I got the X and Y limit switches working. Not perfect, but good enough to move forward. Spent yesterday coding “go home” “go center” and a preliminary “scan slide”.
Slide Center is currently hard-coded based on “eye balling” one of my sample slides. I’ll have to come up with something more scientific. 🙂
The level of magnification determines the steps per photo — inverse ratio (lower the mag, higher the # of steps). It follows the mag values, but I’ve tweaked to give some additional overlap. The steppers and gearing are good enough for over 1000x (the optics, not so much).
I’ve also directly connected the camera to the microscope (not going through an eyepiece). These images are roughly 300x. 3D printed a little 32mm adapter tube.
As I play with it, I’m coming up with a list of stuff to do: add a Cancel/Abort/Panic button (hooked to an interrupt). Allow user to set scan ranges. Allow user to set center. Save default values?
Also thinking through a proper case for it.
Of course, also working through how I’m going to trigger cameras.
It’s so simple a problem that it’s a giant pain in the neck! To further automate the microscope, I need a way for the program to find a “home” position — a starting point. Something that won’t strip gears, crack slides, or otherwise cause mayhem.
Most X/Y/Z devices (3D printers, laser cutters, CNC routers, etc) use limit switches. The trouble is the way the microscope stage moves.
X – Moves just the slide left/right Y – Moves the stage forward/back Z – Moves the entire stage up/down
So, you either start attaching switches in weird places (on the stage, on the body) or come up with something else. Added fun? The top limit of Z should change depending on the objective lens. And, by the way, not the entire stage moves for Y — just the very top of it. But, Z does move everything. You start to see my problem.
How to cleanly find home, inexpensively, with as little modification to the microscope as possible?
I’ve looks at more sensors than I wish to admit:
Time of Flight (Laser) — Good for finding a single direction of movement, but I’d need three at $15 each. I tried some ideas of using just one for X and Y, but that didn’t pan out.
RGB Sensor Through Eyepiece — Find the “dark” / black and you determine the edges of the sample. Works fine, in a controlled (dark) environment, but reflected light (from the room light, bouncing off the top of the slide) messes with the readings.
Hall Effect Sensors — Like switches, but triggered by magnets. Most promising, but when Z is moved, their point of triggering changes. I pretty much need a constant/known distance between magnet and sensor.
Accelerometer / Magnetometer — AKA Gyroscope and Compass — Good for finding Yaw/Pitch/Roll, but since the microscope doesn’t Yaw/Pitch/Roll, not helpful for me. The stage movements are too small to accurately tell if its moving in a particular direction. Perhaps this is a solution, but I’ve not figured out how to tease out good data.
I was also thinking of a proximity sensor, like the one found on our Prusa 3D printer. Kinda like a Hall Effect without the magnets. But, the sensors aren’t cheap ($15-$30 each), pretty big, and I’d need three. And the whole range of motion problems and where to mount them.
So, for now, I’ve admitted defeat and gone back to micro switches. I’m hoping to get them working “good enough” so that I can write code. For now, the focus (ha-ha) is on X and Y. I’ll add Z later. (Z is less important, as the focus really needs a human eye, or smart AI.)
I thought the TCS-34725 RGB color sensor (https://amzn.to/2M0ZqzB) might be the solution, but it’s not. I printed a little slide carrier in black, so I could cut out the ‘scopes light when out of range. That worked great, until the next day and I had sunlight.
The sensor is good at picking up reflected light off the top of the slide. Which is fine for other applications, but no good for us. (Unless we run the ‘scope in a dark room.) I’m not able to get clear in/out of range readings with the reflected light getting in the way.
I still may use the sensor, as a very-rough “print preview” of the scan. It does change values as I move across a sample. However, if ambient light changes, you get shifts in the color. So it’s not very repeatable.