Recumbent trike powered trailer
The Electric Powered Trailer
Jan 2009 - A project by Stephen J. Hardy

Background
Like many recumbent commuters, the one thing I envied about my fossil-fuel-burning colleagues was the way they arrived at work completely fresh and not needing a shower. This motivated me to design a bit of electric assist for my trusty trike, to make that 13 mile morning commute happen without drawing too much perspiration.

The first, and obvious, idea is to add some electric power to the trike itself, like the truly inspiring job done by Matt Shumaker. Although I have a CNC mill, the amount of work involved was daunting. Also, my trike has a lot less room under or behind the seat, so it would have been altogether too challenging. I didn't really like the idea of using a single hub motor on the back wheel either. I'm not really very knowledgeable about bike mechanics, so I was reluctant to do major mods to the trike.

The idea died for a while, but then one day I was contemplating the problem of also giving my wife some electric mobility (she also has a trike recumbent, but it's a delta whereas mine is a tadpole configuration). I also wanted to be able to use my trike more often for small errands around town. The idea suddenly came to me that it would be possible to make a powered trailer. The trailer could be hitched to either trike, and because it was self-powered, it would be effectively "weightless" as far as the rider was concerned.

Specification and Design
Having decided to go this route, I started working out the basic requirements. It had to have at least 15 miles range on flat ground, be able to push the rider and a 200lb load at about 25mph. I really had no idea how much power this would require, but I guessed that about 1000W would do the trick. For ease of construction, I decided to use the brushed Crystalyte hub motors, with 20 inch rims, one on each side. These are not very powerful motors, but having two of them would make up for that. After all, I wasn't that interested in break-neck acceleration, and here in the California central valley the highest mountains are freeway overpasses.

I then needed to decide which battery technology to use. Lithium chemistry is very sexy, but the cost is exorbitant and they lose capacity just by standing (particularly in a hot garage!). NiMH and NiCd were also a bit beyond my budget, so I settled for good old sealed lead-acid (PbAc). Although the cycle life is relatively short, this can be extended dramatically by not deeply cycling the cells. Thus, I decided to employ twice as many batteries as I really needed for my desired range.

Normally, the weight of PbAc is prohibitive, however the trailer actually benefits from the extra weight, since it would otherwise have insufficient weight on the wheels to push the combined weight of 200lb (rider and trike) along.

Anyhow, I calculated that I would need a 48V bus voltage to spin the 20 inch wheels at the desired speed. Minimum energy requirements dictated about 12AH at this voltage, however this means that 20AH batteries are required since the nominal AH rating is for 20 hour discharge. Discharging over 1/2 hour gives only 50-60% of the nominal energy delivered. Then, since I wanted double this capacity to avoid deep cycling, that made a total of 40AH 48V. Fortunately, 12V 20AH batteries are a nice compact 7 x 6 x 3 inches, weigh about 15 lbs and cost about $50. I ordered eight of them from Batteryspace.com.

The hub motors take a maximum of 25A, thus the battery bank delivers at most 50A. This is well within the specification of the PbAc batteries.

As far as overall construction, I wanted to do sleek aluminum and/or carbon fiber tubes, but since I don't have a tube bender, a TIG welder, or a whole bunch of other metalworking stuff, I settled for good old plywood. Actually, very good plywood. The best I could find. The result looks rather quaint, like a piece of furniture on wheels, but it's very strong, and nobody ever suspects it pushes me along.

Let the Over-engineering Begin
With a project of this complexity, I didn't want to be stymied by any unexpected deficiencies, whether theoretical or practical. At the beginning, I had no idea whether various imagined problems were going to be real or not, so I had to be conservative.

As it happens, I really did over-engineer and over-specify this beast in quite a number of respects.

First was the question of whether the motors should be reversible. Yes, I thought, obviously I need to be able to back up that heavy trailer without getting out of my seat.

Should I be able to use kinetic energy to recharge the batteries? Of course!

Should the two motors be wired in series, parallel, or independently? Series is good for torque equalization, but each motor can only turn at half speed. Parallel is like locking the two motors together on a single shaft - no good when turning corners. Also, if the motors are not identical then they might not share current equally, and may waste power. Thus, the drivers needed to be independent, so I would need two 25A drivers, not a single 50A driver.

Because of the special requirements (so I thought) of the motor drivers, I decided to design and build my own. I have the experience in switch-mode design, so it didn't seem like too much trouble (and indeed sounded rather fun).

