Measure once, print twice

I have an MFT like tabletop in my workshop. It has 12 x 7 holes in a regular square grid pattern. They have been cut with a CNC router so that their placement is exact.

I have 3D-printed different devices that fit into the holes with cylindrical pins. Through trial and error I have found the exact correct diameter, so that they have the right amount of friction. That is, they are not too loose and not too tight.

I have also printed a fence that has two pins. The distance between them took quite a number of trials. Here is the strange thing... After some trials, I had the right distance so the fit was perfect where I tested the fence. But when I tried to to plug it into the other side of the table, it stuck and was impossible to fit in. I could dial in the length that fitted in the other side, but then it would no longer fit the first side. My prints had revealed that the distance between the holes varies. It is a difference that is so small that you cannot measure it with rulers or calipers, and apparently so small that it was within the specifications of the CNC router.

For most people, the 3D printer will have a better length precision than anything else in the home.

Work methods

Notice how I found the right dimensions by trial and error, multiple times. That is what I hear again and again, is how you do when constructing your own models. It must be by far the most common work method.

1. Measure once, to be on track from the start.
2. Design the model and the test model.
3. Print a test print.
4. Try it out on reality. If the fit is fine, go to step 6, otherwise go to step 5.
5. Adjust the model. Go to step 3.
6. The model is finshed, and you can start to print real prints.

The method is particularly good if you have a shape that is difficult to describe mathematically. But even when it is easy to describe, this method is good if you want the print to fit with the right amount of friction.

There are a couple of things I have found out, that can ease the work with the test prints.

Make intersection models. Cut away the parts irrelevant for testing, by making an intersection of your model with some simple figures. Often it can just be a box. Maybe two or three boxes.

Use a light color for test prints. That makes it possible to write on them with a marker. That is practical if you print many test prints. I use white PLA, which is often also a bit cheaper than other filaments.

Other methods

can be used when you are lucky that there are dimensional drawings of the thing you are making your print for. It is almost as having an answer key.

I had this luck with my mobile phone. There are detailed dimensional drawings for it. They show dimensions with a precision of 1/10 mm. The print I made for my mobile phone fit in first try.

is to use a 3D scanner. It has very high precision and measures hundreds of thousands of points without you having to do very much. It is able to measure any kind of surface, no matter how complex it is, as long as it is visible. It saves you a lot of time. That is what I have heard, because I dont have a 3D scanner myself. The problem is that such a device, of a good quality, costs 1 to 2 times the price of a new 3D printer.

It takes a lot of test prints before the economy starts to look reasonable.

leverages the high precision of the 3D printer itself. You design measuring gauges beforehand, and print them out on the printer that you later will use to print out your other models. By doing this, you will be able to measure with the same precision as your printer has.

One cool thing with this method, is that if your printer has some small inaccuracy, your measuring gauges get the same inaccuracy. They might not measure the exact correct millimeters, but they measure the millimeters that works with your 3D printer. Because then the parameters you use in your model, compensate for the inaccuracy as you measured them with your printed gauges. (Provided that you print with the same orientation, that your filament doesnt shrink, and so on.)

The fourth method should in theory be "Print a lot of times, measure once". In practice it is pretty common that the print doesnt fit perfectly, and then you have to revert to method 1. Though with a more precise starting point.