I often need to move something big. A pushcart is handy, but too slow for longer distances. A bike is faster, but its rack is too small to carry a washing machine, for example. If we combine the two and hitch the cart behind the bike, we get a nice little lorry, but with a few disadvantages: the bike needs to be carefully secured agains falling over when stopped; two braked wheels are sometimes not enough; a heavy load with high centre of gravity tends to lift the bike's rear wheel when being walked uphill; headwinds are nasty; and you need to visit your basement twice at the beginning and end of every trip. What if we cut off half of the bike and fix it directly to the trailer? Figuratively, of course. Such a cargo trike should solve all those problems easily.
Python class SWB recumbent delta offers the best combo of aerodynamics, comfort, stowability, cargo capacity and ease of build and maintenance. Overall dimensions were clear. Width no more than 82 cm to fit through our doorways. Length no more than 260 cm to fit vertically between floor and ceiling. Cargo bed length short enough to make enough space for a driver in front of it. Width and height of cargo space big enough to contain two banana boxes. Total height as low as possible for stability and aerodynamics. This is the result:
So much for the theory. I forgot to count rear wheels' axles and locknuts into the total width and exceeded the doorway by several centimetres. The trike can't be simply flipped on its side because it's too heavy and unwieldy to carry and I can't just pull it across the floor. Overall length grew by several centimetres after I had to extend the pedal boom (I must have incorrectly measured my leg length or something). If I lift the trike to a vertical parking position, the heavy nose flops to a side and threatens to break something, so strings must be used to keep it straight. So I park the thing either at home, split in half, or in some garage.
Many thanks to Gavin and Velodreamer (mainly here) for inspiration, Jerry for inline wheels, Alex for aluminium box samples, Milan from Azub and colleagues from mailinglist and forum for advice and Tom for letting me use his garage.
It took less than one year: I started to gather parts in summer, began to build in autumn, finished the hard work during winter and last details in April. Total cost exceeded 48000 CZK (roughly the price of two good quality bikes or three cheap ones), half of it being the internal gear hub (I'm crazy, I know :-) ). Tools and equipment used: drill press, hand electric drill, stick welder, angle grinder, vise and hand tools (hacksaw, files, clamps, hammer etc.). This time I tried to design and draw most of the stuff in advance to save later improvisation and fitting. I must admit that going by drawings is much faster than the trial and error method. Non-right angles are much easier to make if you print an actual size template.
After an experience with our local bike shop where I once ordered a 20" rim and after a month of waiting received a different type than expected, I contacted Azub who are familiar with small wheels. They told me thay didn't have rims, spokes and hubs stocked and if I didn't want fully assembled wheels, I'd better contact their suppliers directly - and attached a list of direct links. Perfect service!
Front wheel is built around a Rohloff TS-DB-OEM internal gear hub with a 180 mm disk brake, rear left around Sturmey-Archer XL-FDD (dynohub with 90 mm drum brake) and rear right around XL-FD (just a drum without dynamo). I didn't want cantilevered XL-SD hubs - they would increase track width, which is good, but at a cost of narrowing cargo space, because structural sidewalls would have to be placed on the inner side of the wheels. All rims are Remerx Jumbo 406×23 (the strongest model you can find here), laced with 36 single-crossed spokes. To be sure, I asked Azub if the Rohloff, with all its proprietary accessories, needed any special spokes. They told me it didn't, but I should avoid brands like roubárna Hvězdonice (Marek), CN Spokes and Mach and go for Sapim, Pillar or DT Spokes - differences in quality are remarkable and prices almost equal. After comparing neatly forged reinforced heads of the Pillars with stamping burr of the usual noname spokes, I see they were right (burr is a problem because it makes notches in hub flanges, possibly causing fatigue cracks). I also considered purchasing a spoke thread making tool, but it is quite expensive and would only pay off after about twenty wheels which I'm probably not going to build in my whole life.
