Category: Recumbent Bicycling

The Bafang 48 V 11.6 A·h battery for Gee’s Terry Symmetry mounts on the downtube:

The battery slides onto a plate screwed to the pair of water bottle studs brazed to the tube:

Water bottle studs are (nominally) 65 mm on center. One stud normally appears under the plate’s center hole, with the other stud under either the upper or lower slot, depending on whether the battery fits better mounted lower or higher on the downtube.

However, the Symmetry’s downtube is so short the plate must mount with the lowest slot matching the uppermost stud, putting the lower stud beneath the metal compartment with its complete lack of mounting holes.

Well, I can fix that:

The upper hole in the metal base is 65 mm from the middle of the lower slot in the plastic baseplate, which will be (approximately) centered on the upper stud inside the black plastic mount. The location of that hole is not a free variable: it requires measuring and marking from the slot with the battery plate assembled.

The lower hole in the base puts the bottom of its plastic mount just about even with the end of the plate.

I shortened the battery side of the cable, crimped on (genuine!) 45 A Powerpole pins, and shaped the wiring to put the connector inside the metal compartment, out of harm’s way, and shielded from the weather.

The small bar of white HDPE serves as a cable clamp, held by a pair of M3 BHCS in the conveniently tapped holes.

With all that settled, the final iteration of the 3D printed mounting blocks took shape:

A station number from 1 through 4 identifies the blocks (station 0 is the blank block shape) and, of course, they’re all different. I refactored the OpenSCAD code used for Mary’s Tour Easy to put the feature selection into vectors, rather than convoluted logic:

Latches = [false,true,true,false,false];                // clearance for battery latch clips
Notch = [false,true,true,false,false];                  // notch for battery screw pockets
Recess = ["None","TeeNut","Bottle","Bottle","TeeNut"];  // stud or nut clearance against frame

HarnessCable = [false,true,true,true,true];             // passage for main harness cable

ShiftWire = [false,true,true,true,true];                //  .. shifter wire through sensor
Ferrules = ["None","Both","Front","None","Back"];       // ferrule and bushing ssockets

GearCable = [false,false,true,true,true];               //  .. gear sensor cable

Producing the features for a specific block is now a straightforward series of obvious choices. For example, adding the channels to clear the battery latches at stations 1 and 2 looks like this:

        if (Latches[BlkNum])
            for (i=[-1,1])
                translate([0,i*LatchOC/2,BlockMaxZ - LatchThick/2 + Protrusion])
                    cube([BossSlotOAL,LatchWidth,LatchThick + Protrusion],center=true);

Both parts of the block show the station number to avoid mixups:

Each block requires a bit under three hours of printing time, so they’re produced singly:

Building them sideways produces the best surface finish in all the recesses and holes. Small support structures under the rounded corners make them look Good Enough™ for their purpose.

A test assembly:

The two middle blocks (stations 3 and 2) sit at the water bottle studs. The rightmost block (station 1) is 130 mm from station 2, with the Bafang gear sensor on the rear derailleur cable.

An aluminum plate spreads the clamping force from the M4 screws across the bottom, as seen here below the cable stop cap holding the harness cable:

Those 50 mm screws are too long; a soon-to-arrive bag of 45 mm screws should fit perfectly. The final assembly will use nyloc nuts so they won’t vibrate loose.

The OpenSCAD source code for all the pieces as a GitHub Gist:

The Bafang BBS02 runs a fat “harness cable” from the motor to the four handlebar components (two brake sensors, throttle, and display), with a lump covering the junction where the four smaller cables emerge. Securing the lump to the head tube seemed like a good way to keep the motion in the (presumably) more flexible smaller cables:

From the rear:

I later bound the four connectors into a cluster using cable ties to further reduce the clutter and keep them from tapping the top tube.

