Strain gaged push trailer

I have control issues with my noob project. To add some power to flatten hills I intend to build a trailer with the following features:
- Probably based on the AZ one wheeled Cyclebully
- Easy transfer between different (non powered) bikes.
- Some cargo carrying capability
- Strain gages in tongue measure push/pull to provide primary control signal so:
- Different assist modes from null (carries trailer load) to several extra power modes (especially for hills).
- Regenerative braking to reduce brake pad wear

Basically I want to alter motor current to keep the tongue pull at zero. That would be the null mode. Other modes would have an extra percentage to provide positive assist to the rider.

I thought a Cycle Analyst 3 could help with the control. But I am not sure now. The TorqPAS input produces an output based on the absolute value of the torque measured. For my case I need the output to be based on the present value plus the value in the strain gage. Any thoughts.

Thanks.
--Ray

Would you use some kind of pre-tensioned hitch?

sounds like a great idea, but I can foresee a possibility of ever increasing oscillations causing the trailer to buck like a mule. So at the very least, a quick kill switch like a brake switch on typical bikes.

and uhhh,,, if you do have wires running from the trailer to the bike,, why not just use the fastest reacting, finest tuned computer on hand to control it all. Your brain, connected to your throttle hand. Or,, just PAS.

I'm not scoffing that much, the engineering will be more fun than the ride. But sometimes KISS is good engineering.

Hey guys, thanks for your ponderings.

Gogo, I am npt aware of the advantages of pretensioned hitches. Instead od a ball hitch, I was gong t use a hnge style so that my one wheel trailer cannot fall over... until I do.

Dogman, I expect there will be some entertaining moments during calibration. Kill switch for sure. The trailer will be used on at least two bikes by two people. One ideal scenario would be for trailered bikes to feel exactly like they are when alone, except when steep hills are encountered. Then additional assist could be selected to enable climbing steeper hills without requiring dismounting to walk.

My latest thinking is to use the thrust signal from the tongue (processed strain gage value) as a feedback for an op amp to adjust the existing torque value going into a CA 3.

raystl said:

Instead od a ball hitch, I was gong t use a hnge style so that my one wheel trailer cannot fall over... until I do.

Click to expand...

You might look into BOB trailer hitches.

Great idea. BOB it is. Thx.

You might really need some kind of signal conditioning to get a good response. Oscillations will be hard to avoid with a simple feedback loop. When pedaling, there will be a natural variation in force with each pedal stroke. But you rally want the motor speed to be constant over this time frame. At the same time you need quick response for emergency stops. Maybe a job for an Arduino.

If the motor controller had super fast throttle response, a simple feedback loop might work.

I expect signal conditioning will be the hardest thing to get right. My hope was that there would be existing conditioners that I could tweak to get something appropriate... I don't want to reinvent the powered wheel if I don't have to. I was hoping a Cycle Analyst 3 would fill that role. Maybe there is something else that would be better for me.

I had hoped to use an ebrake signal when I pressed the rear brake caliper to instantly go into regen mode. Because one hill regularly used is challenging to stop on with my pedal bike now, regen is mandatory with the trailer.

PS- Like your signature quote, fechter.

Well, there sort of is,, your brain, your throttle hand. I just come from riding bikes and motorcycles, so a throttle is like riding a bike to me. You never forget how.

Tying the motor on the push trailer to a torque sensing PAS makes the most sense to me. Yer talking just a few wires, but both bikes would have to have the torque sensing bb unfortunately.

One of the bikes could be just a throttle though.

This will be used by pedal cyclists to allow them to climb hills that were beyond their capabilities before. The intent is to give an experience that is 'identical' their bikes without a trailer attached...until an evil hill is encountered. They will also be riding in groups of pedal only bikes. So the load sensor needs to be on the trailer (on the tongue).

On the other hand, I am now quite concerned about the possibility/probability of wheelspin going up hills due to the low weight over the trailer wheel. I agree this is where a direct throttle control could be easiest, but it drastically change the pedalling experience.

I definitely don't want to switch to heavy dead-acid batteries to get more weight over the trailer wheel. Hopefully running a wider tire at lower pressure will be sufficient.

Having used a pusher trailer for a time, I have a few observations. Wheelspin will be a significant issue if you're trying to push the bike, but there are ways to mitigate it. A seatpost-mounted hitch makes the trailer push into the ground under power (but in doing so, it removes weight from the bike's rear wheel). Low tire pressure reduces skidding. Balancing the trailer to minimize tongue weight and maximize axle weight also helps.

Of course, when you're not pushing the bike, wheelspin is very unlikely to occur.

I think if you program the controller for gradual power ramping (like a couple of seconds for 0-100%), oscillations won't be an issue. Whatever elastic and pilot-induced resonance you get should have a shorter period than the power ramp time, and the motor won't lash the thing around.