Single cross lacing is pretty easy (note my ultra-precise cardboard jig for experimental verification of calculated spoke length):
Front brake caliper was a little problematic to install because the ideal place for it is occupied by left fork leg. The problem was solved by placing the brake down rear. It has to be removed to remove the wheel, but only the two adapter bolts need unscrewing, not the other two which need a lot of adjusting afterwards. Rear wheels were easier, I just drilled their reaction arms and anchored them with M6 bolts to the nearest frame tube. To remove the wheel, this bolt just needs to be unscrewed far enough to clear the arm.
If you see a 160 mm disk at the first picture, you see well. At first I thought it would be enough, but changed my mind after a scary experience with overheated front brake on another bike.
The frame is made of 15×15×1.5 mm square tubes, dropouts are 4 mm steel. Everything is welded with Ø2 mm rutile electrodes (exactly 90 of them were needed for the whole trike) at a current of about 60 ampers. Right angle corners are easy to weld, just clamp the tubes tightly to a piece of bigger square tube and nothing warps. Vertical struts in the middle of front and rear sidewalls and the U-shaped braces above the wheels are structural, but the corner struts carry almost nothing, so I saved a little weight by perforating the rear ones and making the front ones from scrap seatstays, welded by bits of their original dropouts at the bottom. Next time, I would made the front ones from perforated square tubes too because they are often used as handles and because LEDs fit neatly into the holes (but first time I planned a fairing).
The floor is made of 9 mm thick waterproof plywood with antislip surface (bought at Albakmen in Most which is reachable by bike from my home), bolted around the perimeter with M5 screws, and with recessions ground on the bottom where the weld beads stick out. Together with the frame underneath, it makes a sandwich strong enough to distribute point loads and carry the forces to the edges where the lattice sidewalls hand them over to the wheels.
Lattice sidewalls are made of Ø12×1 (stainless) and Ø10×1 mm (normal) steel tubes, bolted together with M6 screws. The goal of this relatively complex design is to carry the required 100+ kg of cargo while not weighing another 100+ kg itself. I resigned on the possibility to uninstall the rear wall, which would make loading very large and long objects easier, because the transversal beam under the rear edge of the loadbed would have to be much stronger than it is now to survive without the support of the diagonal struts.
The possibility of vertical parking is essential for everyday use. It is made possible by two stubby feet and two inline wheels on the rear wall, which also allow wheeling it around in the vertical position. By the way: seized inline bearings can be disassembled by prying a thin circlip out with a sharp needle, then the metal dust shield can be removed. Then just wash the balls and cages with WD-40 and repack with fresh grease. If the bearings are not rusty, cracked or pitted, they will work like new again.
I originally planned to make a rainproof streamlined fairing around the whole cargo module, but will probably keep it unfaired after all. The aerodynamic bonus might possibly outweigh the weight penalty, but I think it is more useful to be able to carry long objects which stick out on both front and rear.
I chose steering angle of 65° because it worked well on my first python, and ended up with 63° due to various inaccuracies during the build. Maybe it wasn't the best idea ever, because a rigid trike's riding characteristics differs a lot from a bike's. Most trike builders recommend an angle of 53° or just a few degrees more, saying it decreases turning radius and improves stability and ergonomy. But I'm not going to rework it, the current setup works well enough for me.
Rear half of the fork is perpendicular to the steering axis to make the construction easier, front half is almost horizontal to keep the bottom bracket as low as possible (it ended up 110 mm above the seat), limiting wheel flop and improving leg perfusion. Heels don't hit the ground even at maximum steering lock. Vertical distance of the two steering bearings (109 mm) is a compromise between stability (the lower the seat is, the better), strength (the further apart the bearings are, the better) and ground clearance (must be higher than a typical speed bump or low kerb to avoid getting stuck on them - I chose 80 mm and it's just enough, I wouldn't recomment going any lower).