The Story In Pictures

 View from front during initial construction. Back section (where control electronics are situated) not finished, since at this stage I had no idea how the electronics would fit in. The holes are not for losing weight - they are to allow air flow front to back, since otherwise the batteries and electronics may cook in the heat. The square hole at the front is an alternative position for the trailer arm. The arm is on one side for hitching to a conventional bike or tadpole trike; and it is in the center when hitching between the two rear wheels of a delta trike.

 

Most construction is 12mm mahogany marine plywood, except that the internal braces are 1/2" AC plywood - strong but ugly.

View from right side. The green squares are neoprene foam pads, for supporting the batteries (two of which are installed). The ribbed construction makes this extremely rigid. The triangular corner braces strengthen the butt joints, since plywood is rather prone to delamination. Many more corner braces were added, using solid maple to spread stresses over a larger area. Polyurethane construction adhesive was used throughout, since it is a bit more forgiving of poor joinery than wood glue. It's stronger than the wood itself.
Just visible at the bottom centre is one of the aluminum "dropouts" for the hub motors. These motors don't have much torque, so the reaction torque which exists is resisted by the 3/8" aluminum plate with milled slots. This 6061-T6 aluminum is perfectly adequate for this application. The plates are glued and screwed to the sides. They were carefully aligned by optical sighting. The wheel rims are parallel within 1mm.
 Remote control and display box. Throttle pot on right handlebar. The bent black stick on to of the control is the radio antenna. There is no electrical connection required when you hitch up the trailer. It's all done via the magic of "Zigbee", which is a 2.4GHz, low-speed, data link (similar to WiFi but not as fast). The controller sends throttle and braking signals to the trailer, and the trailer responds with vital statistics to put up on the LCD display. The remote control itself runs off a small LiPo cell, which needs to be recharged after about 24 hours of continuous use.
Throttle pot detail. The throttle pot is a high quality conductive plastic servo position sensor. It rotates continuously, so it has no known "home" position. Whenever a brake is hit, the controller zeros the pot position. The driver then has to rotate the pot forward with the thumb in order to start moving forward. Not visible is a rubber friction device, which stops the pot moving under the severe vibration conditions. The knob on the pot is a toothed rubber belt pulley from an old printer. It can be rotate from full forward to full reverse with a flick of the thumb. Since the rubber O-ring holds it steady, it's like a crude cruise control.
The controller is designed so that the throttle position relates to motor torque (that is, armature current). Thus, acceleration and deceleration are always smooth, although it's a bit difficult to maintain a constant speed with this control mode. Future software will allow setting a true cruise speed.

Although the hub motors don't have much torque, the torque which is there is the same from zero up to top speed. Compared with a car, initial acceleration is slow, but it just keeps pushing and before you know it you are at top speed. I like to give it a boost by pedalling at the start, which also fools other road users into thinking I am some sort of super athlete... nobody ever suspects a trailer to be actually pushing!
Brake switch detail. Both of the front disk brakes have microswitches which are normally closed, but open when the brakes are touched. Both switches feed independently to the controller. If either contact is open (or the switches are not plugged into the controller) then the controller forces zero throttle and/or regenerative braking. One brake activation gives about half the maximum motor braking effort, and two brakes gives full reverse thrust. Only rarely do I have to actually use the disk brakes on the trike.

Once, because I am so low to the ground, some dozy driver cut me off in a roundabout (at 25mph, I'm often right in the traffic). Even though I have a 6ft day-glo flag, sometimes (but thankfully not often) drivers just see right through me. It's the mind-set of seeing only what one expects - other cars - and anything else is invisible.

