My friend and I are building a small direct drive helicopter. We both have a background in mechanical engineering and we found it very difficult to obtain good specs on brushless motors and propellers. Some websites provide a results database, but these databases are frequently for a combination of both motor and airplane propeller, which makes building something different very difficult. When we started rewinding motors, we built a system to measure torque, current, rpm, voltage and temperature. With torque, we can actually completely decouple the motor and propeller in the results, which really speeds up the development and allows calculating ideal gear ratios or find the correct propeller, in order to maximize efficiency.
The code is based on a modified MultiWii system with a series of connected sensors. Currently, the tests are entirely automated to test different pitch and torque combinations. There are safeguards in the code to stop the tests based on measured temperature, rpm and current. The code automatically generates graphs of the results.
That got us thinking. Would people be interested in a community-based website with motor and propeller specs? Our code is already open source and we could make our test rig available for purchase. The rig could be sold on kickstarter for approximately 150-250$ if we sell 40-100 copies. We don't expect this project to generate money really, but the goal is to help the community obtain better motor and prop tests. The system would include torque, thrust, current, voltage, rpm and temperature sensors. The instrument would include a mount for most standard motors. People with test rigs could upload the motor and propeller results on the website and specify the brand, spec or custom winding, etc.
Below are a few more images of our tests. The helicopter is a modified FBL100 and we tested quite a few motors and propellers. Our tests demonstrated that it is very difficult to make a small and efficient direct drive helicopter. It flies though.
If you are interested, please reply to this survey!

Views: 3720

Comment by Chris Paulson on January 15, 2015 at 4:08am

The noise seems quite high in the motor torque direction. Can you elaborate on how you make the measurements? A little smoothing may help the plot. Also, I assume these are static tests? Would it be possible to mathematically model an estimate of dynamic thrust efficiencies? 

Comment by Charles Blouin on January 15, 2015 at 7:16am

@F1P  You are right, we should have specified, the tests were done at a constant voltage.

@Hugues The max efficiency is out of the graph. For the blades we tested (we tested about 10 different types, with length of 5 to 20 cm, symmetric and asymmetric), the peak efficiency is at very low thrust and RPM (under 30g force). We were designing for a 175g helicopter, so our graph is centered on the relevant part.The colors also get very difficult to read when including the very high efficiency at low RPM.

@ Adam Kroll The scale in the first graph is motor efficiency, unitless, from 0 to 1.The second scale represent the thrust per unit of power, in g/w. Quads usually have real life performances of 5 to 12g/w. It would certainly be useful for aircraft endurance, but mostly for the motor. The prop on aircraft has to be tested with airflow, although a static test can still serve to compare propeller profile.

Comment by Charles Blouin on January 15, 2015 at 7:31am

@Chris Paulson You are right about the noise. Since the tests were automated, we could afford to measure a very high number of points, so artifacts show quite will. We could run the tests a few times to do a statistical analysis. Think some of the noise may be due to the backlash in the servos and swash plate, which caused imprecise pitch.

Our test parameters (different pitch and input combination) were all in a spreadsheet. Our program opened the spreadsheet and looked for tests not performed yet. That was useful as we could easily add test points in a zone of interest without redoing the whole test. Having integrated protection in the code also allowed us to run the test without close monitoring, as the program would skip test points that required current or temperature too high. We even had the helicopter detach of the test rig once and the test stopped as the rpm detected was abnormal.

Comment by Nathan Asdourian on January 15, 2015 at 12:14pm
Excellent job! I have been wanting to do something like this but primarily for checking whether a motor sustained damage in a crash or hard landing and should be replaced. Though with a full range of sensors it should be fairly easy to take it further and show if and where the performance has been affected, either from an event or even just use...however obviously for an ongoing diagnostic it would need to maintain the stats over time. ..
I would be more inclined to build it myself maybe with the option of purchasing any custom or difficult to locate parts from you guys, but if that were not an option I would be willing to buy a full setup. ..
Comment by Brian Martin on January 15, 2015 at 3:17pm

Great Job!!  Like many other who have commented on this post i also have wanted to do this.  I have always viewed a multirotors drive train as 3 separate and discrete systems (control volumes if you will).  

