I’d love nothing more than to say “I’m done” and get this product out, but in lieu of it being completely finished, I wanted to give you a sneak peek of what’s coming.   More than that, I wanted to give you a look behind the curtains to see how the Twin Dolphin speed climbing timing system is being developed and what my design and fabrication process has been like.  [Update: First Customer Ship - Done!]

Based on all the iterations and three years of design prototyping in speed climbing competitions, it was time to design it again from the ground up for manufacturability.  To me, this meant that all the mechanical components should be cut by a CNC so they all come out the same and the shapes could be more complex than hand cutting would allow.  It also meant that all the electronic components should be integrated into custom printed circuit boards instead of various sub-components like the light sensor and communication systems of the same circuit being wired by hand.

All the systems I built up until now were essentially built from electronic sub-assemblies I bought or created but were an integration of off-the-shelf components, very often from Sparkfun. It was hard to put together one system, and making a second one the same as the first was nearly impossible.  The prospect of making 10 systems was daunting.  So, the #1 goal was to design the system so it could be built repeatably and consistently.

On any major project, you have to start somewhere, so I started with the most critical part of the system, the sensor design for hands and feet.  Using AutoDesk Sketchbook Pro and a Wacom pen tablet, I started sketching and came up with some of the rudiments of the design based on previous generations, but adding some new features.  This was essentially the back-of-the-napkin phase.

One new feature I wanted was an integrated indicator light within the sensor.  In previous generations of the system, the finish indicator light was independently wired from the master clock and had to be bolted near the finish.  For the installer, it was another part to find a place for and install.  This is an initial sketch of the hand sensor.  The foot sensor works electronically identically as the hand sensor, but the platform is just shaped for a foot.

 

At this point, I was still imagining the light sensor being installed perpendicular to the sensor circuit board.  More on this later, but you can also see the rudiments of a sensor with a built-in indicator light taking shape.  After this, I started fleshing out a few more details, again in sketch form.

After getting the overall picture in my head, I started moving the design into a CAD package called ViaCAD 3D.  Ultimately, I expected to not just use CAD to make the plans or prints, but also manufacture it based exactly on the CAD drawings.  In other words, I wanted to put the full tool chain together from design to fabrication – CAM – computer aided manufacturing.  This way, I could achieve the goals of repeatability and consistency much better than if I was hand cutting and shaping.

Whether a hand sensor or foot sensor, the basic design is a sandwich of two wood parts with wire chases carved as tunnels into the bottom piece.  Furthermore, I wanted the mechanical design to help with aspects of the electronic design including, as much as possible, easy alignment of the laser and light sensor.

After many iterations, I finalized the following design for the hand sensor.  On the right hand side is a cutout to accommodate a part I’m calling a laser turret.  The laser turret allows the laser to be aligned in all axes (pitch, yaw, and roll), and then be bolted into place once it’s aligned with the light sensor.  More about this later.  This is a picture of how the top part of the hand sensor sandwich looks in CAD:

The left hand side with the cutout has a slight indent to accommodate a circuit board containing the light sensor and indicator light drivers.  The slot in the middle of the board is where an RGB LED strip is embedded as a built-in indicator light.  It has a small hole on the left of the slot leading to a slot below in the lower piece (not shown.)  

The bottom part of the sandwich has a slot that allows wires to be hidden and traverse the sensor from side to side.  These wires power the laser as well as carry voltage to the indicator light.  This part has the identical profile as the top portion so that the two parts fit together perfectly one on top of the other.

The top part is 1/2″ thick and the bottom part is 3/4″ thick which is a large part of making the overall sensor stiff and resilient to extremely hard hits by climbers at the end of their race.

ViaCAD does a great job of 3D solid modeling, but the version I’m using (8) does not have a good 3D rendering engine for photo-realistic images of the part which is often important to get a good feeling for how the piece will look when it’s cut and finished.  So for a good 3D rendering of a part, I rely on Strata Design 3D SE.  From ViaCAD I export the 3D model as a Wavefront OBJ formatted file and after importing the OBJ file into Strata Design 3D and a few setup steps, it’s easy to produce a 3D rendered, photo-realistic version of the top of the sensor:

 

Seeing these types of images on a large screen helps me get a feeling for how the part will work and act before ever committing it to be cut.  Even prototype cuts can take quite a bit of time, so it’s much more productive to work out the kinks in the design phase before iterating through the CNC.

