Supplies: Non-Proofgrade 1/4" Acrylic - Op time: 28 mins
I am in the process of finalizing a PCB that I’ve been working on that’s designed to house a stepper driver and a closed loop package. I placed an order with OSHPark and got the PCB about a month ago. I’ve been hand soldering a hand full of them but it’s taking about 1.5hr per board. Yay SMDs!. But why am I doing this by hand? So I started to do some hardcore research on Pick & Place machines and decided to build one myself. Well partially. I currently have a ‘light duty’ laser cutter aka ‘Pew Pew’ in which I am going to repurpose it’s inner workings to a PnP. I have about 85% of it done. But I was having a tough time with the camera alignment.
So I decided to make a targeting grid. Midway thru the design I wanted to add the ability to test and calibrate these new PCBs on an existing CNC. I scrapped the design and I loaded up my mini pallet that I made for the smaller CNC and designed the grid on top of it.
Then created a new sketch and use the project tool to move the features I needed in order to cut. Then import that into AI. Then added numbers and a couple lines of text. Color coded the groups so stuff would cut in the order I wanted.
Ok… let me make sure I’ve got the gist of it. You are making a printed circuit board. Instead of hand soldering you want to use a pick & place machine to set surface mounted devices into the PCB. You are converting an old laser into a PnP. The targeting grid was initially for the PnP, but you decided it needed to be for your CNC instead, and it works perfectly. So what about the PnP? Will this same grid work for that?
Sorry, total newb when it comes to most of this stuff!
The grid will work for both machines. The Pick & Place and the CNC.
The PnP (Pick & Place) in the sense of calibrating the three cameras. And for the CNC to calibrate distance traveled and to make sure the closed loop portion is working correctly. So it’s one part that does a couple things.
High overview of that closed looping does. There is an encoder that’s located on the stepper motor. The pulses that it generates gives the controller positional data. (as a side note there are fancier encoders that also has indexing capabilities and also retains positional info powered down. These are called absolute encoders) This tells it where the shaft is at any given moment. So when you tell the controller to rotate to 16.8 degrees, it rotates to the appropriate steps then the encoder checksums that distance and corrects any deviations. Killer for applications that require crazy tolerances.
Very cool!!! Would love to see the PnP when it’s working!!! Have been thinking of doing the same. Great to see the accuracy of the forge. Would be great if the GF could be a PnP too with the swapable head! Also, what’s on the end of your air nozzle, a chiller?
Great work!
There was a PnP booth right down the aisle from the GF booth at MakerFaire NYC, which only cost $4000 (that seems to be the cost of the new personal versions of industrial machines in that area, since they were right next to Wazer).They were live PnP’ing blinky badges for you, which was pretty cool.
Seems like you are comparing the accuracy of your closed loop system with the accuracy of Glowforge’s open loop system.
I have often pondered how you get an accurate optical grid to calibrate a Cartesian machine with a camera. I will probably start with a sheet of graph paper but that expands and contracts with humidity. Another idea was to print a grid on a plastic sheet with a 2D laser printer, but how accurate are those? My latest one has a row of UV LEDs rather than an actual laser, so I expect the spacing of those is very accurate but who knows?
I would love to be able to buy an accurate grid on a stable substrate for calibration. I have a video microscope so I could also try to look at dynamic accuracy, which can be very different from static accuracy.
It probably wouldn’t cost much to engrave a grid on acrylic, Delrin, etc. with a CNC milling machine, even a router should be accurate to within 0.003" or less. I may do something like that on my CNC mill, but its work envelope is a little smaller than the GF’s bed.
If you get serious about something like that, NYCCNC on YouTube (Ohio) runs a job shop and his Haas mill is probably big enough to do a grid for the GF. (On edit - sorry, forgot that you were in the UK).
I already have a machine 10 times more accurate than that, 6um resolution, so maybe I could use that with a laser diode to make my own grid. I am surprised there isn’t something off the shelf though.
