This is going to get long, so I'm going to break it into chunks that you can skip to if you are interested in a certain part of the project. 

Design Goals

I didn't have all that many specific goals when I first started thinking about designing the printer, but there were a few broad things that were important to me.

The first priority was to make a properly built machine. What I mean by that is I didn't want sloppy bearings, misaligned belts, a wobbly frame, or unnecessary cantilevered forces present in the design.

The second priority was price. If I couldn't make it affordable, then I simply couldn't build it. However, this goal conflicts with the first goal of making a properly built machine. I didn't have an exact budget, but I wanted it to come in at $350 or less.

The third and final priority was modularity. I absolutely love my Prusa MK3 (thanks Hackaday!), but it's a workhorse that isn't designed to be constantly modified and tinkered with. I wanted to make my delta a platform that could serve any purpose that fit its motion system.

What Went Well

These effector arms are awesome: https://www.filastruder.com/collections/electronics/products/12-magballs-6-360mm-arms-for-delta-smarteffector. They are stiff, super smooth, and have surprisingly strong magnets. Thanks to those magnets, it will be easy to create different effectors that I can swap out quickly and easily.

The meter-long 20x60 OpenBuilds rails serve as both the frame and linear guides for the printer. The whole printer really hinges on these rails. It would cost WAY more to build a printer this size with quality smooth rods and bearings.

 I used 4 V-wheels for each carriage to guide them up and down the rail. They only cost about $20 for a pack of 12 off of Ebay. In my opinion, getting knock-off wheels is the way to go. The wheels I got work well enough and saved me $40 compared to the genuine OpenBuilds ones.

Surprisingly, I was able to use an 8-bit board to run the whole thing. I know it could move much faster and at higher micro-stepping if I upgraded to a 32-bit board, but an Arduino Mega and Ramps 1.4 board combo only cost about $20 compared to a Duet 2 Maestro that costs about $130.

I used crowned pulleys to keep the belts centered without flanges. If you're unfamiliar with the magic of crowned pulleys, there is a good explanation here: https://woodgears.ca/bandsaw/crowned_pulleys.html. They do their job well and the belts don't grind against any flanges.

Putting an E3D Titan directly on the effector actually worked out pretty well. I know I'm limiting myself a bit when it comes to acceleration and jerk, but it's still very nimble. Also, I like not having to deal with a super long bowden tube.

Using MDF to hold the frame together at the top and bottom was a cost-effective way to make a very rigid frame. It only cost about $12 to get a 2x2 foot 3/4 inch thick slab of MDF, and my awesome brother was able to get it routed for me on the machines he has at work. Thanks, Bro!

What Went Poorly

OpenBuild's 14-tooth pulleys didn't work out like I had hoped. I might be missing something, but the 3 pulleys I ordered all had incomplete threads for the set screws. I was able to run a normal m3 bolt through a couple times to clear the hole a bit, but I still stripped the heads on a couple of the set screws. It's not that big of a deal, but for $5 apiece I would expect something better.

My belt clamping mechanism also needs some work. I patterned a GT2 profile into the carriages, which worked well enough, but how snug the belt fit into the profile was inconsistent from print to print. Sometimes it was uncomfortably loose, and other times it would be so tight that I was unable to push it all the way in.

Future Plans

Though the 1.8 degree stepper motors I took off of my scrapped Prusa i3 clone work well enough, I can see why 0.9 degree steppers are pretty much the standard for deltas. 0.9 degree steppers are pretty cheap from the Stepperonline Ebay store, so that is definitely an upgrade I plan on making.

On a related note, I think switching to a 32-bit control board is a low-priority upgrade that I will make at some point. The 8-bit one works fine, but I don't trust it to run overnight since it's a low-cost Ramps 1.4 clone sitting directly on top of flammable MDF. If I end up doing a lot of overnight prints, I might upgrade sooner rather than later, but for now I think the 100+ dollars that a quality 32-bit board costs would be better spent elsewhere.

I think swapping in a E3D Volcano is an upgrade that I want to make fairly soon. The high volumetric output of the Volcano would allow me to take advantage of the speed and build volume of the delta much better, and the loss of a few millimeters isn't really a problem when the build height is well over 300 mm.

I want to re-design the carriages at some point as well, I think I'm going to use a clamp design instead of a GT2 profile for holding the belts. Also, I want to create a way to adjust the wheel tension. In the first version, I relied on getting the spacing of the wheels just right so they didn't have any wiggle but also weren't so tight that they prevented smooth movement. I think a better approach is to use 3 wheels in a triangle formation and make one of them adjustable so the tension can be increased or decreased easily.

The Fusion 360 design isn't in a great spot right now. With all of the modifications I have made, it is slow to update when a change is made and overall is just messy. At some point I want to do a complete redesign that is built from the ground up with customization in mind.

Hard-Earned Tips

There are a few things that I wanted to know but couldn't find out easily during the course of this build. Since I don't want you, dear maker, to suffer the same fate, I'm going to put some tips in this section that you may find useful.

Before I started designing the delta, one of the things that I didn't know was what the angle of the arms relative to the effector should be. It took me a fair amount of googling before I came upon one guide that stated the length of your arms should be approximately 80% of the distance between the towers of the frame. I went the other way and divided the length of my arms by 0.8 to get the distance between the towers, and it seems to work well enough. At some point, I want to experiment with different ratios to find what the reasonable limits are, but for now 80% works fine.

This next tip is specific to the people that are building a printer that use OpenBuilds style rail and V-wheels. In my experience, attaching the V-wheels to a carriage before sliding it over the end of the rail can put small indentations in the relatively soft delrin plastic the wheels are made of. This can cause the wheels to make a slight "thunk thunk thunk" noise and not roll completely smoothly. I would recommend assembling the carriage on the rail if possible, or at least file down the end of the rail where you slide the wheels on.