Hydraulics

[DavidEllzey]

I have a unique situation and would appreciate any insight offered. We are developing a full simulation of pedaling a recumbent vehicle to be used for fitness and rehabilitation. This includes VR and a 6DOF motion platform but where we are putting in most of our work right now is the pedaling and steering simulation. We currently use smart servo motors (with gearheads) in torque mode and a custom control board that our software is communicating with as a plugin to a game engine. It works great except it is expensive and requires a lot of design compromise to have a robust enough drivetrain from the pedals to the gearhead.

The idea of trying out small or micro hydraulic motors has come up but none of us have any experience. Before we go out and start talking to potential vendors I'd like to get up to speed.

Question 1: Does it seem feasible to have a small motor, say about 5cm dia X 20cm length capable of resisting counter rotation up to 30nm?
Question 2: I assume we'd use some sort of servo control for the valves, what kind of response time are we looking at?
Question 3: What are the leading companies in the field that I should be considering contact with?

Thanks in advance, Dave

[tician]

It is not really a unique situation as there are plenty of bicycle trainer systems out there that can provide ideas, and I expect a few of them already have real-time adjustable resistance systems that can be interfaced with certain computers and/or programmed with certain "routes".

Cheapest option is simply using a friction brake on a flywheel / brake disc and controlling the force applied to the brake pads to simulate the resistance in real-time. Scooter disc brake calipers are super cheap and widely available in several very common/generic designs, and it would be very easy to connect an inexpensive servo to pull on the control cable as needed (could series connect a cheap force gauge to get a more accurate read of the cable tension if using servo without built-in force/torque/current measurement).

My favorite option would be to use a cheap motor+gearbox solution (banebots P60 or P80 gearbox and motor) as a generator that feeds through an appropriate bridge rectifier into a boost converter to provide a simulated load. The boost converter would be operated in a current control mode based on the input current to create the desired electrical load on the generator which would then create a proportional mechanical resistance encountered by the user. It would be quite easy to add a scooter disc brake to the gearbox shaft for additional resistance at low speeds. The output of the boost converter can vary, but the converter I've been working on a bit for my PV system is supposed to feed into a 16S LiFePO4 pack with a few cheap >100W power resistors (low-side switched by a 100V/360A N-channel MOSFET in SOT-227 package) to burn off any extra power.


As for interfacing the resistance system to the pedals, the easiest and most flexible option would be to just use actual bicycle parts (bottom bracket and crankset) and it might also be among the cheapest if you use OEM, discontinued, and/or take-off items that sell at significant discounts. The only custom mechanical parts would be the bottom bracket mount, the chain ring to motor shaft adapter, and the frame that holds all the pieces together. As for sourcing them, bikewagon has an absurdly large catalog/inventory and plenty of 'bargain bin' items (mostly items discontinued by the manufacturers that they need to clear out of their warehouse); it is also where I am getting the maintenance parts (tires, tubes, brake pads) for my bike.

[tician]

To actually respond on the topic of hydraulics, hydraulic pumps and motors in the >200[psi] rating tend to be very expensive because they are 'positive displacement' devices that require very strong and very precisely machined components to ensure that as little fluid leaks between them as possible. Used/surplus positive displacement pumps/motors in the torque range you are looking at can sometimes be found for <$100 each, but are still thoroughly unnecessary for your purpose since you will never be actively pumping any fluid to counteract the user's efforts as that would always drive the pedals in reverse no matter the flow rate. You will only ever be adjusting how easily the user can pump fluid through the circuit by using a flow constriction device such as a needle valve downstream of the user powered pump. This lack of a pump/motor powered by an external electric motor or heat engine does decrease the cost quite a bit. However, a positive displacement pump typical of 'hydraulics' would additionally require a fluid reservoir, fluid filtering, and possibly a fluid coolant system to prevent damage to the pump/motor. It will also require regular replacement of the fluid and filter with a full flush of the entire system to remove any old fluid and/or particulate that escaped the filter. All together, the cost of a system using commercial/industrial hydraulics would likely be far more expensive than what you already have.

[jwatte]

If you want the cheapest possible solution that is robust: Buy an exercise bike and retrofit your VR and resistant control onto that.

[tician]

If you want the cheapest possible solution that is robust: Buy an exercise bike and retrofit your VR and resistant control onto that.

Not so sure about that as cheap and robust rarely seem to coincide.


Also: $1200 for almost exactly what the thread creator is trying to do - just grab a recumbent frame and replace the drive wheel with the trainer.

[DavidEllzey]

Sorry if I made it sound like this was just in concept, we've actually already completed a successful proof-of-concept prototype (https://youtu.be/wQsJHhuPa8U) you can see in this video. We are successfully using servo motors with gearheads along with a magnetic particle brake assisting the pedal simulation.

Dead torque vs live torque pedaling: The names we gave to drag resistance (dead) versus push resistance (live) are fairly descriptive and why we don't use use a friction device for our primary resistance. The particle brake doesn't start blending in until 6nm and its just a minor percentage of the resistance. The difference in how it feels to pedal live torque is not only profound but more importantly it feels like the real thing. We are quite happy with the simulation but are exploring possibilities that would allow us to simplify the drive train and cut production costs. Servo motors that run in torque mode are not inexpensive, not to mention the cost of the gearhead.

[tician]

Not sure about jwatte, but I very much understood that you had a proof-of-concept system using very expensive industrial servos and gearheads from a place like Anaheim Automation resulting in a gearhead+motor+driver cost well over $600. I did not initially assume you were building the system around a stock recumbent frame and drivetrain, hence my useless suggestion of cheap bicycle parts on a custom frame.

The only real difference between your 'live' and 'dead' loads is whether or not there is anything pushing back against the pedals. It is quite easy to make a 'dead' load such as a friction brake mimic the sensation of a 'live' load's push-back by adding an elastic member between the cassette/flywheel and the disc brake, then ensuring the brake is controlled such that the elastic member is always under load (locking and releasing the brake to maintain a certain minimum load). The illusion of push-back will only last for a certain percentage of a revolution when back-pedaling, but could be maintained indefinitely during a stop as long as the user does not back-pedal too much. A torsion spring would be easiest but probably not readily available in the load and spring rate range you would require, so an alternative is basically building a large diameter flexible shaft coupling and replacing the spider with compression springs.


If you absolutely must have a motor without gearbox, it is not too difficult to make your own coreless brushless motor using a 3D printed bobbin, litz wire, and permanent magnets. This method is common among diy wind turbine builders and permits you to make very low Kv (<20[RPM/V]) motors that would require little to no gearing to achieve the desired torque with a reasonable current rating (<40A). One nice write-up on an example in the radially wound form for an axial flux motor, although I generally prefer the axially wound form for radial flux motors for its higher wire packing efficiency.

Unfortunately, controlling coreless motors can require a controller using rotor position sensors and/or a rather high frequency control loop because the very low inductance of the stator permits very rapid changes in current. Or you can just series connect some inductors to the motor phases to slow the changes in current to let cheaper motor controllers safely drive the motor. No brushless motor controllers in the <$300 range offering true torque control out-of-the-box immediately come to mind right now, but there are a few motor controllers with current sensing that can be easily reprogrammed such as Brushless Rage.