Once I’m happy with the basic design in CAD, it’s time to move to CAM.  I have a Fireball Comet CNC router which I use to cut all the pieces for the system.  The CNC is set up in the front part of my shop where I can both open the garage door as well as more easily contain the dust to this space:

In order to go from CAD to CAM to CNC, I have to first save the 3D model in ViaCAD as an .STL file which is a common 3D data interchange format that my CAM software package can use.  Then using MeshCAM I import the STL model and generate the g-code required for the CNC to know how to cut the part.  

Each of these software packages is a domain of knowledge to master.  Having spent many hours creating parts and taking each part through the entire process to the point of CNC routing, I’ve learned all the right knobs and settings to make the tool chain spit out the part I’m looking for.  

MeshCAM generates the g-code which gets imported by the CNC software, LinuxCNC (formerly EMC2.)  I have a Linux PC attached to my network so it can receive files from my design workstation.  The Linux system is also attached indirectly to the CNC router via a controller which is a box roughly the size of a shoe box.  The CNC controller is attached via a bundle of wires to the CNC router .  The LinuxPC is the computer sitting on the shelf to the right of the CNC router in the photo above. There’s a lot more to talk about just in the CNC, but I’ll save that for another article.

When a part gets downloaded to the LinuxCNC system, when it’s loaded it gets shown as a type of wireframe model except instead of the wireframe being the outline of the part, the wireframe is the actual tool path of the cutter showing you exactly where the cuts will be made as it carves out the part.  In the photo below, I have an array of laser shrouds ready to be cut:

Once part’s g-code file is downloaded and the stock is loaded into the router, the cutting starts.  After 10-20 minutes on average in the CNC, a part will be completed.  The shot below shows the hand sensor top piece being cut.  

The idea is starting to take physical shape here.  I love that the fact that I can go from idea, to CAD, to CAM, to holding a physical part in my hands, sometimes within an hour depending upon the complexity of the part.

 

Below is a stack of hand sensor tops ready to be belt sanded to remove the structural tabs used to hold the part in place while it’s being cut.   As you can see from the photo below, the parts are starting to look much like the CAD drawing.

The hand sensor top is cut from 1/2″ MDF.  Once sanded, the part is ready for priming:

After priming, it’s time for painting:

I use a combination of HVLP (High Volume Low Pressure) paint sprayer and a hobby-sized airbrush to do the painting.  It’s usually at the paint stage that I start to marvel about how much the real part looks exactly like the Strata 3D rendering I showed you earlier.

I’ll talk more about the electronics design in another article, but after the mechanical parts are all painted it’s time to assemble and test the sensor.

Each sensor has a laser turret (another custom part design), a custom light sensor circuit board, and the RGB LED strip used for the indicator light.   In the photo below, the black disks with the green part are the laser turrets and the white strip with the colored wire are the RGB LED strips for the indicator lights.  More about the laser turrets later.

The sensor top receives the posts for mounting the laser turret.  The posts are #4-40 screws that are 1 1/4″ long.  These posts are countersinked on the back of the top piece so the top and bottom piece can fit flush and snug and the posts are fixed in place with a nut and some Loc-tight.   

The RGB LED strip is pressed into the center slot and this becomes the target for the climber to hit when he or she finishes the race.  Once the RGB LED strip is installed and the top and bottom pieces are joined, I fill the slot with a clear epoxy which simultaneously glues the RGB strip in place and protects it from the coming assaults by climbers. 

Another nut is added on each post to form an adjustable base for the laser turret.

The top and the bottom parts are glued and screwed together and the wire used to charge the laser is fed through the bottom part’s wire chase to exit the hole at the bottom of the laser turret.  You can see the LED and laser wiring coming out the bottom of the top piece here as well as the counter-siinked screws for the laser turret posts:

Once the wiring is fed through the top, the bottom and top parts are screwed and glued together, the sensor is clamped until the glue dries:

When the two parts are dry, I add the electronic components on either end – the light sensor board and the laser turret.  All wires are soldered directly so there is no chance a connector internally will come loose from vibration.

The light sensor board goes on the left side and is the target for the laser.  There’s a lot more to say about the design of this board, but it is one of the custom pieces of electronics I designed for the system.  The small crystal-like part in the middle is a right-angle light pipe that will glow when the laser strikes it.  That glowing is detected by the very small light sensor that’s mounted to the circuit board.  The light-pipe straddles the actual light sensor which is roughly the size of a fleck of pepper.

