RepRap: Blog

We're starting to approach the point where it's worth considering practical things to make. Using the Blacksmith analogy, the first thing should be a micron scale gripper - essentially a pair of tongs to hold your work with.

I went for something totally minimalist. The prototype here uses a syringe body 6.5mm in diameter as a sleeve, and an actuator shaft of 4mm dia. The jaw spacing is 3mm. Here is the prototype on a 10mm grid in the open and closed configurations:

 

This is an ad hoc prototype so things are a bit scrappy, please excuse. The sleeve acts on the sloping arms at the base of the gripper, and a single flexure pair bends to allow the jaws to meet. By changing the length of the jaws, I can vary the range of motion.

This was printed on a Prusa Mk4. When printed on a RepRapMicron, I would use a blunt 0.5mm hypodermic needle as the sleeve, and a 0.24mm stainless steel wire as the actuator. I'm sure I'll change dimensions, but just a quick comparison. Scaling down by a factor of (6.5/0.5) 13, this would result in a gripper approximately 1.5 mm x 1.2mm which we know the μRepRap is capable of making with an accuracy of better than 50μm worst case (though the precise layer height is TBA).

The jaw spacing would be about 230μm, which is a bit small for the present capabilities IMHO but this is just an example.

Current concept is to mount it on the probe holder with a micro hobby servo to do the actuating so I don't have to put my shaky hands on it. Some support infrastructure is obviously needed but I'm nowhere near designing that yet.

To attach the gripper to the actuator wire, I would give it a light coating of UV resin before putting it in contact with the printed gripper. When it's in the right location, the resin is cured. Probably needs more contact area for the bond, and a bit more room to prevent the resin blob contacting the sleeve. This also gives a handy way to keep hold of the part when I separate it from the print bed.

Blacksmiths don't just use one pair of tongs - they have dozens. Consequently I'd expect there to be a variety of grippers. Apart from anything else I can foresee a number of them pinging off the end of the actuator wire into the universe beyond forever...

The Maus C design evolved organically, which was handy for prototyping and experimentation. To take it to the next level though some of the idiosyncrasies need to be removed and the size of the build area extended a bit.

One that stands out is the long single flexure used to drive the X axis via the stage - a hangover from the original Maus prototype. There is no reason the X flexures can't be moved directly, rather than using that to push the stage around. This would remove the non-linear motion caused by bending the thin drive flexure both in the X and Y directions.

While that's happening, a complementary flexure should be used to drive the Y axis, rather than the single thin drive flexure similar to the X axis one. Again, that would remove non-linearity.

I've already discussed using what I'm calling a pantograph linkage to make the actual driver motion linear. Somebody please tell me what the proper name is.

One thing I've not wrapped my head around yet is whether to keep the probe's Z axis completely disconnected from XY motion. This makes sense when trying to keep everything linear, but at some point I'm going to want to add manipulators, and keeping those in sync with the bed might be problematic.

It might be necessary to have completely separate manipulator mechanisms, which currently would be very clunky. But when μRepRap is capable of printing the manipulator mechanisms themselves, they could easily be small enough to mount on the print bed. We shall see.

As to practical progress, that's all up in the air as I prepare for the FAB25 presentation early next month. Quite looking forward to getting that finished, and returning to multiple layer printing.

Layers, like an ogre (ref. Shrek et al., 2001). As with most experiments, before attempting a multi-layers print I conducted a trial run using a robust, relatively blunt old probe just to find my unknown unknowns.

The test object was an 8x8 hollow square of voxels spaced at 30μm, layer height set to 10μm. The test pattern turned out to be largely irrelevant due to the size of the probe tip, and an ill-defined square was what ended up being printed, and I ran out of resin at the end of the square. No matter, trial run isn't it? The experiment continued.

A single layer object was left as a one layer control, the probe moved over 500μm, and a second object printed in the same manner.

After the first "layer" the probe retreated to the resin reservoir and the UV LED was activated for 12 seconds. The probe was then dipped, and became over-saturated so I wiped it on the reservoir edge. Not ideal, never mind, soldier on.

The probe then returned to the object origin and was manually lowered until a change in the meniscus of the resin on the probe was seen, indicating contact. This occurred at a probe height of 10μm. The test pattern was then recompiled and printed at that height. Set as before with UV.

After recharging the probe a third pass was made. As anticipated, contact was made at 19μm height, and the pattern compiled/printed. In theory, resulting in a 30μm high object. This was then set for 3 minutes under a 4W UV LED lamp 40mm distant.

So, the result. The slide was put in a slide holder, and inclined at a slight angle under the binocular microscope to get a side-on view of the ill-defined blob. Here I have a comparison between the single-layer and three-layer objects:

Glass substrate is on the left side.  Looks like the layer heights determined manually are approximately correct. Doesn't look so nice from the front, so I'm not calling this a "successful" print, but it was a useful experiment.

Takeaways:

Repositioning of the probe after dipping is accurate to within detectable error.

The aluminium foil reservoir adequately protects the resin, and shields the probe so it doesn't set solid.

The UV LED does not take out the microscope camera.

Blunt probes are no good for accurate resin deposition.

Resin layer with a blunt probe depositing >30μm diameter dots roughly 10μm thick.

More care needed to keep the resin reservoir film at an absolute minimum thickness (swab with a cotton bud if needed).

Printing needs to happen centred more over the UV LED in the bed light well to speed curing time.

Even if accurate layering turns out to not be possible, it appears that creating thicker areas can be achieved, allowing for foldable structures to be generated.

Need to print a proper jig to hold slides at an angle for imaging.

 

What's next? The olive harvest. Nature does not wait for the schemes of men.

Had a thought while trying vainly to sleep last night. I've tried using a parallelogram flexure before to make a pantograph, which in theory works but in practice is not sufficiently stable. I realised that if I made essentially an asymmetric pantograph I could at least achieve stable and constrained motion in one axis:

There's probably a proper technical term for this arrangement but I don't know it. When moving the free end (yellow) towards the (blue) pivot, which would be done with a drive screw, the joint indicated by the green arrow also moves linearly towards the pivot over a reduced distance. This mechanism would be relatively easy to constrain and hold firmly.

The result would be a linear actuator that didn't pivot, unlike the current design which basically relies on a rotating lever which effectively gets shorter towards the limits of its range of motion.

Assuming I can get a 10 degree total movement out of the flexures, and want to have an 8mm range of motion with the same 3:1 reduction ratio, my vague maths suggests the shortest linkage needs to be about 23mm long. That would make the long beam 92mm long, which is a bit longer than the current beam but not outrageous. I suspect the flexures would survive more than this but might not be linear at the extremes, so the beam could be shorter.

Thought: Might have to extend the drive screw to 60mm.

I don't intend to pursue this immediately, so feel free to have a go at it yourselves. I still have much to learn from the existing mechanism, which while not perfect is quite good enough for this stage of the project.