Trailer hitch point on left rear dropout. I thought long and hard about the hitch design, and came up with some outlandish ideas. Eventually I settled on using a ball joint rod end as a simple and elegant solution. As shown, it allows unlimited yaw, and +/-15 degrees pitch and roll. The rod end is very strong, but I made sure this one was "thrust rated". This provides additional resistance to the ball being pushed out by downward force. This may not be suitable for an ordinary two-wheeler, since if the bike fell over, there would be too much "roll" and the rod-end would be damaged.
The piece of aluminum bar is manually milled out on the side which faces against the disc brake mounting lugs (which were unused on this particular trike model, and thus convenient for me to use). The milled recess prevents the part from shifting. I haven't yet had any problems with this design, in spite of the high vibration and weight.
 Hitch end. This is the end of the arm which couples to the rod end. The quick-release pin at the end (1/4 inch diameter) is inserted through the ball of the rod end, which is placed between the upper and lower flange of this part. Just visible is a black rubber pad which, along with one on the upper flange, provides cushioning and damps out vibration from up-and-down motion of the arm. For side forces, the clearance is quite small so there is no noticeable noise.
Adjustment of trailer arm angle. This uses a component known as a clamping handle, along with a special bolt (known as a square shoulder rod end). The multiple leaves of the joint multiply the friction obtained by the cam action of the handle. This joint yields if there is a violent acceleration (such as a crash) but is rigid in normal use. The arm angle may be used to position the trailer further out in the street, or nearer the curb, as desired. It is used straight with a delta trike. The arm can be removed and inserted vertically in the trailer.
In this position, the arm is bent at right angles in order to make a comfortable horizontal "handle" for pushing the trailer around inside buildings etc.
Trailer front, with arm removed. In normal operation, the arm fits into one of the two square holes at the front, and is locked in one of several extensible positions using the 3/8 inch quick release pin. The pin is actually inserted from underneath the trailer, but it's easy to locate by feel. The arm may be moved in and out and locked in one of three positions, according to which hole in the arm is engaged by the pin. Normally, the shortest extension is used, but a longer extension can be used if long items are being carried on the trailer.
 View from front right. Arm in normal position. Blue radio antenna visible. The caster is a surplus hospital bed wheel. It's very smooth running, and has a good locking mechanism, which is invaluable when trying to hitch up on a slope. The caster does not contact the ground when hitched up to the trike. I have added some eye bolts on the fenders, for attaching bungee cords etc. The embedded washers in other positions are also 3/8 threaded inserts, which I use for attaching various accessories.
View from rear left. Camera gives no indication of how bright those tail lights really are! They're painful to look at up close. The tail lights are made from medium-power LEDs, like the ones used in modern automotive marker lights etc. There are 18 in each string, and a 100 ohm series resistor. This gives about 50mA when connected to the 48V bus. At present, they are bare circuit boards, but one day I hope to scrounge some red lenses.
Under the hood. Basically a bag o' batteries. Battery charger at lower left, control circuits at upper left. I decided to carry the charger with the trailer at all times, so it just plugs straight into a wall socket when I get to work, or back home. The cord for the charger is tucked in to the triangular recess at the lower left. One day I'll make a nice little cover for it.
Controller. Xbee transceiver at lower center. This sits on a Rabbit RCM4500W computer core module, which in turn sits on my controller board. Motor drivers are under the aluminum bar, partially obscured, to left and right of controller. There is also provision for a GPS receiver module, but I haven't had time to install it yet. I had great ambition of logging all my travel with GPS, but good software (like fine wine) takes time.
Controller, viewed from right. Main 70A circuit breaker visible at rear. There is a string attached to the circuit breaker lever, and that string is attached to a knob at the front of the trailer. In an emergency, you can pull that knob to disconnect the battery. When developing the software, that emergency stop came in handy a few times, I can tell you! One concern I had was the possibility of fire in case of a short circuit. With a wooden vehicle, fire is that bit more serious. With this paranoia in mind, I made sure the circuit breaker did its job.
The big blue cylinder is a capacitor. This is overkill, but I had one anyway from some scrounged computer mainframe equipment, so I put it back to work in a *real* job.
SMPS and booster circuit. Shows final (rigid) heatsink mounting. Red marks show component positions in the SMPS, so that I wouldn't drill into anything delicate.

Basically, the battery charger is a 320W power supply. This generates a fixed 24V, so the extra circuit board on top is something I designed to boost this up to a maximum of 60V, with a regulated maximum current. The amount of voltage boost is controlled by the microprocessor (hence the gray ribbon cable).

A 12V PbAc SLA battery actually likes to be charged to 15V if used in deep cycling mode. This helps balance all cells, and yet it does not cause too much hydrogen generation. Since I don't deep cycle, the controller limits the voltage to 14.2, which should be optimum for long life. So far, these batteries have been pampered!

The booster circuit worked perfectly first time I tried it. That's definitely not normal for me. Unfortunately, the original heat sink was a loose flap of aluminum which hung over the fan outlet from the SMPS. Good position, except the vibration eventually fatigued the component leads right off. I caught the tell-tale whiff of escaped magic smoke when I plugged it in for a charge, but it was too late by then. Luckily, I had enough components to make another. This time, I bolted that heatsink (the aluminum angle) good and rigid.

As an aside, components are amazing these days. Even though the booster handles almost 300W of power, it hardly gets warm. Most of the smarts are in a tiny little surface mount IC visible just behing the ribbon cable connector. It switches at 500kHz, but even so this booster gets about 98% efficiency! If you don't go for surface mount, some of these cool devices are not available. It's worth learning how to do surface mount. I'm a convert!