The laser turret is another custom designed piece which allows the laser to be adjusted in all axes needed to align it with the light pipe.  The laser turret was designed in ViaCAD as well and cut on the CNC using 1/4″ MDF.

You can see it has 4 slots around the perimeter which allow for lateral alignment of the laser.  The disk sits on 4 nuts and the posts shown earlier fit through the 4 slots.  The center slot cut in the turret fits the laser diode module (the part that generates the beam).  Finally, the hole in the back of the turret provides an opening for wires to come up from below – the wire chase in the bottom part.  

When the laser turret g-code is loaded into LinuxCNC in prep to cut the part, it looks like this on the LinuxCNC screen:

The picture below is the CNC cutting the laser turret parts.  Each row of turrets took about 45 minutes to cut out of 1/4″ MDF with an 1/8″ 2-flute end-mill.

 

 

 

 

 

 

 

 

 

 

 

 

 

When the light sensor and laser are installed and aligned, I power up the whole unit and do further alignment and testing.  This is the laser end of the sensor – you can see that when there is no obstruction, the indicator light is green and the laser is targeted to the sensor on the other side of the platform:

The picture below is taken from the light sensor side of the board and you can see the laser on the other end targeted on the light pipe and the light pipe is glowing.  The light pipe sits directly over the top of a surface mount light sensor component on the circuit board which is the component that determines whether the beam is broken or not.  The climber breaks the beam with his or her hand and the detection of that breakage stops the clock for that lane.

The other thing you can see is that the cutout for the circuit board in the top is automatically exactly in line with the laser path.  Thanks to the CAD design and CNC cut, it’s easy to install the circuit board in exactly the right position every time.  Here’s another shot of the light sensor board with the light pipe glowing from the laser beam:

 

 

 

 

 

 

 

 

 

 

 

 

 

This is a hand sensor undergoing tests on my work-bench prior to the protective shrouds being installed:

 After the board has checked out, the finishing steps are to add the protective shrouds over the electronics on both ends.  Again, these are custom designed parts that have rounded corners on them so climbers hands are safe if they don’t hit the sensor right in the middle.

This is a ViaCAD view of the laser shroud, set upside down for CNC cutting:

When this part is flipped over, it’s flat on top and the internal cutout is sized to cover the laser turret.  The square cutout on the right side in the image above is the tunnel opening for the laser beam to exit the turret.   Here is the laser shroud being cut on the CNC:

Each row of shrouds took about 75 minutes to cut, so this panel represents about 5 hours of cutting time which I did on and off over a period of a few days.

Because the shrouds are cut with a 1/4″ end-mill and the screw holes are much smaller so would have required a tool change in the CNC, it’s simply easier to quickly do all the screw holes with a drill press as shown below.

The part is made of MDF, so in order to not break out the bottom when drilling, it’s important to use a sacrifice board under the part.  This keeps the material from ripping when the bit goes through the bottom of the part.

Here are stacks of laser shrouds ready for painting:

The shroud to protect the electronics is thinner and is shaped to fit the sensor board, provide a cutout for the laser beam to enter, provides a hood so the light pipe is mostly in the dark unless the laser is pointed on it, and finally provides a slot for the RJ45 connector to exit the top of the sensor.  This ViaCAD shot of the light sensor shroud shows it upside down in the way it will be cut since it will be hollowed out by the router and the top is flat.  The CNC I have is considered a 2.5D cutter so it can only cut on a Z-axis coming straight down.

The sensor shroud goes through the same tool chain as described above and cut on the CNC:

 
Each row of sensor shrouds took about 1 hour to cut.

Once the sensor shrouds are installed and the T bracket used to bolt the sensor to the top of the climbing wall is mounted, the sensor is complete.  Here’s a pair of hand sensors, front and back:

 

 

 

 

 

 

 

 

 

The bumpers on the back are critical for IFSC style speed walls that are slightly overhanging.  They allow the sensor to be firmly snugged up against the wall to keep the back of it from slapping the wall when it’s hit.  This was a simple fix for a lesson learned the hard way.  In one competition, before having these bumpers installed, one teenager hit the sensor so hard, it slapped the wall and sprung back so the T-bracket was bent about 30 degrees out from the wall.  Everything survived, but ever since the bumpers have been an effective, permanent fix for that problem.

