"Precision costs, how much do you want to spend?"
Once you get down to those resolution and precision requirements, an SLA printer might be the way to go. The build volume is a bit of an issue for the lower end SLA printers though. You are going to have a lot of angst using a .2mm nozzle yet alone a .05mm one! If you go that route, my recommendation is to use a Bondtech QR extruder, it is the only thing that will push plastic through nozzles that small reliably.
On the precision front you need to take a few things into consideration:
Mechanics and Motion Control. Let's look at each of these.
Mechanics
Overall Design
Delta motion is based on trigonometry. You can learn more from Steve Graves excellent
Johann C. Rocholl (Rostock) Style Delta Kinematics. But in a nutshell, resolution decreases as the effector moves from the center to the periphery of the build surface. As resolution decreases, so does positioning precision. If you look at a "Kossel style" printer you will see the columns and carriages are much further away from the edge of the build plate than the Rostock Max. Take a look at the drawing below. Remember the
Pythagorean theorem A^2 + B^2 = C^2 ? On a delta the hypotenuse (C) of the right triangle is the arms and is a fixed length. The carriage movement up and down the tower defines (A) and the constraints from the other two towers through the effector is (B). As (A) and (C) get closer together (the angles marked in the drawing) it takes less movement of the carriage up/down to have a big effect on (B) (movement in the X-Y plane of the nozzle). Think about it. The larger the tower-arm angle is at the very perimeter of the build plate the higher the precision you get.
So the take away here is, look for a delta design that positions the towers as far from the edge of the build perimeter as practical. 50% of the radius is a good compromise - this is what the Kossel printers and variants use.
Mechanical Precision
Next, in order to get the best precision you need to eliminate non-controlled motion in the system - commonly called backlash or slop. This requires freely moving yet non-sloppy joints and bearings. The areas to focus on are the carriage "wheels" and the arm joints. I say "wheels" because these could be bearings like in a linear guide. The absolute "best" in terms of precision are high quality linear guides. The first Mini Kossels used these. I have them on 2 of mine and they are remarkably precise and quite machines. Moving away from linear guides are the common "V Slot" style of wheels that run in the grooves of the tower extrusions. The Rostock Max's carriages are a variant of this. However, only 3 wheels are needed - 2 on one side, 1 on the other. More than 3 wheels is over constraint and makes it more difficult to adjust the system for free motion without excess slop or unanticipated movement. Here is a simple but good explanation of how this applies:
Over Constrained Linear Systems. Put simply, 3 wheels on the carriage with a simple adjustment to remove slop is ideal.
Finally, the arm joints themselves. There are a couple of options that have been used on Delta printers:
1) Johann originally used Traxxas ball joints. These are simple and reasonably effective. There are other brands, some that are much better. The Achilles heal with this form of ball joint is that they are typically not adjustable (at least not in the small sizes we need and for a reasonable cost).
2) magnetic ball joints. These are gaining popularity and have some significant benefits. Magnetic joints are virtually backlash-free. The only (perceived) negative is that they do have the possibility of popping apart.
3) the SeeMeCNC ball cup joints. These are a great design and have the best attributes of the simple ball joints and magnetic ball joints. The cross-arm "spring" removes backlash. The only down side is they aren't really RepRapable at this point although you can buy the barbells individually and TrickLaser now sells just ballcup arms so you could fabricate your own within the size constraints of the barbells.
Motion Control
Next let's take a look at the motion control. Carriage positioning is a function of the stepper motor resolution and the diameter of the pulley on the stepper shaft. Steppers come in 2 primary flavors - 1.8° and the newer higher resolution .9°. These translate as:
1.8° steppers are 200 steps per revolution
.9° steppers are 400 steps per revolution
The more steps per revolution, the better the resolving precision. So .9° steppers will be much more precise than 1.8° steppers for the same motion. However, it gets a little more interesting when you throw in
microstepping - probably the most misunderstood and appreciated aspect of stepper motors! The motor's field windings are electronically manipulated by the stepper driver (like those built into RAMBo or Duet controllers) to position the shaft between full steps. What people don't understand is, there are two primary tradeoffs when you introduce microstepping:
1) holding torque decreases as number of microsteps increase. This is important because at some point, friction inherent in simply moving the carriages could cause the stepper to miss steps.
2) the resolution of microstep positioning decreases as the number of microsteps increase. This is what is meant by the statement
"Then, as the microstepping divisor number grows, step size repeatability degrades. At large step size reductions it is possible to issue many microstep commands before any motion occurs at all and then the motion can be a "jump" to a new position." in the Wikipedia link above.
So, starting with .9° steppers gives you better precision out of the gate. Then running them at reasonable microsteps (like 16) gives you double the resolution of 1.8° steppers.
The stepper driver itself also factors into this. In general, use a controller that has modern drivers - like the Duet WiFi. You will be amazed at how quiet the steppers are driven by this board.
As you increase pulley diameter, you also decrease resolution. So you want to use the smallest pulley practical. For delta printers in this class, 20T pulleys are the standard. A 20T on a 1.8° (200 steps/resolution) running at 16 microsteps with G2T belts (2mm pitch) gives you an overall positioning resolution of 80 steps/mm. The formula is:
S(rev) * (MS) / pitch * N(teeth) so: (200 * 16) / (2 * 20) = 80 steps/mm
The original Rostock MAX V1s had 16T pulleys that gave 100 steps/mm resolution
But there are tradeoffs with using smaller diameter pulleys - increased belt wear being one of them.
So sticking with G2T belts and 20 tooth pulleys with a .9° stepper (400 steps/revolution) running 16 microsteps you get: 160 steps/mm resolution, double what you get with a standard 1.8° stepper.
