1 micron increments.
Steps were consistently inconsistent. (see linked spreadsheet)
At smaller steps such as 1 micron there would often be no movement even after a command.
Did initial testing on squareness of axes. Initial testing with a 2 micron indicator and a granite square suggests that x and y axis squareness is not a issue.
Did initial testing of how parallel the vacuum held wafer was to the axes.
I did not bring my course indicators because I assumed the error would be in the micron range. The error was higher than what my indicators could measure.
Initial testing suggests that the vacuum is bending the wafer.
Performed manual PI tuning for CNC machine spindle motor.
Problem 1
Did not get to test positional repeatability of axes. repeatability of limit inductive sensors, and hysteresis of limit sensors.
Proposed Solutions
Perform test on Tuseday 02/04/2025
Problem 2
Testing by manually typing g-code is slow. Even though most of the time spent was manually adusting the dial indicator there is speedups fr
Proposed Solutions
Create a gcode script if testing is going to be an ongoing thing.
Measurement:
Test the positional repeatability of axes, limit inductive sensors, and hysteresis of limit sensors.
Converts notes from testing to repeatable SOPs.
Test longer travel distance accuracy of axes.
Machining:
Order tooling. (carry over)
Order material. (carry over)
Tram the head of the CNC machine. (carry over)
Nano Positioner:
Work on CAM for the top plate of open source nanopositioner. Because there are no tolerances in the paper create a dimensioned drawing with best guess tolerances.
Work on creating a simpler way of testing piezo nano positioning. (Sanity check)
Inaccuracies of small steps (<10 micron) seems to be from motors lacking torque to micro step. friction/binding in the system.
Mechanical accuracy of z axis is so bad that its functionally unusable. (a fix has been proposed and is currently being implemented by Carson )
The axis orientation of the stepper do not follow industry norms. This should be updated to prevent confusion.
even though the surface of the wafer is almost atomically flat. the bottom and top layer are not necessarily parallel.
The 3d vacuum wafer holder is tilted significantly.
Important Notes
The stepper components and axis orientation were changed between the last measurement and the tests performed on 2/8/2025.
Worked on documentation on test procedures for measuring stepper.
finished positional repeatability measurements of x and y axis.
finished repeatability testing of homing switches.
Attempted CAM (Computer Aided Manufacturing) on open source piezo nano-positioner.
There were some DFM (Design for manufacturing) issues that makes the design unnecessarily complicated.
read papers on friction drive nanopositioning.
3d printed the stl of nanopositioner. (printing it feb 10 morning)
Problem 1
Did not order some tooling and material for nanopositioner.
Proposed Solutions
add items to purchase sheet before tuesday.
Problem 2
The parts specified on nanopositioner have micron level tolerances. Therefore the nano positioner when stacked on top of each other might not be anywhere close volumetrically to nanometer accuracy.
Proposed Solutions
talk to team about it. Should not be a big problem.
Problem 3
Have big due date for major courses. Have to set up things on thursday for design school career fair.
Proposed Solution
Let leads know on tuesday meeting.
Measurements:
Check with team on proposed positional accuracy of redesigned nano-positioner.
Check with team on proposed redesign of piezo nano-positioner.
Machining:
Add spotting drill, drills, taps, chamfer mills,ball endmill, roughing endmill, and collets to purchase sheet.
Nano Positioner:
Add pre-ground aluminum bar to purchase sheet
Machine bottom plate part when end mills arrive.
Will be a lot of work.
Probably going to use sacrificial workholding instead of making custom workholding.
Worked on CAD for redesigned nano-positioner
Worked on test procedures for positional accuracy.
ran test cuts on scrap aluminum
Problem 1
Tooling for machining did not arrive.
Proposed Solutions
Machine the week of feb 17
Problem 2
Did not complete CAD for nanopositioner to a suitable state to machine.
Proposed Solutions
Work on cad before presentation date.
Problem 3
NanoPositioner linear rail out of stock
Proposed Solutions
Design around an alternative THK bearing slide.
General
Prepare presentation.
Measurements Finish up documentation for positional repeatability, step accuracy. and backlash measurement testing.
Bought an LVDT probe that "should" be able to measure double digit nanometers across very short distances. If time allows will redo some measurements.
NanoPositioner
should be able to make all the parts for one axis of nanopositioner IF tooling arrives.
micron motion on piezo actuator!!!
Used a function generator to output a 12v square wave to piezo.
Problem 1:
Did not get anything else done other than measuring motion with piezo element.
Solution:
revaluate timeline.
Problem 2:
Got scammed by ebay seller. The lvdt was not the high precision version.
Solution:
Buy another lvdt probe off of ebay.
Try to measure movement accuracy of previous years piezo actuator.
NanoPositioner
if pcbs arrive start assembly.