I don't think you will have to reinvent the electric wheel, but you may have to reinvent the throttle.

A strain gauge on a trailer tongue or hitch is likely going to have too much noise to be useful. While the bike pulling the trailer will have a fixed amount of strain, every bump, rock, and breeze is going to add stresses the trailer. All of that stress will be transmitted to the hitch arm in the form of strain and vibrations. So you're going to have a very high noise environment for the gauge to try to work in.

I'm not sure how to eliminate that while keeping strain gauges, short of a lot of strain gauges and some fancy computing algorithms.

An alternative might be to go for something less fancy. A sliding sensor could be made out of throttle parts, and would give higher fidelity to the strain signal from the hitch. Build your trailer so that the hitch it's self is on an arm that can slide in and out of the trailer arm. ass springs and stops to give it, say, 3 inches total movement. 1.5" in compression, and 1.5" in stretch. add the hall sensors and magnet from a conventional ebike throttle, and possibly an ebike brake lever, and you would have something that could be used with a Cycle Analyst to achieve your goal.

And if you added an Arduino to that, you could add other features like a drag brake feature for long hills, regen braking to stop, or emergency braking if you hit the brakes hard on the bike.

Drunkskunk said:

I'm not sure how to eliminate that while keeping strain gauges, short of a lot of strain gauges and some fancy computing algorithms.

Click to expand...

Maybe look into JustinLE's skateboard thread and see what he did to deal with using strain gauges on the trucks for board control? I'm sure that has to deal with some of the problems a trailer might.

raystl said:

I have control issues with my noob project. To add some power to flatten hills I intend to build a trailer with the following features:

- Strain gages in tongue measure push/pull to provide primary control signal so:
- Different assist modes from null (carries trailer load) to several extra power modes (especially for hills).
- Regenerative braking to reduce brake pad wear

Basically I want to alter motor current to keep the tongue pull at zero. That would be the null mode. Other modes would have an extra percentage to provide positive assist to the rider.

Click to expand...

This is a pretty neat concept and approach to doing an electric assist trailer. The proposal of sensing the force on the hitch and having a feedback loop neutralize that by increasing/decreasing the motor torque is certainly doable, but it's not a noob type of electronics project. You'd need to be pretty comfortable hooking up and wiring sensors and as Fechter mentioned there are lots of avenues for feedback oscillation in a system like this. In principle you'd want to integrate the force signal on the hitch and have that be the throttle/torque command going to the motor controller. Then the throttle would move up and down to whatever value results in zero net force on the hitch, and to have a steady offset for having a net push or pull you'd just add an offset term to the equation.

The V3 Cycle Analyst could do that with a custom firmware, but the stock firmware doesn't have any naturally integrating inputs. I did experiment at one point trying to do something very similar of having a PAS assist mode that would do its best at forcing the riding to maintain constant pedal power output (so it would integrate the difference between the target human watts and actual human watts for determine throttle output level), but scrapped it in the end since it didn't work nearly as well as I'd envisioned for an ebike PAS mode.

Anyways there is another approach which might work just as well at achieving your goals and could be made with readily available parts, and that's to use an inclinometer sensor rather than a strain gauge sensor on the hitch. By sensing the %grade hill that the trailer is climbing, you can quite linearly scale that into a motor torque which will perfectly cancel out the gravitational drag of the trailer. It does mean that as you add/remove weight from the trailer you'd need to adjust the scaling factor, but this can all easily be done via the Aux input of a CA3 device, and while riding you'd want a control on the handlebars anyways to increase/decrease the assist provided.

Let me know if that's an approach you'd consider pursuing and I can give some more pointers on how it could be implemented. As a starting point though you'd want to go to sparkfun, adafruit, or similar shop that would have an accelerometer mounted to a breakout PCB that is set to provide a 0-5V voltage signal based on the incline of the circuitboard. If you have a torque throttle motor controller like a Phaserunner or Adaptto, then you could do this all without any CA in the system. Just use a potentiometer to adjust the gain for your weight compensation and wire up the inclinometer output to be the throttle signal going into the controller.

Thanks all for your ideas. Looks like I've got or a lot more pondering to do.

Chalo, I will be mindful of your observations on preventing wheelspin. The seatpost hitch is not an option for me since my recumbent does not have one... BOB hitch points will be used.

But the weight shift principle got me thinking of load equalizing hitches used when pulling large trailers behind cars. A similar effect should be possible by running mounting 'springy' bars that bridge from behind the BOB axle mounting points on the hitch to the rear forks on the bike.

Drunkskunk, my original plan was to use some sort of sliding tongue similar to what you described. For some reason I became convinced that regen braking was not going to be possible so I went to the stiff tongued strain gauge approach. Perhaps that was before I needed an explicit ebrake signal. Time to revisit that approach.