The clip captures the cable tie in those indents:

The overhangs require easy cleanup with a square file to get rid of a few droopy threads. Avoid the temptation to print it standing up as an arch, because you want the perimeter threads to go around the whole thing, not across the thinnest sections. Trust me on this.

The OpenSCAD source code:

module HeadClip() {

CableOD = Harness[OD];

    difference() {
        linear_extrude(height=HeadTube[LENGTH],convexity=10)
            difference() {
                hull() {
                    circle(d=HeadTube[ID] + 2*WallThick,$fn=FrameSides);
                    translate([0,-(HeadTube[ID] + CableOD)/2])
                        rotate(180/(FrameSides/2))
                            circle(d=CableOD + 2*WallThick,$fn=FrameSides/2);
                }
                circle(d=HeadTube[ID] + HoleWindage,$fn=FrameSides);
                translate([0,-(HeadTube[ID] + CableOD)/2])
                    rotate(180/(FrameSides/2))
                        circle(d=CableOD + HoleWindage,$fn=FrameSides/2);
                translate([0,-HeadTube[ID]/2])
                    square(0.75*CableOD,center=true);
                translate([0,HeadTube[ID]])
                    square(2*HeadTube[ID],center=true);
            }
        translate([0,-(HeadTube[ID]/2 + CableOD + WallThick - CableTie.z/2),HeadTube[LENGTH]/2])
            cube([HeadTube[ID],CableTie.z,CableTie.y],center=true);

       for (i=[-1,1])
            translate([i*(HeadTube[ID]/2 + WallThick - CableTie.z/2),0,HeadTube[LENGTH]/2])
                cube([CableTie.z,HeadTube[ID],CableTie.y],center=true);
    }
}

I briefly thought of holding two pieces together around the head tube with M3 screws, but came to my senses: a cable tie is exactly what you want when holding a cable in place. Right?

The old-school “aero” brake levers on Gee’s Terry Symmetry bike have rubberoid cushion covers, so I slid the Bafang brake sensors inside:

They make the grips somewhat wider, but I can’t figure out a less destructive way of installing the things.

I glued the magnet inside a holder contoured to fit the space available:

Knocking the corners off makes it much more finger-friendly.

It’s unobtrusive with the handle released:

When you squeeze the lever, your fingers are nowhere near the magnet:

The lower edge actually slides along the brake lever housing without touching, but it’s a near thing.

Those are the same magnets I used for the Bafang brake sensors on Mary’s Tour Easy, once again aligned to aim the strongest volume of the magnetic field toward the sensor. The brake sensors activate just before the pads touch the rims and release with the magnets a few millimeters away from the sensors.

A complete coat of JB Plastic Bonder urethane adhesive covers each magnet to both isolate it from the weather and conceal the fact that they’re recycled from a power toothbrush.

Now that I know they work in this position, I must ease adhesive underneath the sensors so they don’t move around under normal hand pressure.

The OpenSCAD source code snippet:

module BrakeMagnet() {

    Magnet = [10.5,3.0,5.5];
    Plate = 2*ThreadThick;
    BrakeRad = 10.0;            // brake handle curve Radius
    Holder = [2*BrakeRad,7.0,Magnet.z + Plate];


    difference() {
        intersection() {
            translate([0,-BrakeRad,0])
                rotate(180/24)
                    cylinder(r=BrakeRad,h=Holder.z,$fn=24);
            translate([0,BrakeRad - Holder.y,Holder.z/2])
                cube([2*BrakeRad,2*BrakeRad,Holder.z],center=true);
            translate([0,0,-2*BrakeRad/sqrt(2) + Holder.z - 3.0 + BrakeRad])
                rotate([0,45,0])
                    cube(2*[BrakeRad,2*BrakeRad,BrakeRad],center=true);
        }
        translate([0,Magnet.y/2 - Holder.y - Protrusion/2,Magnet.z/2 + Plate + Protrusion/2])
            cube(Magnet + [0,Protrusion,Protrusion],center=true);
    }

}