Started soldering circuitboards for piezo driver.
Worked on a design for a single axis piezo actuator. Below are some design considerations and criteria.
Design has to be easily machinable.
uses bellville washers for preload (simple to adjust when compared to coil springs)
for tight tolerance portions uses premade bushings loctited into bearing block.
tries to minimize multiple machining operations.
reduce/eliminate the number of machined flexures.
Let precision be dictated by the stage side. Basically have the actuator be accurate and precise only constrained to one axis and let whatever stage its actuating against dictate the precision of the rest of the axes.
Problem 1:
Mcmaster Carr materials did not arrive
Proposed Solutions
Reorder.
Problem 2:
Don't know if its possible to have piezo element with friction element in this configuration.
This puts a shear load on the glue interface between the piezo and the friction element.
In a standard configuration there is no shear load on the glue joint between the ceramic friction element and the piezo.
Reorder mcmaster material.
Order mcmaster materials personally.
Solder surface mount components on piezo driver pcb.
work on a design with a square actuator shaft rather than the cylindrical one in the above design.
There might be tigher machining tolerance requirements
might be easier to attach actuator to stage.
Work on consolidating information for measurements into one spreadsheet. Currently its spread across multiple spreadsheets.
Prepare mid semester documentation.
100 nm back and forth motion
85v output from piezo driver (pdu 150) correlated with ~100nm of motion.
2 micron back and forth motion
7 volt input and 150 output resulted in 2 micron motion.
Roadblock 1:
Trying to output a sawtooth wave through pwm did not result in translational motion. I'm not sure if this is due the piezo needing a higher duty cycle pwm signal than what an arduino can provide or some mechanical issue.
Potential Solution:
Get an arduino that has a dac.
Roadblock 2:
not a big roadblock but measurement resolution might be an issue going forward since we know that the piezo can move sub 100nm.
Order an arduino that has a built in DAC.
As a simple first step I could swap out the magnet in the system with a weight to provide more preload. This could validate my spring based preload design without actually machining parts.
A second step would be replacing contact point with a ball bearing. These can be prototyped before thorlab parts arrive.
Salvage piezo actuators off of previous years nanopositioner.
semi controlled translation movement.
Motion was achieved through a 10khz 5v sawtooth wave.
The load capacity of this stick slip mechanism was almost nonexistent. This is an issue because we have a stacked stage configuration where one stage is above another.
Swept through frequency to see if motion behaved linearly to frequency. Taking a slow motion video and counting frames seems to suggest that 10khz is twice as fast as 5khz.
Attempted to measure step accuracy. I was not able to use my dial indicators to check for accuracy because the spring loading on the measurement tools were too strong for the force of the actuation.
Roadblock 1:
Did not get quantitative information on distance traveled per step.
Proposed Solution
Use a calibration slide on a microscope to measure travel distance.
Roadblock 2:
Thorlabs ceramic contact points and piezos have not arrive yet. I've tried to salvage existing piezos but the superglue mounting sometimes breaks the piezo.
Proposed Solution
Try to get at least one working salvaged piezo.
Temporarily use a ball bearing instead of the ceramic contact points from Thorlabs.
Roadblock 3:
Adding more preload to the current flat contact point actuator just binds up the system.
Proposed Solution
When ball contact points are made or when thorlab parts arrive check to see the optimal amount of preload.
We have three things we have to test and do for the upcoming week to have progress for presentation 2.
test having a ball contact instead of the current surface contact
Test how much preload results in a usable load rating and consistent actuation.
Using a scale and weights incrementally increase until we see a usable actuation force. Usable right now should be able to overcome the spring force on my measurement tools. A reasonable target to hit might be 750g on each axis.
Test the effects of sawtooth wave frequency, amplitude and a combination of both on actuation with updated prototype.
Start off at 10khz and actuate for 500ms then measure distance.
Decrement the frequency by 500ms and actuate and measure distance.
repeat until there is no measurable movement.
For measuring the effects of amplitude on motion. Set a frequency such as 5khz, and a actuation time of 500ms. Start off at 5v and decrement voltage by 0.1v and measure distance.
repeat until there is no measureable movement.
Because shipping is outside of my control. these are things I can work on while thorlabs parts are in the mail. I think right now my time won't be best spent on doing CAD because there are a lot of variables we don't know yet that will end up influencing the mechanical design.
Write code to interface piezo actuator with our motion controller.
We use grbl for our motion controller. As far as I know grbl can only output step and direction.
The first step would be to find a good ratio to map the step signals from the controller to a frequency of sawtooth waves.
Test the effects of sawtooth wave frequency, amplitude and a combination of both on actuation for the existing prototype.
Repeatable motion.
Because the current piezo prototype does not have enough force to move a dial indicator we put it on a microscope and measure motion from the microscope.
