Week #5

🛠️ Hardware

We began working on the attachments for the last stepper motor, which includes a redesigned slanted ‘chair’ with tracks along the sides for us to secure a 3D-printed part to hold the stepper motor in place.


We also redesigned the camera mount to fit better, with a little easter egg on the inside 😉. The final camera adapter fits snugly.


However, we did run into some unexpected problems. The platform could not handle the weight of the motors and succumbed. We reprinted it, but got a rude shock the next day (our one day off!). We returned to this:

Our print got destroyed 🙁


Mind you, we printed a platform, not Shrek’s dinner! The printer is probably jealous of our success and wants to give us some setbacks. The printer is a hater 😡. It’s safe to say we will not be using that printer again.

We also worked on a solution to fix the slanted chair in place. Our initial prototype involves a holder which is screwed into the wood. However, due to the chair being top-heavy, it didn’t work very well as the tower was still shaking.

Initial prototype of our holder


💻 Software

I started the week off with an insane discovery… The Arduino code for the stepper motor was wrong the whole time… After making the change, the motors finally worked properly.

We also fixed the auto-focusing algorithm. By changing the way we calculated the sharpness of the image, we could consistently find the best position at which the platform had the best focus.

In more technical terms, we used the Tenengrad method instead of finding Laplacian variance of the image, which gave a better measure of the sharpness for our use. Although calculating the Laplacian is a very commonly used technique, it failed in our use case. The reasoning for this was because the Laplacian variance calculates the overall detail of the image, while the Tenengrad method calculates the strength of sharp edges in the image. Since our microscope images (as seen below) have a large dark background, it has overall low variance despite having sharp edge at certain focus levels, while the Tenengrad method can pick up on the strength of these sharp edges.

Now, all we have left is to print a screw mount for the lens rotator motor, before Cooper can finally start full test runs.