Thursday, December 29, 2005
a short motion sequence with head on
I'm not happy with the way Knewt "leans" when shifting weight. This is one of the issues I set out to solve when Knewt was started. The change in 2.0 to add more weight shifting ability, has re-introduced this leaning. It's not so bad really, but I would prefer that the head was kept perfectly level. I'm going to have to look into this more closely.
Tuesday, December 13, 2005
two heads are better than one
This photo shows the finished metallic head next to the un-finished white head. It was a couple of days work to get the head to the metallic state. I'll outline the steps here.
The STL process leaves a groved surface in random patterns as a result of the building process. I consulted Scott Washington who won awards at the IPMS/USA 2005 nationals modelling competition. Scott reccomended filling the gaps with Squadron white putty mixed 1:1 with Testors liquid glue. This is a pretty toxic smelling brew so be warned. The glue thins the putty to make it more spreadable. The pitfalls to avoid in using this on an STL model is to make sure you cover all the gaps. Since the entire surface is textured, it's easy to miss a spot. I put the putty on with a brush wich leaves brush strokes, creating a texture of it's own. I will try using a popsicle type mixing stick to apply the putty in the future to see if that works any better.
This putty dries pretty quickly. I began sanding after about 3 hours using successively finer grits of paper. 150, 220, 320, 600. This is the step when you find the spots you've missed. Stop here and fix these spots, then re-sand. The paint won't fill the gaps.
I sprayed the metallic paint on avoiding drips and runs. The next day when that was dry, I gave it a coat of Future Floor Polish. This makes the metallic finish have a slightly lighter color and smooth's the surface like glass, increasing reflectivity.
Notice the logos on the sides. These are custom decals made with Testors custom decal kit. The kit contains special paper, a spray fixative, and software. I wasn't able to get the software to even load, so I used photoshop to compose and print. The decal material is transparent. This fact slipped my mind and I was suprised to see that white areas on the logos were now silver when I applied them. This makes a cool effect though on the large red background area of the IronCAD logo, as the red is translucent and the silver highlight shows through. Another thing to be aware of is that the decal material is flexible. This is a big advantage to covering curved surfaces without wrinkles. But beware of sliding the decal around as the water dries. I pulled on a corner and stretched the decal material, slightly distorting the image. Be sure the decal is where you want it before drying away the water with a towel.
After the decals were dry I added another coat of Future floor polish.
I'm dissatisfied with the silver paint and am looking for a high quality replacement.
Friday, December 09, 2005
I got the STL parts for Knewt's head this week. As luck would have it, two of them on the same day. Stuart Rolf prepared the STL file and Ken Apple ran the job in Dekalb Tech's CAD lab. Barry Smith of DrafTech provided a head and a great poster that's in the next photo. Thanks a million everybody. (this is a perfect opportunity to use the colloquial "y'all", and even "all y'all" in this particular case").
Barry Smith gave a bottle of "Ambroid Pro Weld" to repair and strengthen the parts. These particular ABS plastic STL models are held togeather by a slight fusing of a .013 bead of melted plastic. These small threads can be torn away from the part if you try, or if it takes a bad fall. Considering how much Knewt falls, that's bad. But by saturating the part with the plastic welder, the fibers are much more strongly fused togeather and become almost as strong as a single piece extrusion. I coated both heads with it inside and out and didn't even use 1/4oz. It's like water and doesn't smell "too" bad.
After brushing the part with this stuff, it becomes shiny. A new advisor to the project, Scott Washington, recommended a filler paste and sanding to cover the texture left by the machine. I tried this and it worked quite well. Lots of had sanding is required. The grits I used were, 150, 220, 320, 600. The errors I made were to leave small gaps in the sanding putty. These were most evident on the vent slots on the top of the head. So I sanded and painted it again. I think by the third time I paint this head, I'll be up to amature status.