The more I think about it the more I agree with everyone who said strain gauges would be difficult to use here. I am wanting to operate trying to keep the gauge force near zero while all sorts of bumps and other irregularities would be constantly giving a high mechanical noise value. As Justin says, an accelerometer/inclinometer could be used to measure the same thing thrust&incline (F=ma) and, since it could be mounted on a compliant backing, would be much less sensitive to high frequency and random chatter.

Lots of good ideas to ponder. Thanks.

justin_le said:

It does mean that as you add/remove weight from the trailer you'd need to adjust the scaling factor,

Click to expand...

The scaling factor could also be automatically adjusted for the weight of the trailer by adding a weight-measuring scale into the trailer itself, though I expect that could also be complicated. :/

If I understand correctly, Justin, you are essentially recommending a system that models the torque requirements of the trailer and then use the throttle or a pot to do fine tuning on the fly. I can see where that would be more stable than a system based on measuring signals around a zero net thrust.

If so, I can see the system working well on grades and for low speed acceleration. How would rolling and air drag resistance effects be included at higher speeds where they would become significant? Hmm, maybe I can add an input from an air speed indicator :wink: .

I did visit adafruit to see the goodies there.

Wind resistance starts to really impact performance around 20MPH, but it gets noticeable around 15MPH, when I'm using wider two-wheel trailers with boxy loads on more normal bikes.

If it's a one-wheel trailer with a low profile it's probably not going to have much (if any) more aero impact than the rider on the bike itself will, on an upright.

On a 'bent it depends on the aero of the 'bent itself and trailer's relative size/aero compared to it.

You can also put a shell around the trailer to make it more aero, probably enough to not really worry much about it's impact at cycling speeds.

raystl said:

If I understand correctly, Justin, you are essentially recommending a system that models the torque requirements of the trailer and then use the throttle or a pot to do fine tuning on the fly.

Click to expand...

Exactly. We actually built a prototype ebike PAS sensor setup like this about 8 or 9 years ago and it worked really well, and I've been meaning ever since to make a product around that (PAS sensor pickup with built-in inclinometer) but there are always way more projects to do than time to pursue. Another detail that's not obvious at first is that the inclinometer/accelerometer also responds to you pedaling harder to accelerate even if you aren't going up a hill, and so it will naturally give an extra boost when you are trying to speed up the same way that a strain sensor would too.

If so, I can see the system working well on grades and for low speed acceleration. How would rolling and air drag resistance effects be included at higher speeds where they would become significant?

Click to expand...

Rolling drag is a steady offset, it doesn't increase with increasing speed. The effect of air drag on the trailer itself will partly be masked by the fact that the trailer is basically drafting your bicycle and is itself not contributing that much additional wind resistance. But in any case you'd still have a dial or up/down button on your handlebar to increase / decrease the gain setting if you find you want a bit more or less power from the motor while you are riding. I wouldn't sweat too much about trying to make it perfect at always exactly cancelling its own drag.

Another, cruder approach would be to use "mechanical signal conditioning". If the hitch has a spring-loaded sliding section that actuates a linear pot or hall/magnet setup, you could add a hydraulic damper to slow down the response. The damper could have check valves for different rates in acceleration vs. deceleration. Sort of like the setup on a fancy rear shock absorber. There would be a lag when you first start out and pedaling would be harder, but it would gradually catch up. A variation of this would be let the sliding section be un-damped but have a springy connection the linear pot and use a very small damper just on the pot slider so it always moves slowly.

Continuing with the idea of roughly modelling the trailer loads, the motor would have to overcome, the steady speed loads, plus acceleration and grade loads. Steady speed torque load would consist of rolling drag and air drag. Rolling drag would be constant with speed but increase with weight. Air drag increases with the square of speed. Because of its small size and sheltered location, air drag of the trailer should be lo relative to the bike with rider. So accurate modelling of the air drag might not matter.

Regardless, it looks like the throttle input to the controller needs to be the sum of at least two effects the acceleration/grade load with the steady state load.

So I will need hardware for the accelerometer/inclinometer plus something to provide the steady state load simulated portion. I think it is time to bite the bullet and use something like an arduino to model the steady speed losses, read the processed accelerometer/inclinometer signal, then output the sum of the two values as the throttle signal.

I'm thinking additional assist torque should increase with speed. That would be more comfortable to se than just a simple offset. I am thinking using current throttle signal would be easiest to emulate. Using a CA after the arduino might male sense to keep that well behaved. If so, not sure whether to add the assist offset to arduino duties or use the aux pot on CA.

Oops. It appears I missed the last comments from Amberwolf and Justin when I sent my previous comment.

