Wednesday, December 22, 2004

New Head Design for Knewt #2

Stuart's head design
Keith Rowell Design

Stuart sent this head design last week. It was designed in SolidWorks and exported in the SAT format first. This format didn't work for import into IronCAD so we tried it next as a Parasolid which IronCAD liked just fine.

Stuart's abilities with CAM as well as the design software shows the kind of skill and talent he possesses. Expect to see these two versions and maybe more, chicken walking around all over the place come this summer.

Stronger Servos for the Foot Motor

42oz/in foot motor 01
Keith Rowell Design

There was some concern from the beginning as to whether the feather weight servos would suffice for the foot motor, because it's the only motor that experiences the full weight of the robot during walking. The HighTek HS-55 feather servos are rated at 18 0z/in at 6V. Over powering might take it up to a higher torque, just before the magic smoke escapes.

Phillip FitzSimmons did a rough estimate of the forces needed to hold the weight of the bot on one motor and surmised that 15-18 oz/in wouldn't be enough. Not being able to find any servos in the same size format that had more torque, I ordered the next larger size. An even stronger one is available in the same size format: And one with metal gears:

48pitch gears are available for this servo spline shaft too. I ordered a 18tooth pinion and 66tooth spur that will replace the ring and pinion used on the smaller motor.
An 18 and a 66.
Metal ones are available

The smaller motor with it's smaller spine shaft required a tricky manuver to make the pinion fit. And it requires a pillow block to hold the pinion in place. The larger motor is much easier to fit because of these issues. I would still like to be pleseantly suprised to see the smaller motors carry the day. But I'm planning their replacement just in case they don't.

I'm also looking for travel cases for the knewt. It's overall size is 7x9x12". The pelican water tight case is ideal with the foam inserts but is expensive at over $100. If anyone has access to travel cases for lot less, please let me know.

42oz/in foot motort 02
Keith Rowell Design

42oz/in foot motor 03
Keith Rowell Design

Wednesday, December 15, 2004

Head and Shoulder Plates Finnished

shoulder plate and head plate with head turn motor atached B
Keith Rowell Design

Stuart has finnished the first shoulder plate, a very complex part to make as we mentioned last week. All of the weight of the bot is carried through the big bearing, and then to the legs, through this plate. It's un believeably light.

The head plate holds the head onto the neck using the same large size bearing. The servo is responsible for the head turning movement via a mechanical connection with the servo horn, through the center of the bearing.

I don't anticipate getting a new camera for Christmas, so WYSIWYG.

shoulder plate and head plate with motor attached A
Keith Rowell Design

In this orientation, we are seeing the top of the "right" shoulder plate. And the top of the head plate/mtr. Stuart did a fantastic job on these parts and we hope to get a movie of the cutting process this week. It's very exciting to see these parts in the "physical" world.

Mojo Mods

testing mojo board with leg
Keith Rowell Design

We've recieved the second ServoMojo board and are modifying our set up routine. Until now we were setting all the motor presets with a terminal program directly to the Mojo, and running a pic grogram to control the Mojo afterward. Bob's plan now is to do all the communication with the Mojo through the pic using a "passthru" scheme for testing, and adding a "setup routine" in the pic code to initialize the Mojo on start up.

The Mojo has the set-up saving features that store servo acceleration, speed, trim, etc. But reading, checking, resetting, these from the controller is a step above the manual procedures we've used so far. Brian Dean of BDMicro has been very helpfull as we learn the in's and out's of the Mojo board. Thanks Brian again for your assistance.

Thursday, December 02, 2004

shoulder plate physical
Keith Rowell Design

shoulder plate model
Keith Rowell Design

Shoulder Plate Progress

shoulder plate progress

Stuart has begun the really complex parts and I was lucky enough to get photos this week. This shoulder blade part is about 2/3 complete. It started out almost 3/4" thick. The steps used to make this part were 3 setups so far. Stuart set the CNC up for a cut from the "end", using that surface to set up the tools. Then he set up again for one side and cut away the bulk. Then, set up on the other side and cut the holes. It's a very big deal to figure out how you're going to hold the part for cutting. And the surface that you use to touch off the tools for setting up the CNC must be perpendicular so that the various setups will be "synchronized" or the part will have misalignments in assembly. It's quite a complex part. These photos are a disappointment but you can see which part it is by comparing to the model.

shoulder bracket
Keith Rowell Design

shoulder bracket
Keith Rowell Design

First Leg Assembly

leg assembly
Keith Rowell Design

This leg assembly has been togeather since before the holiday but I only got a photo today. We've had the leg and foot parts for some time, it was the ankle hinge parts that completed the assembly to this impressive point. Tape is holding it in a standing position.