Fork legs are made of 40×10×2 mm rectangular tubes, dropouts of 4 mm sheet. Much later, I realized that reaction torque of the hub is, with the chain transmission used, almost equal to the torque at the cranks, so the left dropout might be too weak. We'll see, so far it holds well. The four sideways bends in the legs must be welded after all the connecting parts are welded to the ends - now I know, it cost me three cuts and some electrodes. Front connector is 40×20×2 mm and pedal boom 40×40×2. At the rear, I repeated the proven design from last time: a vertical 20×20×2 mm square tube, with the steering axis going loosely through it, reinforced by two 20×10×2 diagonal tubes. There are 4 mm thick flanges at the top and bottom with precisely reamed 12 mm holes for the axis. When assembled and tightened, the spherical inner parts of the bearings make visible dents in the flanges which clearly show how nonparallel they are. I shaved them until the dents extended to more than half the circumference.
Handlebars and their stem came from a bike wreck which I had stolen from a dumpster. Before cutting the stem, I laid a wide bead around it to increase its wall thickness from the original 1.5 mm to 3, making it possible to weld it to the upper flange. At first I planned some vertical bar ends, but they are not needed after all because I can reach the flat bars as they are (although not 100 % comfortably, they should be about 2 cm further back).
The bottom bracket is movable. The shell was taken from the same wreck as the handlebars; it was a welded steel frame, so it could be filed to a nice round shape and welded easily. A simple design with two vertical sideplates didn't work exceptionally well last time, so I used short horizontal 20×20×2 mm beams fastened by four vertical M6 bolts. It works, but it needs some shape lock to prevent turning in the direction of chain pull and left pedal push; I used thin metal tubes on the bolts - hope they won't shear off. Chainline is chosen to put the big chainring perfectly in line with the sprocket because it is used most often. The small ring is just a reserve for super steep hills.
Main beam is a 40×40×2 mm square tube, seat brace and rear arms are 40×20×2, flanges for attaching to the cargo module are 4 mm thick (M8 bolts), two angled struts under front edge of the loadbed floor are 20×20×2 tubes held by M6 bolts (20×20 with a cutout fits perfectly around a 15×15). The contraption around the steering pivot at the front is a 40×20×2 mm tube and a 40×4 mm flat bar. The rear arms diverge at 94° which is a value where least material is needed. The point of this Y-shaped design is to minimize bending loads of the cargo module: weight of the front part of the vehicle hangs close to the rear wheels, so the two front struts only have to resist negligible frame twisting and imbalance of cargo.
Steering pivot is made of two SKF JAM 12 rod end bearings, the same as last time (inner diameter of 12 mm, outer thread M12, 12.2 kN of strength). The axis is 12 mm solid steel rod, dust covers are made of soft leather. An innovation is the possibility to adjust distance of the bearings, because tightening the big nuts so that the bearings end up exactly in the position of minimum drag is near impossible. Now it works like this: first tighten the bottom bearing's nut (the other one is inaccessible, welded inside the main beam). Then tighten the two nuts on the axis, pulling the bearings together. Then tighten the two nuts of the upper bearing, after which the bearings probably seize. Then turn the small vertical M5 screw or nut to set the correct bearing distance (few tenths of a millimetre); the two plates slide along each other and the upper beam bends a little. When the bearings stop to drag, tighten the two M8 bolts, fixing the plates together. It may sound complicated, but it is actually an easy five minute job.
Cornering stability of the trike would improve significantly if the whole front half with rider could lean into turns. That could be achieved by modifying how the midframe connects to the cargo module: instead of three fixed points, there would be two bearings, front and rear. Handlebars on the front fork would be replaced by two handles sticking out forward from the loadbed and ending besides the seat. Arms would control the leaning and keep the trike upright after stopping, legs would steer. High speed cornering would improve a lot, but some disadvantages would appear as well, which made me abandon this idea: it would be necessary to balance just like a singletrack bike (Python, of course), some locking device would be necessary to prevent falling over when parked, the loadbed would have to be shorter to clear the tilting seat, and three long cables spanning between front and rear part would need to be disconnected during every disassembly (plus some negligible extra weight).
