XY stage

XY movement

For the Medjed system, the XY axes is handled by a pair of crossed Snapmaker 168mm travel linear axes. These use the TMC2209 stepper driver and feature standard 1.8deg steppers (to my knowledge). The axes are lead screws, not ball screws and feature rather large leads of 20mm (XY).

As such these axes are not really suitable for high precision motion. Their repeatability is well below 5um (Medjed's target pixel size). In their intended application (3D printing and comparatively high force CNC milling) this is not an issue.

For the use as a MLA system, we can deal with the low mechanical precision of these axes with methods outlined in the openMLA wiki. The axes have the benefit of forming a very rigid system with plenty of mounting points for accessories. In addition, the proprietary cables and connectors (while somewhat annoying to source) provide a nice interface to connect to custom boards. Footprints for the connector headers can be found in the KiCAD PCB files.

Encoders

Encoders are key to making the low precision Snapmaker axes produce acceptable results for lithography. Given that Medjed system targets 'only' 5um resolution we should be able to make of magnetic encoders with high interpolation factors. Going with the assumption that we need an encoder resolution 10X better than our step size we would need an encoder step size of ~500nm.

The selected encoders for the system are the RLC2IC by RLS, which is low cost (just under 100 euro for a single axis, excluding tape) and has a universal A/B/Z quadrature output. Documentation is pretty good and CAD drawings are available and the product will remain in production for at least 5 more years. Tapes with markers at specific locations can be purchased, thereby for allowing reliable homing with external (and less accurate) endpoint switches.

Keep in mind that since these encoders are magnetic, they will not be very happy in applications with strong magnetic fields such as inductive linear stages (but that would typically require optical/capacitive encoders for precision anyway).

For the Medjed system we want the highest encoder interpolation factor of 2^13=8192, which yields a step size of (2mm / 2^13) = ~244nm. That is about 1/20 of the desired pixel size so that is plenty.

The 2nd core parameter that is to be chosen is the minimum edge separation. Higher velocities will require higher A/B/N signal sampling to ensure no (interpolated) steps are missed. For the Medjed system we do not need extremely high velocities, but we do have a higher interpolation factor, which means that even with low/moderate speeds we will already need fairly fast sampling. Finally we must also ensure the signal speed is compatible with the TMC424 encoder input chip.

Let's run some numbers to motivate the chosen configuration.

The Medjed system will need to do 'fast' movements when moving from marker-to-marker positions, and (possibly) when moving to the next write line. Let's look at what the maximum velocity is at edge separation of 2MHz. It takes two edges to transition from one marker position to the next, so we could pass at most 1 million markers every second. At ~244nm marker distance (interpolated) that means we have a speed limit of 1E6*244E-6 = 244mm/s. Given that Medjed targets 4-inch wafers (100mm) it should be sufficient.

❓ At the moment I do not know what a realistic 'peak velocity' between steps of each exposure field is. I think it should be possible to keep it below 244mm/s.

Finally, the TMC424 requires (datasheet table 3) that the edge separation is at least two TMC424 clock cycles, and that the count rate is not faster than the TMC424 clock. The TMC424 will be run at a frequency just above 16MHz. As long as we stay below 8MHz edge separation we should be okay. Faster would be possible, provided we run the TMC424 at higher frequencies (up to 60MHz).

📃 Just for fun, if we think about a higher resolution MLA system, one may use optical encoders with an interpolated step size of 5nm. We would then (keeping everything else the same) have a 100MHz encoder signal. Too fast the the TMC424.

Altogether that yields the RLC2IC with specific configuration 🔗.

The Encoders come with a Molex 52745-1197 connector, which mates with a 11-'circuit' Premo-Flex FFC Cable 15020, which can be found at most suppliers. They are fairly pricey but I think it's worth it. It is a simple connector with cross section 6.1x0.43mm.

We need to order 3 different cables for the Medjed system.

X-axis: travel range is a little less due to geometrical constraints of Z-axis, so there is only ~155mm of travel on that axis. So our FPC needs to be at least 155/2 = 77mm. We need a bit more to account for the Z-distance of connector to the encoder and general flexibility limitations. A 102cm of 76cm cable might suffice.

Y-axis: This is the most complicated as it needs to account for both the X-axis movement and that of the Y axis. We essentially create a nearly identical fitting to the X-axis (requiring ~77mm of FPC), do a 90 deg cable direction change and then run another ~100mm of FPC for the Y-axis part. So this axis would need around 180mm FPC. Maybe closer to 200mm with the turn.

Z-axis: This axis is not as complicated as the Y-axis, but it is a long connection. We need to run the full travel range of the axis (168mm), base of the linear axis (another 75mm or so) and then towards the PCB (another ~75-100mm). So we are looking at about 318. Largest I can find is 302cm so that will have to do.