Ankle Hinges

Ankle Hinges
Keith Rowell Design

Stuart made these the week before Thanksgiving. They're very nice and use welding rods for hinge pins. See them in use in the leg assembly above.

Friday, November 12, 2004

Assembly Task List


While the last of the machined parts are being made, I've put together a list of the assembly tasks required. These are tasks that require some minor machine tasks like taping holes to receive screws and cutting standoffs to size. This is the list as it stands now. The numbered tasks correspond to the leaders in the attached drawings.

Mechanical assembly task list

1. Cut ankle ring gear x2
2. Drill & C/sink ring gear x6
3. Tap bottom ankle link for ring gear x6
4. Drill and ream ankle hinges for hinge pin x6 ---11/23/2004
5. Turn bottom outrigger post for ball end x2
6. Cut, Drill, Tap, top outrigger post for ball end x2
7. Make bearing washers from Teflon / nylon type mtl. x22
8. Turn shoulder balance posts for ball end x2
9. Turn hip spreader posts for ball end x2
10. Drill, ream for alignment pins x6
11. Cut neck standoffs x5
12. Cut neck motor standoff x1
13. Tap holes x22
14. Make neck motor / bearing connector
15. Make shoulder bearing posts x2
16. Drill, Tap, Mount leg motor ball end. x2

Machined parts not yet finnished are:

1. hip bracket
2. shoulder plate and mating tail plates
3. leg motor mounts
4. head
5. assembly parts mentioned above

Iso bottom
Keith Rowell Design

Iso top
Keith Rowell Design

Keith Rowell Design

Keith Rowell Design

Monday, November 01, 2004

Knewt in IronCAD Gallery

Looks really good guys..... Keith, you really do have too much time
on your hands.


IronCAD posted the Knewt Model on their site this week. Thanks to Kevin Devol and IronCAD for the software upgrade.

Brian Dean of BDMicro also responded this week with a very through explanation of the "extended mode" feature of ServoMojo for getting the full 180deg rotation from the foot motors. The settings are made by setting a minimum and maximum position address for each servo. We'll be testing this feature this week.

Philip FitzSimons (who did a great demo of his 5Amp, 30Volt, servo controller at the AHRC October meeting this week "FADC-01") will be helping us look into making a larger heavier, more powerfull, version of Knewt in the near future. (at which time "Knewt" wont be an appropriate name, perhaps "Knossos") How many lbs. of lumbering metal do you think 250watts will carry?

The ankle hinge is under way. I bought polished steel pins some time back for the purpose. But when I got them out this week, they were pitted with rust. The container I put them in must have had something corrosive in it at some point in the past. I have alot of .092 music wire, and we're going to use that as the ankle hinge pin.

View a movie of the lego test jig in action here.

Friday, October 22, 2004

Successful PIC / Servo Tests

Lego test jig for testing leg motors with Servo Mojo (in the foreground)

Our tests with the PIC and the servo Jig were, after a short detour with the wrong baud rate, successful today. ASCII "motion files" were uploaded to the PIC from the PC using RealTerm. And the program was successfully played back from the pic to the ServoMojo with correct servo motion on the Lego Jig. The Lego test fixture is just a crude mock-up that we use for testing... Next week we'll take a movie of the Lego Jig in action, and do a demonstration for the AHRC monthly meeting. Thanks to Bob for the photos... movie here (1.37m).

Stuart serves up more parts

More aluminum parts this week

Keith Rowell Design

Stuart joined the circuits and programming team this week with parts in tow. There are now only 7 unique parts left before complete assembly. Unfortunately, these are the most difficult and "intresting" parts, and hence take longer to make. We projected complete mechanical assembly by January / February. Next week might see the ankle hinge added.

The First Step of Real Assembly

Foot Plate and Bracket Properly Assembled

This image shows the footplate and bracket assembled. This is the first two parts to be more than "fitted" togeather but actually "fastened" with screws. The cutting of the 2D parts being the more simple process, they were rather quickly produced. The final drilling and tapping and
3D parts comming up will take more time. I intend to join the machine class in January to finnish up the assembly. We have parts for at least two complete bots, and assorted parts for a third.