Motor driver. Components at rear left are the 30A hall-effect current monitor. Blue component underneath on right is part of a temporary setup for measuring voltage spikes. The mosfets (for those interested) are Fairchild automotive rated devices, which can handle over 60A at 60V, at up to 175 degC. This is a full H-bridge design, which turned out to be overkill since I don't really need reversing like I originally thought I would. Also, if I was to do this again, I'd go for the 75V parts, since I did encounter some initial trouble with voltage spikes.
Designing these things is not as easy as it looks. Now, I check everything with the Linear Technology "Switcher CAD" program (free from Linear). I did manage to keep voltage spikes within 3 or 4V over the bus voltage, but it's a bit close to the limit for comfort. Then again, I'd like to try brushless motors next time, and that takes 6 mosfets...
Remote control and display internals. Radio antenna added as an improvement to the original design, which didn't have enough power to get through all the layers of plywood. Hence the hacks on the rear cover. Connector for throttle and brake switches on bottom. This is also used for charging the small lithium cell in the center. Another Rabbit RCM4500W is used for the computational power. Normally, this would be considered total overkill, but since I work for Rabbit they were happy to "lend" me a few parts for an interesting project.

Actually, the radio link was the part of the system which took the longest to get right. I started off hoping the the inbuilt antennas would be sensitive enough for reliable communication. Unfortunately, many things were conspiring against signal integrity. First, the trailer controller was at the far end of the trailer, and there was a layer of plywood above it. Then, the remote control/display was at the front (furthest) end of the trike, with its own layer of plywood. Worst of all, there was this great lump of a rider directly between the two radios. The upshot of it all was that the trailer would lose connection about 5 or 6 times during the commute, and often at a most inopportune time - stopped at a light in top gear (have you ever tried starting off a 300lb load in 27th gear?). After trying various remedies, I have finally added an external antenna to the remote, and provided a front-mounted antenna on the trailer.

Remote control, showing normal display of Volts, Amps, Watts and Speed (derived from motor back EMF). Also shows kilojoule usage at top right. Slight wrinkle in the keypad is from getting drenched. This is a fair-weather friend! The display is a standard part number from Rabbit. It doesn't have a really big display, but it's good for a few lines of text or graphics at the current 3ft reading distance. While riding, it's possible to lean forward and fiddle with the switches, to change displays and so on, but most of the time I leave it on this display unless diagnosing a problem.
The big silver button is the power on switch. The knob is hooked up, but is currently not used. One day it might be a cruise control setting or something.

Making an electronic enclosure out of wood is not a common practice and, it must be said, looks pretty low-tech (as does the wooden trailer, of course). Well, my wife wants me to build some furniture, so I thought I could kill two birds with one stone and get in a bit of joinery practice.
Spot the missing component! This is what happens when you ignore vibration as an enemy of electronics. This was the first construction of the battery charger booster. An ignominious end for a hard-working circuit.
Some specialized components from McMaster Carr. A clamping handle, square shoulder rod end, thrust rated swivel rod end, marine grade quick-release pins. All up, about $90.
Controller board on the left, battery balancer on the right. This controller board worked perfectly, but suffered irreparable damage when a motor driver caught fire, shorted out, and forced back 48V into the controller. The battery balancer ended up not getting used, since lead-acid batteries automatically balance provided they are "overcharged" occasionally.

I used Eagle CAD to enter the schematics and layout the boards, and Advanced Circuits manufactured the prototype PCBs. Components came direct from the manufacturers when possible, or via Allied Electronics or Digikey otherwise. Normally, I would not have made so many boards, but would have tried to use off-the-shelf hardware. In this case, however, I wanted to try my hand at design and surface mount tasks, so I took the plunge. Kind of expensive, but worth it for the hours of amusement.

Rough layout and sheet cutting plan. Since marine-grade plywood is expensive, it pays to plan the cuts meticulously. Enough was left over for the fenders and a box for the remote control as well. Dimensions in mm.
Rough plan for remote control box. Not very professional, but only made a few minor changes during construction. Good enough for a one-off.
Trailer hitched up to a Tri-Sled Gizmo. 
In action. Picked up a small passenger who insisted on going along for a ride, regardless of the lack of basic amenities (like a seat).
Another view. This time, tooling along the sidewalk. 
Packing up. The hitch arm is inserted and locked into a pocket on the top surface. One day, I'll bolt a shopping cart basket on it, so I can take merchandise direct from the supermarket aisle to the refrigerator.

