All posts by David Black

Day #25 – 1/28/2014

Lab open time 1:30 PM

Today was not an official meeting day, but we got the baseplate riveted to the side rails, for both frames.  A little bit of “match drilling” was required, but it wasn’t too bad.  Also, we cut and drilled 6 holes the longitudinal member that spans the center of each robot frame.

riveted_frameLab closing time: 4:30 PM

Day #19 – 1/22/2014

Lab opening time: 1:30 PM.

The rendering farm is complete!  With iTalc, we have full abilities to power on the 4200 lab, render jobs, and power down without ever setting foot in that lab.  We have 6 servers there, 5 of which have 5 slaves (counting the server), and one with three slaves (again counting the server).  The 28 workstations there, plus the 16 workstations in the 6302 lab makes for 44 machines or 176 cores of rendering power!  We should be able to do more detailed renders much more quickly now.  And also, if the 6302 lab is being used, we can just push the renders to the 4200 machines.  Testing shall commence tomorrow.

Whilst planning of the rendering farm implementation was taking place, the students, Mr. Black, and the school principal Mr. Dall discussed names for the so-called “Choo Choo” mechanism.  Mr. Black initially thought of the name RAIL, an acronym meaning Rotationally Actuated Impulse Launcher.  While Mr. Black still likes the name, we realized that no part of the mechanism has anything related to a rail, so we went back to “Choo” looking for engineering related words to make it into an acronym.  After an exhausting vocabulary search, we determined the following (with a final edit as of this blog post):

CHOO – Cycloidal High-Output Operation.  We also decided that since we’ve been referring to it all along as the “Choo choo” mechanism, there needs to be two of them.  If you take one choo, and then you add another choo, you get a choo choo.  So, we plan to design the robot with one of these mehanisms on each side, instead of just one in the center.

Mr. Black and Mr. Widholm discussed the design of the Choo-Choo mechanism and began to calculate preliminary gear ratios.  They speculate that 64:1 ratio on a Mini-CIM motor will be approximately optimal to pull down the launcher mechanism in under 1 second time.  This is operating under the assumption that two choo mechanisms will be used, each powere by a MiniCIM, and both mechanically linked to be in sync by means of hex shaft between them.  Tomorrow, we will attempt to design this gearbox.

Below is a fantastic explanatory graphic from user “Ether” on the Chief Delphi forums which storyboards the motion of the Choo Choo mechanism.

chooLab closing time: 6:45 PM.


Day #18 – 1/21/2014

Tuesday, January 21st was officially a non-meeting day for rest, recovery, homework, and life, but a few still came in to work.

Mr. Black fabricated a mount to attach a new HP 8300 Elite SFF PC to the rolling cart with the 40″ Sceptre display, and bundled the wires neatly, then opened the lab at 2:20 pm for the Eco Challenge and animation teams.

The animation team configured and tested iTalc and Backburner in the 4200 lab for Distributed Bucket Rendering over the network.   We are seeing great performance in rendering full frames across the network slaves.  The decrease in render time is basically as many “fold” as the number of computers you use.  If a frame takes 8 minutes to render on one workstation, it’s taking 1 minute on 8 workstations.  The 3rd-gen i5 processors in the 4200 lab are rendering approximately 15% faster than the 2nd-gen i5 processors in the 6302 lab.  Also, we’re having a blast watching CPU loads on all computers simultaneously in real time.  It’s getting warm in here.

We’ve set up one server for the  i5 workstations (8 slaves in the cluster), and one server for the Core2Quad workstations (8 slaves in the cluster), and one server in the 4200 lab, with 7 slaves in the cluster.  We’ve noticed in the 6302 lab, the render job maxing the CPUs on 7 of 8 slaves, and in the 4200 lab, maxing the CPU on only 5 of the 8 slaves.  So, we plan to reduce the cluster size to 5 slaves in the 4200 lab, and run more server machines there.

We experimented with exporting to TARGA format with the Split Scan Lines option ticked, but it seemed to hang on a particularly tricky part of the render.  Further testing is needed.  Tomorrow we plan to finish setup of the remainder of the 28 rendering slaves in the 4200 lab.



Day #12 – 1/15/2014

January 15th was the second non-official work day, but a few came in to work nonetheless.   The lab opened at 12:00 Noon.  Nick C. fixed the dimensions of the frame in the CAD model.  Mr. Black discovered a major issue with the design of the gearbox plates in that as-designed, the chain would interfere with the lower standoffs.  The lower standoffs were moved, and Nick C. put in new pocketing for weight reduction.  Mr. Black reviewed the model and found and noted approximately 10 other minor items that need resolution before sending the models to machining.

Cynthia explored design of a motor-driven “Choo-choo” cam mechanism to pull and release the spring-loaded launcher  and made a cardoard mockup.  Mr. Black tested the SMC quick-exhaust valves with flow controls, and found them to not exhaust quite fast enough.  New exhaust valves are on order with a much higher specified flow rate.

Mr. Black installed a coolant washdown hose to the HAAS Mill.

Lab closing time: 5:00 PM.

Day #6 – 1/9/2014 – Mechanism Design Meeting

Lab opening time: 2:20 PM.

Finished machining the first operation on seven of the wheels.