A rigid wire was glued on to use as a pointer.
A dial indicator was set against the microscope stage and after each movement a measurement was made using a mahr supramess 0.5 micron indicator.
with a 50ms step at 5khz we were able to get a repeatable 18.5 micron motion. repeatable to the half micron.
Note that this is a comparative measurement because we are measuring the movement of the microscope stage rather than the piezo stage itself.
Got cad to a good enough stage to machine. All parts for machining and assembly have arrived.
This iteration differs from previous designs through the use of a flexure.
The carriage of the linear rail is rigidly attached to the stage. and the linear rail is the moving component.
Preload is done through spring washers.
The piezo module chosen is a shear piezo from thorlabs. Even though a shear piezo is more expensive it makes machining and assembly easier. Having the glue surface normal to the preload direction means that the piezo is not under any shear loads.
Sacrificial softjaws were made to facilitate machining operations. these jaws are meant to hold on to the flexure component of the actuator for second operation machining.
Roadblock 1:
Did not measure the effects of frequency on motion.
Proposed Solution
Ignore task for now because making a improved prototype is a higher priority.
Roadblock 2:
Did not get step and direction from grbl to interface with piezo driver.
Proposed Solution
Like roadblock 1 ignore task for now because making a improved prototype is a higher priority.
The next steps to get a working piezo actuator include making the workholding to machine the backplate (see referenced 3d cad), flexure. And programming the machine to make the parts.
The backplate workholding is relatively simple because the second operation does not require any sort of precision.
The flexure on the other hand is a relatively complex part. it requires machining very thin walls which can cause vibrations.
Because I focused on CAD and CAM for this week I will need to work on software for converting step and direction to sawtooth waves as well as testing the effects of frequency and voltage on motion after machining the updated prototype.
Links to CAD and CAM files
Set up tooling on CNC machine as well as CAM (computer aided manufacturing) software. Each tool's height was measured to 10 microns and runout of the tool in the toolholder was dialed to <5 microns.
Programmed workholding to machine the backplate and flexure.
Machined backplate for positioner
Backplate was machined in three operations from 6061 t6 aluminum. Aluminum is a good choice of material because its easy to machine and relatively strong.
The first operation included all of the tight tolerance features. By designing all reference surfaces to be machined in one operations errors from removing and re-clamping part in a vise are reduced.
The second operation removed the part that was clamped in the vise. This operation has the part flipped and clamped lightly to reduce the effects of the clamping force of the vise on the final part dimension as well as allow for internal stresses in the material to relieve themselves.
note the area surrounding the part, it is not machinable in the first operation because it is clamped in a vise.
The third operation drilled a hole to allow for a screw to hold preload springs.
Threaded holes were hand tapped.
programmed and machined flexure piezo mount.
machined from 1045 steel with a 0.4mm thick flexure. 1045 steel was chosen because aluminum is not a good material for flexures.
The part was done in 4 operations. The first operation drilled and interpolated m5 holes for mounting onto the backplate and for work holding for the second operation.
The second operation had the part mounted on a sacrificial aluminum block. This allowed the flexure part to be supported while being machined. To hit a tight tolerance on a thin part the flexure was machined with 0.1mm stepover and at a low 100mm/min feed rate.
The third operation drilled a tapped hole for spring preload.
A piezo was glued onto the flexure and a sma connector soldered on.
By reducing the amount of tight tolerance features to two straight edges the cost of the actuator is significantly lower than that of the open source paper I originally referenced.
Each mounting hole is drilled to a clearance to allow for adjustability.
Adjustment is made by using gauge blocks. By using a known standard, that has traceable length and parallelism. It reduces the need to machine to a tight tolerance therefore reducing cost.
Roadblock 1:
The shear piezo I wanted to use did not move when tested. This is probably because shear piezos have a shorter travel distance. 1.3 micron movement was not enought to get translational motion.
Proposed Solution
Modify design to use standard piezos that have longer extension (6micron)
Roadblock 2:
Did not get to testing translational motion on new machined positioner.
Proposed Solution
Test translational motion on 4/7. Calibrate distance traveled to number of waves similar to what I did in the previous update.
Test the effects of different spring loads on new positioner. this can be done by swapping out springs and or adding weights.
Test the effects of frequency on travel distance and speed. Originally I was also going to test the amplitude of sawtooth wave but because of the resolution of my measurement equipment the effects of amplitude are not going to be measurable.
Prepare presentation 2.
Tested different spring rates
achieved through gradually reducing the flexure thickness by 50microns at a time until actuation was achieved.
because my cnc machine is not double digit micron accurate each reduction in flexure thickness had to be measured by hand using a dial indicator.
Two sawtooth waves per 500nm. ~250nm per actuation.
Frequency of sawtooth waves were 10khz and voltage was at 100v.
adjustment was made by stacking gauge blocks and moving the flexure down 10 microns at a time.