Results and Summary
So how did it turn out in the end? Well, it certainly works. If I was to make another one, though, I would lose the reversibility. It turns out that reversing this rig is an extremely difficult skill to acquire, since the hitch point is right on the rear axle. This means that the trailer is almost uncontrollable once it starts veering off-course. Reversing a trailer on a car is not so difficult, since the hitch point is aft of the rear axle, and thus responds immediately to steering corrections. With the trailer, it is impractical to reverse more than a few feet. Thus, it would be better to simplify the circuit and just require the rider to get off the bike and pull!

The next simplification would be to parallel the wheel motors, and use a single 50A driver. This is because the initial objections to this hookup mode are simply not a problem in practice. In fact, parallel motors give the equivalent of a "limited slip differential". One simply does not turn corners often enough for the inefficiency to be noticeable, and in any case the normal power loss in the armature resistance is far greater.

Finally, the ability to regenerate braking energy sounds nice, but it turns out to be not worth doing for just that purpose. However, as it happens you can get this for "free" with a modern, efficient, driver design, so it certainly doesn't hurt. On my trailer, there are no mechanical brakes, but the motors give very good deceleration if you use them as generators to recharge the batteries. In the normal course of events, I never actually use mechanical braking. And, it's a good feeling knowing that at least some energy is going back into that battery.

During initial testing, where the motor braking was not working, I found that the twin front disc brakes on the trike were not really powerful enough for a quick stop. They are very effective without the trailer, but the additional trailer weight overloads them. Thus, motor braking is recommended for taking most of the load off the mechanical brakes.


Overall, this has been a very rewarding project. There were plenty of moments of heartbreak and desire to throw the whole thing in the wood chipper. One such incident now seems funny, but tragic at the time. I had just completed everything, and had for the first time experienced the thrill of being pushed by smooth electric power. I was in at my work office, since there is a whole lot of test equipment available. I wanted to try it out in the great unrestricted outdoors, but to get there I had to take an elevator downstairs. I unhitched the trike and the trailer, wheeled the trailer to the elevator, and went down, leaving it just outside the elevator in the building lobby. It was a Sunday, so nobody was around, and the entrance to the building was locked. I went up again to get the trike. This time, I thought it would be quicker if I just carried the trike down the stairs. So I picked it up. Unfortunately, the remote control was active, and at that time I had not put a friction lock on the throttle. I accidentally brushed the throttle control to full forward thrust! Out of sight, but within earshot, there was a sound of spinning tires and scooting, pilotless, 100lb trailer. Too late, I realized what had happened. The next second, bang! It impacted the drywall mere inches from the glass door leading to the outside world. Stupid, stupid, stupid! There was a gaping hole in the wall. I wasn't going to ride anywhere that day!

On the positive side, with every self-made disaster there's a very good lesson to be learned (ummm... put more corner reinforcement at the plywood joints, get a brain, etc.).

As far as the ride is concerned, it lives up to expectations. It's even a bit scary how fast it goes along those bumpy county roads. Some of the ride in to work I have to back it off a little to avoid getting pinched tubes and punctures. Rail crossings are the worst.

One annoying fault of the design is that it has rigid suspension. The only "give" in the suspension is in the tires. Everything else is so stiff that it has very little natural damping. Thus, it tends to bounce along behind. The knobby tires howl anyway, but with the oscillation, it sounds like a rapid woo woo woo woo... The cure for this would be to put proper suspension and shock absorbers. Mechanically challenging, so I'll leave that for when I get a welder and make an all-metal version.

The battery pack is more than adequate. The 13 mile trip to work uses about 1200kJ. The entire pack contains about 1kWh, or 3600kJ. Thus the estimated range is 39 miles, although I've never actually taken it that far without recharging.

At full speed, on the flat, it draws about 800W. On the way home I usually drop that to 600W and do a bit of pedalling, for health. Several times I've pedalled the whole thing more than 10 miles without power assist. The added load means that I use the next smaller chain ring with the same cadence. Obviously, that means a longer ride, but at least it is possible and in an emergency (unlike a car) you can get home without calling for help. One time I had to pedal because I had to go home in heavy rain. The trailer itself is fairly water resistant, but the remote control flaked out when water got in the keypad. A plastic bag over it would have prevented that problem, but nevertheless this is intended as a dry weather vehicle.

Stephen J. Hardy
Questions? email Stephan: mean_taipan at yahoo dot com

Back to the electric assist recumbent page