Analyzed CD using home made microscope, compensate tilt by rotating sample

In metrology tool there is always artifact on the tool. especially there is artifacts coming from mechanical tolerance or calibration. Mechanical stuffs are usually calibrated but it's difficult to be handled as it needs tool down and limitations in physically controlling small things. I wondered whether it's possible to compensate it by measuring images in several perspectives and processing it. so in here I tried it by rotating sample. I rotated sample and grabbed the images then compared how it looks like. 

 

 

Here is some updates about the HW side. 

Rotation stage for sample is added as like image below. Actually, it's just a motor, not enough to be called as stage. I got the motor from the CD ROM of old PC. Before finding this motor, I tried other DC motor which is commonly used for RC car, but those motors are too week and fast to rotate the sample reliably. But this motor from CD ROM is quite nice. It's strong enough to start rotation and can even control the angle, altough 1 tick is 60 degrees. Motor is controled with motor driver L298N and Arduino Uno.

 

Here is some analysis of CD

I grabbed the images while rotating the sample. sample is rotated by 60 degree everytime. Below images are the raw image without processing. By putting the axis of rotation in the FOV, I could keep same feature in FOV for all of the images during the rotation. 

 

 

Although same feature can be obtained while rotation, the image looks different in reality. Below 6 images are images grabbed with 6 different angle(0, 60, 120 ...) and profile is grabbed from the middle of the images to compare how it looks like. As you can notice on the profile, i) p2p of image is different, ii) shape of top is different; some profiles are round while some have 2 peak. iii) releative intensity of each peak/valley is different. It is expected behavior though. There must be something that's not on axis, and it will make difference between 0 deg and 180 deg measurement, so as the other angle. 

 

What I did was averaging the images measured with different angle. I first averaged the image of 0 deg and 180 deg. Below image is the result of it. Same images were used for the analysis although ROI is different. As you can notice, image looks much uniform inside the rectangle comparing to the image with single angle. But you can still notice the non uniformity in the profile. Observing the Profile on the dark area, local peak in the dark area is shifting in 1 rectangle as like indicated with red arrow.  i.e.) we still have non uniformity although some of it is canceled out by averaging 0 deg and 180 deg.

 

 

As conclusion, I analyzed the images grabbed with different rotation angle and confirmed that Image can get better, removing the non uniformity coming from tool artifact, by averaging the images grabbed with different angle.