It was wierd seeing the head on for the first time. All the screw holes and bosses worked just fine. It's held togeather with #6 selft tapping screws from Austin Electronics. All the holes lined up ok. I had to cut an inconspicuos part off one piece to allow it to fit. There are several improvements I would make on the next pass. The're are ribs in the lower part that didn't print in both cases. But the intresting thing is that they weren't the same ribs in both cases. I reinforced those areas with plastic plates after ward.
Removing the part from it's building plate and extracting it from it's scaffold material is a pretty big deal. I've heard that the DrafTech office has a water soluable scaffold material and you throw it in a bath to seperate the two. Stuart said that part of his head broke just on scraping it off the plate, and again more cracks from trying to remove all the scaffold material. There was still a little of this light gray scaffold material left when I got it home and It was indeed difficult to remove, especially in small crevices and inside screw holes.
The eye holes are sized to 3mm LED's and I have mini O-rings on the protruding green lites. I inted to leave this head translucent. I like the fact that light passes through it and will try to incorporate a mouth "indication" behind the face without cutting pass throughs. You can see the base of the silver head in the forground.
Here you can see (or almost see, anyone want to buy a camera?) logo's added to the white head and circuit boards mounted inside the silver head. We've decided to build the second body up as version 2.0 also and work on motion routines in tandem.
I chose silver for the painted head to emulate the finnish on the aluminum. White and black and clear are the most common plastic colors on the body. There is a rainbow of colored plastic wires, other wise the silver on the aluminum is the most prevelant color. What color would you choose for the head?
Does anyone have a suggestion for how I might tint the white head from the inside surface. It would need to be a translucent dye or the like. Perhaps a blue dye added to the plastic welder solution would do the job. I bet art markers would do it. Does anyone have suggestions for a pattern on the head?
Actually this is a perfect time for a critique. I'll get some better photos up soon.
Wednesday, December 07, 2005
Knewt at Draftech
I had a good time Friday talking about robotics and CAD. There were about 20 - 25 people who came over to my table and had a good look at Knewt. 90% of those had some comments to make like "very cool" or "pretty neat". About half inspected the mechanism closely and offered a question about it's design. It seemed to break the ice for even the most quiet attendees. I met some old workmates from Nordson coincidentally. We reminisced.
The Lithium batteries died halfway through, but I was able to hook the power supply back up pretty quickly to finish. I really like the poster alot. I'll take it along on my next presentation and show off Draftech along with Knewt. I wish I had gotten a photo when people were milling around but I was busy dealing with the bot. Scott Walker, you would have been the perfect candidate for that task. What was I thinking?
The event was the 4th annual SolidWorld user conference held by DrafTech Systems Inc. in Lawrenceville. The white cabinet to the right of the table is the Dimension STL printer that produced one of Knewt's heads. The printer lays down a .013" bead of ABS plastic, building the prototype part layer by layer. Thanks a million to Barry Smith for the invitaion and the wonderful posters and especially the STL parts.
DRAFTECH Systems 1730 Spectrum Drive Lawrenceville GA 30043
USAPhone: +1-770-963-8856 Fax: +1-770-963-7056
Monday, December 05, 2005
testing ramping on ServoMojo
After having used the Parallax controller for a while I'm beginning to miss the gentle ramping provided by the BDmicro ServoMojo. Bob wired the "position pot" on a standard servo to the scope, so that we could view the position feedback directly.
This screen shot shows a normal servo command that is a straight line from starting to ending point. This translates into a servo motion that goes from full stop to full speed at the beginning of the motion, and vise versa at the end of the motion. The "speed" that the servo moves is choosable with the Parallax board (rate), but the transition from stop to moving is still one step. The abrupt changes in speed cause rocking and jerky motion for a bot like Knewt. It's an issue we considered trying to solve with MCU management of the "rate" variable every 20ms. But why do that when it's built into the ServoMojo?
In this screen shot, you can see the curves at the ends of the motion. The servo transitions from stop to full speed in a gradual fashion and slows to a stop in the same manner. This is shown by the gentle "S" curve shape on Bob's scope. Some of the original videos of Knewt showed this fluid motion. We're driving the servo from the ServoMojo using a ramp setting of 15.