Only the mounting hole pattern changed on the proven wooden design (template here), the plywood beams were beefed up a bit to better resist lateral forces (and then trimmed down again in some places to make room for the handlebars), and the width increased by a few centimetres to make leaning of upper body easier. The profile is the same as last time. Materials: Ø10 mm grooved wooden dowels, 15 mm construction plywood for the main beams, 5 mm furniture plywood for the side beams (better to wait with their installation until you're done fitting and filing all those holes in main beams) and some acetone glue for final assembly. Fore/aft position of the seat is not adjustable, because the steering axis must stay exactly under the driver's hips. The only adjustable parameter was the tilt angle, but I canceled it in the end because the most laidback position (34°) suits me best and provides most handlebar clearance. The seat is held by two M10 bolts: one goes through the upper rectangular tube of the steering pivot assembly, the other through an extra brace which is placed where it leaves most free space for an underseat bag - the loadbed is big enough for anything, but you can't reach there for a snack while riding. I used an old mattress foam for the seat cushion at first, then switched to Ventisit soon after first test rides.
Headrest is almost the same as last time: two pieces of wood bolted together, cushioned by plumbing insulation foam and covered in jeans fabric. The only difference is I finally managed to drill the mounting holes properly, so the thing holds in place and doesn't rattle.
There is a classic road double chainring with 42 and 52 teeth, the cranks are shortened from 175 to 155 mm. At first I wanted a single ring, but it was harder to find and now I'm glad for the second ring: the lowest total gear is almost exactly 1, which in theory is the same as getting off and pushing (in practice, the pushing is easier). I planned to use an old Romet front derailleur modified for reverse operation (it would work even when stationary), but there's not enough space for it to work properly, so I now ride without it, downshift with my heel (poor man's Schlumpf :-) ) and upshift by hand; there's usually enough time to pick up a leaf or twig (or a parched flat toad, eww) while trundling uphill to keep my fingers clean.
Mudguards are true Scrapheap Challenge products. Front one is made of two leftover standard 26" front mudguards bolted together. The enamel cracked when I hammered them to smaller diameter and had to be repaired by wire brush and paint. Side extensions are made of high pressure laminate sheet peeled off a discarded table, with the glossy waterproof side facing inside and the other side painted (strangely enough, they smell exactly like burning brake pads when wet). Their radius is slightly larger than the mudguard's, so they keep a slightly conical shape and don't rub on anything. Their depth is chosen so that the tyre isn't visible from the side - from my experience, it's enough for 100 % protection against mud spray; covering the whole upper half of the wheel is not necessary. Another affair is water spraying from the point where the tyre hits a puddle: there's no defense except lifting your hands from the handlebars which get sprayed (or slow down, but that would be a shame, right?). Rear mudguards are boxes made of scrap 5 mm plywood glued together with leftover plumbing silicone and secured by aluminium angles. Each of them weighs over a kilogram, so when I'm bored someday, I'll learn to work with laminate or find some sheet aluminium and try to make something lighter. Standard mudguards wouldn't be enough, they are located inside the cargo hold and must withstand being buried under a heap of heavy stuff.
Rear brake cable is hidden inside the midframe's main tube, the mantle ends on a bridgelike stop at the tube's end, the bare cable continues under the loadbed where it splits in two by a quick-release rocker arm. The two cables then enter two shorter mantles anchored on the loadbed frame, turn around and approach the wheels from the rear. A nice detail is that these two cables are fairly short and held by pinch bolts on both ends, so I can finally reuse all those cable leftovers without end stops. A remark for next time: before cutting a mantle, it's good to check for forgotten cables inside :-].
A trike doesn't fall off when stopped, but runs away on every slightest slope, that's why it needs a parking brake - equally useful thing as a kickstand for a bike. First I made it of a rubber ring cut from an old inner tube, slipped over the left handlebar and holding the left brake lever if needed. Simple and easy to use, but the rubber deteriorated and failed quickly, so I now use a styrofoam chock. It is coated with sticky tape to improve durability and equipped with a velcro strap to hang it on some horizontal tube.