So I will take their advice and stop fussing about air drag. That means that the system boils down to a simple inclinometer/accelerometer with an on the fly tweak 'throttle'... as Justin originally suggested. Justin, I would appreciate any of the implementation pointers you had mentioned earlier.

I am assuming an appropriate accelerometer package would be involve sensor board and Arduino board or equivalent. Output from that could go into a CA V3 or directly into a torque throttle.

Although I want to get the trailer working soon, I expect to be tweaking it for some time to 'improve' it. For example, after using it for awhile I will have a better idea how much wheelspin is an issue. If so I expect to be getting to get to know the Arduino more intimately.

The Adaptto controller is interesting. It can handles wheelspin. Not sure how suitable it would be for the sub 500 W powers I would be using since the smallest Adaptto is 4 kW. It is also an expensive solution.

I will be using a 20 inch wheel and want a 500 W motor. That should be more than enough power for this project, especially since wheelspin will probably limit what the wheel can use.

Not sure what to use for a motor. I was looking Magic Pie Edge motor, partly for its internal controller and regen capabilities. But a torque throttle would be out.

If I used an external controller that widens the choices to something like Crystalyte SAW20 with a Phaserunner. That costs more but gives me a torque throttle.

Any other thoughts on what I should be considering?

raystl said:

I am assuming an appropriate accelerometer package would be involve sensor board and Arduino board or equivalent. Output from that could go into a CA V3 or directly into a torque throttle.

Click to expand...

Exactly. If you're familiar with arduino's or other microchip platforms you can do all your signal processing there. Otherwise though, I would recommend trying to get a device that powers directly from 5V and has a useful signal range in the -0.2 to +0.2 g range like this Murata part here:
http://www.mouser.com/ProductDetail/Murata-Electronics/SCA61T-FAHH1G-1/?qs=ez%252bYtvyjdJJE48NnxcjtuA==

Then you can go directly from the output of the inclinometer to either the CA3's torque sensor or throttle input lines or use it as a straight up throttle signal for a torque based motor controller like the Phaserunner or similar. A lot of the more common breakout boards for accelerometers are like +- 2g or higher (like this here https://www.adafruit.com/product/163), so the signal range in the area of interest will be difficult to measure. A fairly steep 5% slope is equivalent to 0.05g, so on that adafruit board you'd have 1.65V on level ground and 1.6725V up a 5% hill, an additional signal amplifier would be required before you can send that to a controller or CA device. But on that Murata part every 10% of slope corresponds to 0.4V of signal change which is a usable amount.

raystl said:

For example, after using it for awhile I will have a better idea how much wheelspin is an issue.

Click to expand...

From our experience wheelspin on electric push trailers only really happens if you are riding bumpy gravel roads where the trailer wheel can get airborne, and that's in situations where the trailer drive wheel is doing much more than just pulling its own weight. So I think you'll conclude that this isn't much of a concern.

I did not realize that there were sub 1 g acclerometers. That simplifies a lot. Found some electrolytic tilt sensors which might be suitable such as http://www.frederickscompany.com/product_category/tilt-sensor-assemblies/

To allow more flexibility, have decided to feed the processes inclinometer signal to a CA3 as a torquPAS value and use auxpot to vary the assist. Since I need torque from zero speed on a steep hill, I will need a fake PAS signal.

Being mindful of the intent of the min PAS speed threshold to prevent runaway bikes, I will add a deadman throttle (separate button?) press requirement when below a threshold speed.

I will probably use an Arduino to provide the fake PAS, deadman throttle handling and to handle everything I else can't think of now.

I have finally finished the design and have ordered the electrical components. It uses an off the shelf inclinometer with a Cycle Analyst 3 and some custom trailer processing module circuitry. It will (hopefully) power the trailer to normally have a null effect on the bike. It will have proportional regeneration. The aux pot of the CA3 will scale the torque applied to provide bike assist, especially for steep grades... and also to just be able to wheee.

The DOG2 inclinometer (http://www.te.com/usa-en/product-CAT-TSI0007.html) has a +/- 25 degree range with a 4.5 volt maximum voltage. It provides the acceleration/grade load signal for the trailer load processor. Since rolling and acceleration load both vary with weight, rolling drag can be handled as if a small additional grade.

The circuitry consists of one quad op amp chip, one diode, one potentiometer and some resistors. These will be set in an Altoids Small tin. The trailer module will send a torque signal to the CA3 operating in pedal assist mode.

The Ebrake signal will combine with the torque signal to make a throttle value that is always zero when the brakes are not pressed. When the brakes are pressed, a deceleration or downhill grade resisted will be converted to a throttle input for the CA3. The Ebrake signal automatically disables pedal assist operation.

While under power, the auxiliary pot will scale the torque upward required to provide more assistance.

That's the theory. All electrically related parts have been ordered. While waiting for them I have some metal bashing to do.