Tested non precise translational motion
Roadblock 1:
Linear rail surface was not a precision surface. Therefore actuation was only consistent in millimeter ranges.
Proposed Solution
Choose a tigher tolerance bearing, most importantly having the surface that the piezo actuates upon to be a known precision surface.
Roadblock 2:
Wear from ceramic on the piezo rubbing back and forth on steel linear rail results in an unusable linear rail after a couple minutes of actuation.
Proposed Solution
Switch to a ceramic or tungsten carbide bearing.
The first thing to relatively easily prototype is having a two stage system. With course adjustment done with micrometer. and fine adjustment done with piezo. This would require the current stepper stage to have a slightly better movement resolution. We would need a +- <3micron resolution on the stepper to fit within the movement range of a piezo displacement.
Assemble laser interferometer kit.
Fix design flaws in current piezo positioner design. Change the bearing choice. Maybe design it so that preload can be adjusted without having to go through the multi hour long process of using gauge blocks.
Designed a fine positioning flexure stage
The stacked stage approach proposed in the previous update is relatively easy to implement. But the issue with having the piezo in line with the micrometer is that the errors of the micrometer stage are not insubstantial when attempting sub micron positioning.
Another issue with this approach is that there is slop in any sort of ball bearing based system like the cross roller bearings used in stepper v2. This slop takes away the already small displacement of a piezo chip. This may result in the piezo actuation not having enough travel to compensate for the limited resolution of the micrometer stage.
By using a flexure xy stage stacked on top of our stepper stage we can remove issues with bearing tolerance. The image below is the initial design of a flexure stage. Size is 75x75mm
This flexure is different from normal XY flexure such as the one below because it does not compensate for rotation when actuating. This should not be an issue because of the limited travel of the actuation. Without using FEA analysis and just treating the flexure centers as joints the rotational error should be less than 0.006 degree per 6 micron piezo displacement.
Roadblock 1:
Only have one working 150v piezo chip. need 2 for each prototype (4total). Do not have the right type of steel to make flexure.
Proposed Solution
added more piezo chips and materials to purchase sheet. hopefully they arrive before final presentation.
Roadblock 2:
Hitting the limits of measurement resolution. The highest resolution indicator I have is 0.5 microns and you can interpolate it visually to 0.25 microns. But we know that piezos can have much smaller motion.
Proposed Solution
Show in documentation why I believe that the designed flexure stage can have double digit nanometer motion. Use non offgassing epoxy in assembly so that future groups can test it with
Program and machine flexure based piezo positioner.
Wait for materials to arrive.
Make fixtures for machining positioner. Because flexures are very thin its important to hold each part of it rigidly so as to not introduce vibrations.
Work on documentation for stick slip piezo design.
Specify new bearings.
Specify new wear surface
document how I got from no actuation to having actuation.
Measure the difference between the piezo "dry" actuation and when its mounted in line with micrometer stage. (note that measurement tool resolution may be an issue)
Programmed the CAM file for machining for the flexure.
Redesigned the flexure to use aluminum rather than steel. Because materials did not arrive.
Tested the flexure dimensions for aluminum.
Did the same thing as the previous flexure design. By gradually reducing flexure thickness until I get the largest usable flexure thickness. (this took 4-5 hours)
Machined off 50 microns at a time until a working thickness of 0.35mm.
tested aluminum flexure on the previous stick slip stage.
Measured a piezo without gluing it onto a flexure it behaves linearly at least with the available measuring tools.
Roadblock 1:
Materials did not arrive. Don't have material to make flexure positioner nor do I have enough piezos.
Proposed Solution
Work on documention for piezo stick slip design.
Roadblock 2:
Had to work on other courses because took a week off from my other classes to work on hackerfab things.
Proposed Solution
finish hackerfab documentation on wednesday.
Program and machine flexure based piezo positioner.
test flexure positioner
work on final documentation
upload cad in a universal format to git repo.
work on 2d dimensioned drawings with critical dimensions and notes.
Looked at the cad files for this open source piezo nano-positioner.
Link to testing results spreadsheet:
Worked on a modified design for a single axis nanopositioner, (modified from ).
followed test procedures from pdu 150 datasheet.
Try to test out my theory that inconsistent motion that people have seen when trying to replicate r is due to a magnet based preload.
Links to CAD Files
h
Modify flexure to allow for a standard piezo versus the Shear piezos I tested . This will require machining a new part to attach the friction element perpendicular to the axis of motion.
Modified positioner to accept a standard 150v piezo from thorlabs.
Achieved sub 500nm actuation.
A second idea that is worth exploring is using a piezo to rotate a micrometer. I think it would not be reasonable to get a working design of this by semester end but gathering literature on this might benefit future efforts.