Parts Arrangement
Keith Rowell Design

Tuesday, September 21, 2004

Laser Cut Parts Arrive

Outline files

The activities this week are to test the variables saving features of the ServoMojo board. This includes trim or centering, home position, acceleraion and speed, etc. Also, I'm putting togeather a spreadsheet to store servo locations for all the general motions required. This includes stand, sit, turn R / L, walk forward / back / arc, and transitions between all of the above.

All of these should be easily settable using either narrow or wide mode. Have you settled on which mode you will be using, or perhaps are you using a combination of the two modes?
One way to move to the "home" position is to generate a reset, which will cause the saved positions to be reloaded. Note, though, that the servos will snap back to their home positions as fast as they can at that time. This is due to not being able to know the actual servo positions at reset time. I could easily implement a "Home" command, though, that moves all the servos to their saved positions, without having to perform a reset, which also implies about a 3 second delay as the stored configuration is loaded from the on-board EEPROM. That command would allow the servos to be positioned and also honor the current velocity and acceleration settings, and eliminate the startup delay following a reset. The firmware you have does not support that - but it should be easy to add. I can send you a new chip that you can replace on the board to have that feature. The processor is in a socket to the replacement should be straight-forward.

There's a sample of how the Japanese guys do it here...

Unfortunately, I cannot find a description of the RCB-1's communication protocol in either the software nor the hardware documentation. Without that, I cannot say how easy or difficult it
would be to match it. However, their PC side software is similar to what I envision for ServoMojo. But I don't expect to have that piece in place for some time. I do think it is an important component, and one that I plan to have in place before I present the ServoMojo board for general availability. One hold-up for me has been the lack of a PC virtual machine that runs on my G5 Mac in order to do the development - the current version of Virtual PC supports up to the G4 processor, but not the G5. Just starting recently, however, Microsoft has begun taking pre-orders for the latest version of Virtual PC which finally supports the G5, so at least there is light at the end of the tunnel - but they still don't say exactly when it will be available.

I like the name of the software, "Heart to Heart". I don't know the data output format, but I suspect it is'nt compatible with the ServoMojo. Do you know if this is the case Brian? Is conversion worthwhile?

I ordered parts from Pololu in polycarbonate as a test of the material and the service. The Pololu laser cutting service is very inexpensive and their operator Jan Malasek is very helpfull and conscientious.

Not so much can be said about the aluminum watter cutting service DC waterjet. These guys didn't respond to emails, and after bugging them a few times they finally came back with estimates.

Pololu can't cut metal nor mat'l thicker than 3/16". So I only sent them the parts that were 2D and 1/8" thick. Their charge for these parts was $28, and that includes the polycarbonate material, checking the file for errors and getting back to me on fixing them...I really like using them. I've not yet ordered the parts for the 1/4" material. Stuart is considering making them locally so we'll have one all aluminum bot and another plastic/aluminum hybrid.
The same parts cut in aluminum from DC water jet is $90. The 1/4" parts are $114. (not including material) was $118. All the 1/4" parts require extra machining after cutting out the outline.

The polycarbonate is considered bullet proof by the battlebot guys, but not suprisingly, when you cut it this thin, it's flexible compared to metal. Dale suggested that this could be an advantage, absorb shock, etc. Re-Ordering in 3/16" mat'l is an option. Stuart's aluminum version will kick this one's butt.


The following is the central body made up of the back plate, battery retainer, balance servo, hip/turn servo, shoulder bearings w/mounting flanges, the neck standoffs, and the batteries



The batteries and motors fit nice and compact. Dale made the 3 bearing mounts in aluminum, two shoulder and one for the neck. That's the two shinny parts in the top view. Thanks Dale.

Stuart has redesigned the connector for the pinion gear on the foot, basically turning the gear around, mounting the tooth end next to the servo. This allows room in the material to put a std servo horn spline into the center of the pinion, and the pillow block on the shaft end. The image shows a recess cut into the foot to accomodate this change in the location of the servo.

Keith Rowell Design

Wednesday, September 08, 2004

Center of Gravity

Should it be higher for smooth walking motion? Or lower for stability?

A lower CG will tend to be more stable, but a higher CG will, in my opinion, give a nicer looking walk (less waddle). One of the projects I've been working on is a very small muscle wire powered biped. I was originally thinking of making a smaller version of MiniMechadon with a pair of AAA batteries mounted horizontally. I have since decided to mount the AAA's vertically and make it a humanoid (about 3" tall) to improve the walking characteristics.

