Sunday, September 03, 2006


One way out of the bramble patch...

Okay, here goes. You all saw the last blog entry that I did. We have a real problem with the shaft encoded gear motor control theory once you talk about trying to use them in an xy positioning system.

Having that disaster may just be a blessing in disguise... or maybe drag. :-p

I've been looking at the conversion that NASA to turn Shuttle technology into the Aries I and V launch vehicles for the past week and a few days ago it hit me that that the concept was brilliant, a view that not many people agree with. I intend to write an article for Agora Vox in a few days that presents my thesis.

The point of the NASA development is that you don't have to develop any new technology to speak of to get back to the moon a lot more efficiently and safely than we did the last time.

Back to us though. Of course, we've had a number of real problems with positioning systems in general for some time, but then we've been politely ignoring them and hoping that they will go away. Foremost amongst the problems was that we like screw drives because they're accurate. The problem is that unless you make them very expensive they are also very slow.

I think I have a way of killing both birds with the same stone. Let's take Reiyuki's basic design as a starting point.

It's simple and it's easy to fabricate.

Now, put a turntable on the horizontal platform and you're good to go.

Vik demonstrated that you can make workable large gears last November.

Since then I've written scripts that let us design involute profile gears.

If we were to create a 125 mm radius gear for the turntable and drive it off of a gearmotor with a 25 mm radius gear we'd have a 5:1 ratio. With our cheap AS5035 shaft encoder one pulse would mean an angular translation of a shade under 17 minutes arc resolution. With the AS5040 we would get just over 1 minute of arc translation per pulse.

Now, here's how it works. The y-axis that moves the green horizontal work surface can run at pretty much any low speed. So can the z-axis that raises and lowers the extruder(s).

The x-axis needs to run fast if we can manage it, but it's not a deal breaker if it doesn't go too fast.

In operation the extrusion stroke will only use the x-axis motor. The z-axis and turntable align the worksurface and can take their sweet time doing so. Then, the x-axis carries the extruder across the work surface and lays down an extrusion thread.

If we are doing a perimeter it runs as long as there is a straight line segment then stops picks up a fraction of a mm and waits for the z-axis and turntable to align it for a new stroke in a new effective direction. With that approach we can create a true vector loop for the outside perimeter of an object, something that we couldn't do with a straight cartesian machine.

The best news comes with infills, however. Since infills are a series of parallel extrusion traces all the z-axis and turntable have to do is get the alignment right for the first infill stroke. The x-axis then lays down the polymer thread and all that is required for the next thread stroke is for the z-axis to advance one step.

Nothing to it.

One of my cheap little 16F628/754410 control boards could handle all four motors. They don't even have to be synchronized the only time dependent thing that is needed is for the x-axis controller board to know that the turntable and z-axis boards are finished doing their thing before it starts. The relatively slow serial comms loop that Simon put together will be more than enough to handle that and cook breakfast, too.

We really don't even need to do PWM motor control on the motors, either. It's probably better if we just feed them straight 5v or 12v power since we don't have anything to coordinate that would make speed control anything more than an additional, unnecessary complexification.

There's also nothing to stop somebody from using this approach with steppers, either.

If we can't hit 0.1 mm accuracy with this rig we need to hang it up and go to work at the neighborhood supermarket sacking groceries. :-)

Seems like this would work. It does add yet another level to the 3 axis system (It becomes a 4 axis system - 3 linear axis stages, and one rotational axis stage.)

All this just so you can use PWM to control motor speed? It seems like a lot of work, especially now that with this kind of set up, you don't need PWM at all: Just run the X-Axis at full speed or full brake, with nothing in between.

That is, unless I've misunderstood your idea.
All this just so you can use PWM to control motor speed?

Not at all. I specifically said that...

We really don't even need to do PWM motor control on the motors, either.

PWM would be a good idea for the x-axis that does the extrusion stroke, but not for the others. Even then it would only be to keep the speed steady so that the extrusion would be even. There are other ways to do speed control besides PWM, though.
Ah. Okay. I missed that part. :)

It still seems to me like a software solution might be better than a hardware solution -- I think I'm going to continue to work on this with the spline driven solution.

You're approach does have some interesting concepts, specifically, it has some of the characteristics of the dual disk design I proposed last month -- for example, it might be easier to make gears with it by driving more than one horizontal platform motor at the same time.
I think I'm going to continue to work on this with the spline driven solution.

By all means. The more ideas we've tried out the more tools in our toolbox of techniques to handle problems as they appear. has some of the haracteristics of the dual disk design I proposed last month

Yup, it sure does.
Post a Comment

Links to this post:

Create a Link

<< Home

This page is powered by Blogger. Isn't yours?

Subscribe to
Posts [Atom]