Saturday, August 23, 2008

Ping, Boing.

Progress, of some sort.
I wish I could say it works great, but there is still a lot of work to do. The picture shows the assembled roller carrier - or 'nut'. The fitted rollers are the tile cutter discs - two on each axle separated by a 4mm ID x 6 OD x 9 long sintered brass bush. All these parts are standard and no machining is required.
The axle rods are 4mm steel, with notches files in each end.
Compression is provided by standard circlips resting in the grooves.

So what doesn't work? The brass rollers shown are M8 plastic inserts - usually moulded in when injection molding parts. They really grip the filament well and with the thread drilled out and the 4mm to 8mm bushes lightly finger pushed in (also standard parts) they roll nicely on the 4mm axles. Alas, they don't roll nicely on the filament. The knurling pattern isn't random or angled enough, so with the massive circlip pressure they 'bump bump' round. No good.

Next up is the tile cutter build. This is nice - the edges are about 90 degrees and whilst not rounded, they arn't sharp like a knife either. They grip the filament really well at low pressures. Shows promise if I can get the loading right.

Also failed the experimental progress is the circlip system. The un-sprung ID of the clips is 18.5mm. At 18.7mm ID the pressure is too much and the filament and the tile discs send the free end of the filament pinging over my shoulder!

So on to spiral wire compression springs. These also go pinging over my shoulder but mainly due to my eagerness to install and operate the nut without securing them properly. Patience required.

The heavy ones are too heavy, but now I have the range of motion to adapt to varying filament diameter. I have a roll of HDPE (the slipperiest one) which is oval - 3mm max, 2.8mm min.

The light ones ('recovered' from a dismantled inkjet printer paper feed path) are not strong enough. I have been to Dad's to get some alternative ones from his 'spring box'. More tests to follow but whilst no quite as simply elegant as the circlips, I think this will work.

One drawback with the tile cutters is their expense - about £1.50 each from B&Q and 6 are required. Not too bad and being tungsten carbide they will last for ever running on plastic. And that is another problem. Sometimes they make a groove that is just far too deep and all feeding stops. To solve this I have machined a couple of plain aluminium cylinders - 12.7 OD (as bar stock) with 4mm bore and 12.5 long.

These plain rollers roll - even on oval filament, but grip is a problem and if I increase the spring pressure then ovality may become a problem. I will try this, but I also want to try some other ideas -

1. Heatshrink on the rollers. This is easy to fit and repair and may just add the grip I need.
2. Machine (eek, not ideal) the roller OD to include pointy ribs much like the tile cutters, but leaving flat section to prevent the ribs penetrating more than 0.1mm or so into the filament.
3. Using shorter plain rollers and the tile cutter roller in a stack (OD will need careful matching/machining).
4. Putting a short taper section on the roller to aid initial filament feed-in.
5. all of the above together.

I am trying to design this to be reprapped (Plastic housing, nut etc in place of Laser profiles) and hand made from 'standard' parts. I have access to my Dad's workshop - ML7 lathe - nice, Milling machine etc. But I do have to drive 12 miles to get there. Proto and test parts from this source are ok and when (IF) it works then I can think about making it home-makeable later.

And the pencil sharpener? to make the filament end a bit pointy to help feeding of course !


aka47 said...

For your feed rollers have you considered using brass/steel studding. it has a thread already on it to feed with.

If you make sure that rollers made from studding don't roll but turn with the nut assembly (whilst remaining sprung) then the thread will act upon the polymer and drive it forward.

Lampbus said...

aka47 - I had thought of somthing similar to that. The problem is you are still dragging the threads against the filament. This forces it to twist - limited by the Darwin extruder by putting it in a channel and driving on one side only.

If the nut were to rotate as you describe, it would be even harder to prevent the filament twisting.

The reason I am rolling my rollers in a balanced pressure arrangement is to remove as much of this dragging as possible. This leads to (hopefully) much more consistant and reliable feed. Should be quieter and use less power too. Less ware on components.

Remember this part runs almost continuously and can rack up the hours quite rapidly.

aka47 said...

Must admit I am struggling to see (probably my problem rather than your oversight) how if the motion of the rollers isn't either locked or synchronized to the nut (but not the filament) how it will actually drive forward rather then just roll round the filament.

If the rollers are geared to the nut then the nut assembly should be able to drive forward maybe a bit like a gripping planetary gear box. But only because of a mismatch in diameter of the filament versus the amount of turn of the rollers.

If you can get hold of one have a quick look inside an old fashioned school type pencil sharpener (The one with the windy handle). But even these need the grippy thing to stop the pencil from turning whilst it is sharpened.

Lampbus said...

Yes, it needs the filament to be restrained from twist - but I am expecting the reel & heater loads to be more then enough.

The rollers roll without needing to be geared - think of one roller. If it is alongside the filament with its axis aligned and you swing it's axle about the filament centre, it will go round, rolling, and after 360 degrees it will be back where it started.

If however, you tilt it slightly, it will move along the filament's axis and after its axle has swung 360 degrees, it will be displaced along the filament. (or filament displaced...)

The only twisting force on the filament is from unevenness in the section and rolling friction from the roller bearings. Oh yes, my latest rollers have weeny ribs round them which cut into the filament (like the tile cutters but controlled) and this will add to the twist a bit.

If you take a threaded rod and feed it through threaded rollers (on the same axis), it will displace the rod very nicely. I tried this first with the thought that the threadded rollers would press enough of a thread into the filament. It does, but the result is an oppersite thread (roll a length of studding across the table - it travels at 90 degrees to the roll axis - no axial displacement.
Try the same on plastercene - it makes nice thread-angled groves, but also dosn't traverse axially.

It will however if you roll it along existing groves of any angle other than its thread angle.