That being said, if you're designing Knewt to compete in RoboOne-type competitions it would probably be better to have a lower CG to gain better inherent stability at the sacrifice of walking "style" and efficiency.



Friday, September 03, 2004

Laser and WaterJet Parts Cutting

Laser and Water Jet cutting estimates

I solicited cutting estimates from 3 sources.


Pololu can't cut all the parts. Their maximum thickness is 3/16" and they can't cut aluminum. Their cost for doing the 1/8" parts in polycarbonate was $27. I ordered those, they were more flexible than I expected. Not the ideal stiffness. Dale mentioned that this could be an advantage...they're certainly light.

DCwaterjet took several days to get their estimate to me after a couple of reminders, their cost was very close to eMachineshop at around $200 - $300 using waterjet on aluminum.

Thursday, September 02, 2004

Project updates
The new project name is "Knewt" (knowledge in a small package) This week we explored a problem with the gears intended for the foot motors, and decided they were'nt a problem after all.
There is a 5%-7% descrepancy in the tooth pitch of the bastard gears from the surplus bins. But further testing and measurement revealed that a little wear on the aluminum gear, and the errant 7% is worn away. I'm worried about backlash, but as the foot only sees resistance in one direction, there isn't much reason to worry about it. A spring may be added at some point if play becomes a nieucance.

My method for measuring the fine dimensions on these gears was to scan them at a high resolution directly on my scanner, then import the image into CAD, in this case intellicad. (oddly enough, the photoshop PSD format was the only one that made the image visible in the workspace, jpg's and bmp's came in as icons...something I didn't expect) I then used 3 point circles to outline the id and od tooth dimensions, splitting the difference between the od's of the two gears to get the pitch. I then measured the pitch of one tooth on each gear along the pitch line, and found a 7% discrepancy. Even though these are faily high resolution images, there must be some error which I estimate to be less than 3%. So to change the diameter of the pitch circle and remove the discrepancy, I mounted the steel pinion on a drill and ran it against the aluminum ring gear for a few minutes and saw marked improvement in the smoothness of their mesh. I'm hopefull that with a proper jig and some lapping compound, the gears will mesh very smoothly....the clock builders who make alot of gears said that backing off the clearance between the gears will work.
Alternately, we'll go back to linkages...


The Finer Points of Surplus Parts

7% difference in the pitch of the surplus gears I've scrounged up for the build

Keith Rowell Design

We can always fix this gear problem with other answers if they fail to please.
I've deduced from the high res scans that there's a 7% difference in the pitch. I wondered if just "running them in" with a drill motor and a jig to apply preassure, or such, would "wear" them down to fit enough to make up that meager 7%. Ultimately either the gaps in the pinion teeth are too narrow or the teeth on the ring gear are too wide, or both. I believe that 7 percent will be an easy modification to reach, because even the slightest wear moves the pitch line alot. Maybe running them on a jig with lapping compound in their teeth would match them up...(I must sound like a real hill billy!)

what do you think?

Is this a regular discussion in hill billy circles?
Don’t worry about the gears I think we can use them as they are.
See you later


The percentage of difference you're test shows suggests you have one metric and one fractional gear. Metric is measured in "Modules" and fractional is measured in "pitch".


Monday, August 30, 2004

First Part

The first part is made

Stuart made the first of many parts this week. It's so great to see it join us in the physical world.

Some Feedback on the New Look

A tribute to Brian Dean's Servo Mojo Eval Board
Keith Rowell Design

Excellent! Careful ... that's tempting to put up on my web page :-)
BTW, I like your new "head" design a lot better than the original.
Looks clean and modern. You ought to turn this into a product once
you get it fab'd. A kit would be great. I know I'd buy one as long
as it was affordable.
Great work!

Hey Keith!

Looks like you are making great progress! Are you machining the parts yourself or are you using a fab shop? I have been thinking about getting a table top mill to do things like that, but it's hard to justify the cost. The lateral joints at the knees are interesting. Are there also lateral joints at the hips? About the batteries in the head. You are correct - the high center of gravity requires less shift left or right to balance the robot over a foot. I haven't done any formal calculations, but I think that the higher center of gravity would also improve the efficiency of the walk. A biped with a low center of gravity will need to "waddle" back and forth more to shift the CG over a foot. The overall result is that the body of the robot travels in a zig-zag as it walks forward. With a higher CG, the robot waddles less and the body travels in more of a straight line. The distance between the hips will have a similar effect - the farther apart the legs are for a given height, the more the robot will have to waddle to shift the CG over a foot.

