4

I'm designing a series of small drones for use in a swarm configuration in combination with terrestrial / marine / submarine robots. As part of the process, I've designed a test rig to measure the thrust performance of various motor / propeller combinations. The general layout of the rig is shown in the Sketchup diagram below:

Test rig layout (click to enlarge)

The motor mount is attached to a plate, which is bolted to a MGN12C linear bearing block mounted on a 100mm rail. (If anyone is interested, the details of the MG series linear bearings can be downloaded here). This particular bearing was just one I happened to have available.

At the other end of the plate, a load cell protrudes through a slot, and this is used to measure the thrust force generated by the motor / propeller combination. The mount is designed to allow the load cell to be easily changed (I have quite a few 1kg, 3kg, 5kg, and 10kg load cells).

The data, together with temperature and vibration measurements from an accelerometer, will be gathered using an Arduino (or possibly a Raspberry Pi if I decide to integrate the rig with other test gear).

The length of the plate means there is ample room to mount the ESC between the motor mount and the load cell.


It occurs to me that it would be fairly straightforward to modify the rig to allow the twisting force, or torque, generated by the motor / propeller combination to be measured. Probably the simplest way with the parts I have available now would be a pair of load cells fitted between a modified motor mount and the plate attached to the linear bearing as shown here:

torque measurement rig (Click to enlarge)

Minimal modification to the electronics would be required to add this to the design, and I have more than enough load cells available to add this to the rig.


We had to re-design arms in a previous model when we found the forces generated by the motors caused them to deform while testing the unit on a static rig, which is why this modification appeals to me. However, I wonder whether anyone else has any experience of using this kind of data in drone design.

So, how useful is it to know the torque generated by a motor prop combination when designing multi-rotor drones?

1

I am a fan of "More information is better". I believe that torque can be used to calculate system efficiency. I just do not know what those calculations are. I have seen one commercial thrust stand like your concept/design that includes one or two load cells that measure torque. I do not believe that measuring torque would be included in that thrust stand if there was no merit in it. Look at this similar thrust stand which also comes with data collection and reporting software. It is linked to a computer through a USB: RC Benchmark Thrust Stand It does not use a linear bearing. I like your concept very much.

I encourage you in your quest for a deeper understanding of the entire performance characteristics of your drone design. I would have a use for a sophisticated thrust stand like yours.

| improve this answer | |
  • 1
    Thank you for that. I hadn't seen that unit before, but their website has a lot of useful background information. Based on some of that, it looks like we really may need to capture that data to help us with the designs for the algorithms for our swarm controllers. This is really helpful information. – sempaiscuba May 24 at 18:56
3

Not very useful.

If you're designing at the absolute limit of what a frame can handle, such that it's as light as possible, then maybe you'll want to know the maximum torque, but you should be designing for orders of magnitude more strength than that (It's not hard, small BLDC motors aren't very torquey).

There are control algorithms that take into account frame torque and how the forces are applied. I forget off-hand, but it was a matrix-solving control algorithm, as opposed to PID. Unless you're writing your own flight control software, that's not an issue.

When changing motors, it may give minor insight into tuning the yaw axis (these motors have more torque, so less P-Gain may be needed), but yaw is done by applying torque to a propeller, so again unless the motor can barely spin the propeller (this should not be the case in a design), then this isn't a concern.

Where you may see an advantage is when computing efficiency - ie how much torque do I have to apply to propeller x to get y grams of thrust? However, this is generally done by using current figures since those are more relevant to power system design.

Another possibility is how much braking/accelerating torque a given ESC can provide - high performance machines needs rapid changes in motor RPM, so a given ESC may be better (due to FET latency, resistance) at control algorithms than others. This would be interesting data to see, but I don't think any ESC is necessarily better than another, since braking is just "FETS on" and accelerating is just pushing the control algorithm to its max. Noisiness is much more important than acceleration torque in practice.

| improve this answer | |
  • All good points. However, information from the RCBenchmark web page (linked from the other answer) suggests there are quite a few potential uses for the data in our use case. (For example it looks like it may be of particular use when designing our swarm flight-control systems) – sempaiscuba May 25 at 3:04
1

The torque a brushless motor is capable of producing is particularly important when dealing with control loops. The torque is directly proportional to the rate of change of RPMs, which is directly proportional (though non-linear) to thrust. How quickly the motor-prop combination can make adjustments has a massive impact on overall stability. That may or may not be necessary depending on how agile you need your swarm to be, but in larger applications where stability and agility are key, it's a very important factor to know when optimizing power systems.

That being said, measuring torque directly may not be specifically necessary if you have high enough resolution in the rest of your data. When I was setting up the test equipment for Mini Quad Test Bench, I abandoned the complexity of multiple load cells in favor of higher resolution RPM, voltage, and current data. By measuring those values at a very fast rate, I can detect the rate of change directly which is actually the more critical information than simply the torque number. I also found consistency problems with using linear motion of any kind and ended up mounting the prop vertically on a steel arm above the load cell with no mechanical interlinking. I've been able to achieve extremely repeatable results with essentially zero drift. My calibration numbers haven't changed in years of heavy testing, and we were able to replicate the stand and match the data at a second location. Unfortunately the inexpensive load cell amplifiers I'm using aren't capable of more than about 4ms read rates, so that data lags behind the rest of my data. If you get a more expensive amplifier though, you can up the resolution of that data as well, and really get a good picture. This is where the RCBenchMark setup falls down. I got one, but ended up not using it, as the data was simply too low resolution to be useful for what I needed.

For more information on our setup, and the steps we went through to get there you can check out the equipment page over at MQTB. The setup is fully open source, and all the code and list of hardware is available on GitHub.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.