When the climber breaks the laser beam which runs right over the top of the indicator light the indicator light turns from green to red to help the climber and audience see that the finish got detected properly.

This is the laser side of the sensor with the shroud.  Even if the climber took a slicing swipe at the light and hit the shroud, his or her hand would hit a rounded part, not a corner.

 

 

 

 
This is the light sensor side (left side) of the hand sensor:

The T-bracket on the back of the hand sensor is screwed into the back with screws that are long enough to go through the back and part way into the top part of the sensor.  The main hole in the bracket is large enough to accommodate a standard 3/8″ diameter rock climbing socket head bolt with a washer.  This mounting design was carried over from previous generations because it was proven to be very strong and effective.

 

 

 

So, a route setter installing the sensor only needs to bolt the T-bracket to the wall and plug in a single cable into the jack on the top of the sensor.

The foot sensor uses exactly the same electronics but is mounted on a different platform in order to accommodate a foot.  The foot sensor is used for false-start detection and not to start the clock.  If the official starts the race and the foot sensor sees there is no foot in place, it immediate indicates a false-start.

The foot sensor ViaCAD design looks like this:

Overall, it’s narrower than the hand sensor because it doesn’t haven’t be hit while moving, but it provides a very stable base that won’t rock or twist on uneven surfaces like climbing gym pads.  The indicator light is at the heel of the platform because the official needs to be able to see the climber is properly in the foot sensor.  The climber is still able to see the light as well if he or she needs to.  The climber will step on the slot extending across the middle.

One of the things I found in previous generations was that the painted line indicating to the climber where to step would easily wear off, so this design dispenses with the painted pin-stripe line and carves a permanent slot in the middle to show that placement. 

The bottom piece is similar to the bottom piece of the hand sensor in that it provides additional support and thickness as well as slots for the wiring to be embedded and hidden.

The wiring chase goes from side to side and also extends to the heel in order to power the RGB LED strip used for the indicator light.

The Strata 3D rendering of the foot sensor top looked like this:

 
This is a little more like what it will look like in profile from a climber’s point of view:

It might not be obvious, but one of the hard lessons learned from the first year I fielded a timing system to USA Climbing nationals (2009) was to not put the sensor cable  directly out the front box of the foot sensor (they were sort of U-shaped back then.)  The gym had a slab start instead of an overhanging start and so climbers wanted shove the sensor as close to the base of the wall as they could get.  This ended up crimping or smashing the cable and RJ45 connector against the wall.  So, this design has a built-in toe-kick to keep that from happening.  Also, on IFSC style speed walls and routes, the climbers almost always start with right-hip into the wall, right foot on the first hold and left foot in the foot sensor.  

So between the built-in toe-kick and the left-hand light sensor placement, the cable is not a problem.  The same basic placement has been tested in national competitions over the last several years and has worked very well.

The same type of construction is used for the foot sensor except both the bottom and the top are carved out of 7/16″ OSB.  OSB has some nice properties for this part because it’s naturally got some texture built in – pressed slices of wood going every which direction.  Secondly, it’s extremely strong with little flex.  OSB is often used as flooring in houses.   Two slices of 7/16″ OSB means the foot sensor is not quite an inch thick.

The same posts are used to set up the laser turret:

Indicator light is installed in the heel:

 

 

 

 

 
The top and bottom of the foot sensor are glued and screwed together forming an extremely stout, unified piece:

 

 

 

 

Once the laser, electronics, and shrouds are installed and aligned, the foot sensor operates identically to the hand sensor.  Like the hand-sensor, the indicator light slot is filled with epoxy to create a translucent light bar that is very rugged and can easily tolerate being stepped or stomped on – though being at the heel make that less likely. 

You can see the glow of the light-pipe on the light sensor well protected by the light-sensor shroud.

 

 

 

 

 

 

If you made it this far, you are amazing for sticking with it.  And this is just the creation of the sensors!  In a future article I’ll discuss the electronic design of the light sensor and also the master clock and display clock portions of the system.  Stay tuned.

Right now I’m expecting the first commercial system to be available for sale by the latter half of June/2012 and will be opening up for preorders shortly.

Landon Cox – May 24, 2012.