Please keep me informed of your progress!



Thursday, August 19, 2004

A New Look

Hey, I like the new 'head' design. Sort of looks like a mechanical fist walking around to find something to punch! Bill Hotch

This shows the current new head design, and what it would look like if the bot were anodized in gold or yellow.

Keith Rowell Design

In the intrest of avoiding any conflicts with the Lucas Films guys in the future, I've been advised to avoid derivitive design based on the Star Wars AT-ST (though I love that design so much).
So all future work will be original in style, (all comments and critiques welcome). Hence the funky head shown in this week's update. Please respond with your first impressions. I will incorporate all ideas possible. Don't get stuck on this first one, The sky's the limit until it comes time put in the electronics and start testing. Lets hear your ideas...

The project has had many leaps forward this week.

First of all, IronCAD has chosen to sponsor the bot. So we'll be sporting the IronCad logo, on the bot it'self at all events...

The first parts have been made and they fit perfectly! I place the bearings into the lower ankle rod myself today, and they fit like a gove. You can spin this thing on the bearing and it runs forever, zero friction! I like it. Thanks to Stuart Rolf for his formidible experience, ability and generosity to provide the machine work. I can see 2:1, 4:1, ad infinitim scale versions of this in the future. Watch out world!

Draftech is expected to print out a copy of the new head shape on their 3D STL machine tomorrow, in abs plastic. Asuming I can get the file ready.

Another area I would like ideas on is weapons. I plan to enter it in the first (American) Robo-One competition, which will probably happen next year. And a major requirement for competetiveness is the ability to get up when knocked over, and some kind of ability to knock over opponents. So get your viking battle magic thinking helmets on and offer some suggestions. I'll try and summarize the rules for that competetion here soon.

Keith Rowell

It laks some of the angular agressiveness of the original, and it's not prickling with guns.

Keith Rowell Design

Tuesday, August 17, 2004

IronCad Sponsors the Project

I designed the whole project in IronCad so far. I really like the interface.

Looks like we can go ahead with the sponsorship

Keith Rowell Design

Hello Keith,

Looks like we can go ahead with the sponsorship. Our VP of Marketing is asking for "use of the IronCAD model files and images, promotion of the IronCAD name on any website or write ups and placement of IronCAD sticker on machine".

Best regards,
IronCAD -
700 Galleria Parkway Suite 330Atlanta, GA 30339
Toll-free: 800-339-7304

From: Shaun
Sent: Monday, August 16, 2004 5:39 PM
To: Kevin
Subject: RE: Knewt walker project


This sponsorship should be fun. Please ask for the following marketing promotional benefits (use of models and images, promotion of IC name on any website or write ups and placement of IC sticker on machine). If they go for this lets do it.

From: Keith Rowell
Sent: Monday, August 16, 2004 4:15 PM
To: Kevin
FW: Knewt walker project

Thanks Kevin for the the menu advice,
The "TBC" file problem happened only once before to me, many years ago, I'm very gratefull for your help.

We're getting ready to start cutting parts for this robot now. All the motors and batteries and electronics have been aquired and tested. I plan to enter it in the first (American) Robo-One competition, which will probably happen next year. I plan to use it for trade show promotions, other robot events, etc in the coming couple of years. It's completion date is scheduled for next spring.

The following email is my last update to the group that's involved in the build team. I have electronics types, mechanical engineering types and machinists contributing. I've been very lucky on this project in regards to attracting high quality advisors and I think it's mostly because of the "coolness" factor.

I approached you once before about the possibility of sponsorship in return for some some marketing worthy models. I'm hopefull that this project qualifies. Can you pass this message around to anyone who might be able to make that happen? I can offer logo space on the bot, and do some promotion events as well.


Keith Rowell

Monday, August 16, 2004

I recover from the crash, with an improved design

The head sits directly above the weight shift motor. The head plate and bearing attach to the neck plate. I can't wait to start cutting some parts. I'll start making drawings this week to make parts from. I have lots of screws and bearings and gears to show you Stuart. I want to get your advice on how to do some things.

Keith Rowell Design

Having started at the foot and worked my way up, we've now reached the head. Getting the neck support in around the motors was getting to be tiresome, but after 3 tries, I think this will work. I wanted all the wires to come through the neck but the simplicity of this "head turn motor" design convinced me to just bundle them from the back. All the electronics will go in the head. The back of the plastic toy head comes off and the servo motor driver board fits in nicely.

Keith Rowell Design