First operation on the wheels


Tomorrow, we hope to finish the first operation on all the wheels, and setup and begin the second operation to finish the wheels.

Mr. Black did preliminary design review of 3-motor shifting gearbox, found no major issues (although a few minor ones), and put in an order with VexPro for the gears to build 5 gearboxes.  Mr. Black upgraded his teacher computer to 16GB of RAM to prepare for working with the whole robot model, and cascaded the existing RAM to upgrade the business room computer to 8GB for better performance in Inkscape and other design software.

While this was happening, the entire team had a meeting about robot mechanism design. Although we had TONS of success with a 1114 style ball puncher (videos in previous blogs), we decided to pursue a catapult design for the rest of the season. While a puncher claw would have been ideal for shooting and intake on different sides of the robot, the added complexity and weight of the arm made the design undesirable. Our current sketches of a catapult system involve being able to intake from the front and back of the robot, as well as “funneling” deployable side panels that would aid the robot in human player feeding and catching the ball.


Catapult Prototype


One major pending issue with the robot design is making a final determination of the robot’s width and length.  These parameters may be dependent upon the type of mechanisms.  We would like to design the base plate over the weekend to remain on schedule, and need these dimensions first.

We missed ordering from Coast Aluminum today, so we likely will not have material for frame rails this weekend.  This may put us an additional 3 days behind.

Lab closing time 9:45 PM.  The HAAS and MakerBots are doing a bit of lights-out manufacturing, which is super cool.  All the machines are still working well after we all go home!  The HAAS will be done around 10:30 PM.  To run a lab with the ability for robotic machines to make parts for other robotic machines while no humans are present is just fascinating and stupendous.

Cleanup score : 92% (one major issue with an item left out).

Day #2 – 1/5/2014 – Manufacturing Progress

The day started at about 10AM with me taking the Canyonero, err, Suburban to Ghanal Lumber in Pasadena, where a nice gentleman cut all the pieces of plywood for the high goal for a modest fee.  Total cost, approximately $100 for the two sheets of plywood, a dozen 2x4s, and about 26 cuts.  It was a big time saver to have them cut it on the panel saw at the store.  Hopefully the field crew can get this together soon.  I brought in an air nailer, but this tool will be for adult use only (for now).

In the CNC router room, special thanks to Mr. Hoard and Mr. Widholm for finishing the awesome new dust collection system.  Check it out when you get a chance.

On the CAM/CNC side of things, I got the 1st operation of the bearing blocks programmed, which took an hour.  I decided to make the two bearing blocks in one go out of a single piece of 1×2 barstock, 5.9″ long.  There are 7 tools in all, but it’s a rather straightforward part.  Programming took about an hour, and machine setup took about an hour.  Big thanks to Mr. Widholm and Julien C. for setting up the two vises on the HAAS!

I decided to use a thread-forming tap, which is different from a thread-cutting tap.  It’s supposed to not create chips, which is what causes a tap to bind and break.  Good idea right?  So on the first part, Tap goes 0.3 down into a 0.45 deep hole, and promptly proceeds to pull the part out of the vise, and shear itself off.  Well, there went about $15.  Need to get more taps.  I knew the speeds and feeds were correct.  10-32 tap, 800RPM, 25 IPM.  What gives?

After further investigation, I found OneCNC XR5 CAM software did not post a proper G84 tapping cycle code!  It was feeding with a G01! So, we had non-synchronous motion between the spindle RPM and the feedrate.  This is a dangerous bug in OneCNC that I’ll be reporting to Patrick Matthews at OneCNCWest.  When setting up a tapping operation in OneCNC, you MUST select “Machine Cycle -> G84” NOT “Automatic cycle”.  If you select Automatic cycle, it treats the tap like a drill.

After resolving that, and a few other minor adjustments to feedrates and cutting depths, we were back in business.  I had to tweak the countersink and corner rounder tool length offsets for a good fit and finish.  1/2″ endmill roughs it out at about 100 IPM, and finishes at about 50 IPM, all at 5500 RPM. Single-flute HSS Countersink is good SLOW at about 450 RPM and 8 IPM.  Drills are drilling at 6k RPM and about 30 IPM.  Tapping at 1440 RPM and 45 IPM.  I’m still not really sold on the form-tapping, as I’m not sure if we have the same thread depth/engagement.  It may just be an optical illusion when looking in the thread.  A screw seems to fit in okay.  Total part run time is about 5 minutes for the pair.  9 good sets of two bearing blocks were completed;  enough for two robots plus one demo set. A picture of today’s progress on the bearing blocks is shown below.


I consider this progress to be one full day behind.

In the morning, I’ll wake up at 6:00 AM to call Fry Steel to try to get some metal on the truck for delivery for the next set of parts.

Tomorrow, I will bore softjaws to flip these over, and program and machine the backside to make the coaxial joining tube features.  Then we’ll do the 5 bearing blocks for the gearboxes out of 2×3 bar stock on-hand, then we’ll get to sprockets from 7075, wheels from 4″ round 6061, and wheel axles from 7075 PG 1/2″ round rod.

We need more loc-line and fittings and nozzles.  Two just isn’t enough for tools of all different lengths.  Also, I ordered two more ER-25 3/16″ collets, because we have a lot of 3/16″ tools we use often.

Shop closing time today: 10:30 PM.