Flat handlebars are the best hideout for spare spokes, all you need is a bit of foam to prevent rattling:
Lights are described in a separate article.
Bell is the little "ding" type with a plastic hammer on a spring which didn't work very well last time, but now, mounted in the correct position on top of the handlebars, people have no trouble hearing it. That implies these bells emit sound in a non-uniform pattern and have a deaf spot in the direction of handlebars - good to know. Second noise-making device is an AirZound (a.k.a. Biologic Blast) air horn. It is about as loud as a car horn, but sounds higher, similar to classic handlebar trumpets powered by a rubber balloon. The air tank (actually an ordinary half-litre PET bottle) is filled by a standard Schraeder pump and is enough for three or four long honks. I haven't used it against cars yet, but it repels stray dogs quite effectively.
Underseat bag is made of waterproof backpack fabric bought at a famous hobby shop at Florenc, Prague, and is tied to the seat by several cotton ribbons. Lesson learned: recycling zippers from discarded clothes sounds like a cool idea, but don't try it with the old ones with metal teeth on soft cotton base fabric. The whole point of an underseat bag is defeated if you have to operate each zipper with both hands and constantly pay extra attention to prevent jamming it.
In one word: easy. An obstacle ahead? No problem, just stop. Too high a gear for getting going again? No problem, just turn the grip anywhere you like. A bug under your glasses? No problem, you have two free hands to get it out. Other people's reactions? All positive.
Due to the three wheels, no learning was necessary, every possible control problem is solved by slowing down or stopping, there's no falling over. At first, steering felt twitchy at speed and tossed me around in unexpected directions, but those were just ingrained old reflexes from singletrack bikes where the steering is reversed. It improved over time. Ergonomy is OK, I was lucky to find the correct BB distance very quickly and no tendons or joints hurt, even after hundred-kilometre trips for several days in succession.
Tipover stability is good enough for normal riding, not good enough for slaloms and other stunts. It's pretty easy to lift a wheel by a sharp quick turn at speeds over 25 km/h. Above 30, I have to lean even into corners of standard road radii. When turning near the steering lock, leaning is necessary regardless of speed - centre of gravity gets very close to the line connecting the two outer wheels. I was glad for being able to reach the ground to prevent tipovers on banked and bumpy forest trails. The stability slightly improves with increasing weight on the loadbed.
Traction is very good, I can climb pretty much anything, including loose dirt paths. Front wheel carries about 64 kg, which is 58 % of weight of the 43 kg trike with its driver. Cargo makes no difference because it only loads the rear axle. Offroad, the trike rolls like a tank, unless it gets stuck on a root or sinks in mud. Acceleration is not bad, with all those slow gears at my disposal. Top speed on level ground is 35 km/h with all my power, comfortable cruising speed slightly over 20, average on long trips cca 16..20 depending on hills and cargo.
Brakes work well enough. The front one is not as sharp as I hoped, but strong enough if pulled hard. It does overheat on long slow descents, thermodynamics can't be cheated. Rear brakes are sharper than needed, it's easy to make the rear wheels skid, but they become more effective (and more necessary) with increasing cargo weight. Due to the cable geometry and friction, they engage at random order, so I had to get used to a momentary steering interference. Once fully engaged, the imbalance disappears. It's better to only brake with the front in turns because the unloaded inner rear wheel skids easily.
Safety in traffic seems to be slightly better than on a bike. Getting into a turn is faster because I don't have to start with a balancing swerve in opposite direction. Locking all wheels during an emergency stop doesn't cause a wipeout. I don't have to watch the road surface so intently, skids are mostly harmless. Having the widest spot at the rear makes other drivers give me more space when overtaking (and most of them slow down to have a look at that strange thing). The only weak spot are fast right turns: if I had to suddenly stop or turn tighter, I could tip over and roll into the opposite lane. So I drive carefully and slow down in advance.
It didn't go exactly according to my plans and wishes, but it works and serves its purpose, so let's call it a success.