  • The first CV is the ESC their is obviously a power loss through the ESC but given the 3 phase output i am not entirely sure how to measure it (maybe some EE's out their can help)?   
  • The second CV would be the motor and we know how to quantify this.
  • The third CV is obviously the propeller (its efficiency is heavily dependent on the control mediums (air) properties and conditions (ie incoming flow speed, etc)) 

But anyway as i said before i also wanted to build something like this (you guys beat me to it :-)).  What sort of dynamometer did you design; inertial, prony brake, or some other type?  My preliminary design will use an inertial dyno with a hall effect sensor to measure speed with an arduino but i have not built it yet.  I would be very interested to learn more about your project, and sign me up for one of your testing rigs.  


Comment by Brian Martin on January 15, 2015 at 3:40pm

oops i just read your first comment your dyno sounds very similar to a prony brake.  I see that you have to wait for the load cell to settle.  You could link the BEC of your ESC to a micro-controller (like arduino) and write a script ramp-up the throttle response very slowly so your load cell has time to settle.  I'm sure they make fast response load cells, maybe would have something.  A carefully designed inertial dyno i think would yield smoother results although their is some data that you cant get. my 2 cents :-)


Comment by Charles Blouin on January 15, 2015 at 7:51pm

@Brian Martin Thank! I would add that the design for the ''control volumes'' is also even more difficult as the propeller needs to be match to the motor.

Right now, the whole motor and propeller assembly is mounted on bearings. There is an arm attached on the motor mount that pushes on a load cell that looks like this . On the load cell, there are small strain gauges and the geometry does a really good job at measuring only the force in a single direction. The arm length times the force is the torque. We calibrated it with precision weight, and it is quite precise and linear. I will look more into the response time of the load cell.

Comment by Greg on January 15, 2015 at 7:54pm

I'd buy one on Kickstarter. Just make sure it can handle big motors and props (think U10, 30" props).

Comment by Brian Martin on January 15, 2015 at 11:06pm

@Charles Blouin Thanks for the response.  Yes i have some experience with these load cells.  I think i understand your set up so you are measuring the torque at the motor shaft but you also have the propeller attached.  Is this correct?  

If so then what you have is a chassis dyno because you are measuring the torque that the motor delivers to the propeller, but you are still measuring the motor propeller system.  But i thought the goal would be to characterize the motor, by generating a proper torque curve over its entire speed range.  Then reattach the propeller and conduct the same test (the dyno now becomes a chassis dyno not an engine dyno).  Then since power = torque x angular speed you can simply multiply torque and speed at every discrete measurement point to obtain the power curve.  

You do this for both test that you ran

  • motor only test
  • motor prop system test

then we would have 2 power curves as a function of angular speed.

  1. Motor only power curve 
  2. Motor propeller system power curve 

Subtract the 2 and that should give you the propellers power curve as function of angular speed.  If you add a load cell for measuring thrust for the second test (or you have a corresponding thrust vs angular speed curve for the test prop.) then you can create a thrust vs engine power curve.  Then using the motor curve and prop curve (assuming there was a database for all of these) then you could select the best motor for a specific propeller or vice verse.  Its worth noting that that propeller curve would only be valid at the test altitude temperature and pressure, and for zero free stream air velocity (assuming that's what you conducted your test at).  Like you guys i am a mechanical engineer too but it has been since 2009 since i dealt with this stuff so my logic could be flawed.  Time to dust off the old books ;-)

Anyway i feel that this form of testing is necessary to further advance the efficiency of multirotors (trial and error only gets you so far).  Keep us all updated on your progress with this and if their is anyway that i can help just let me know.


Comment by Charles Blouin on January 16, 2015 at 12:18pm

@Greg thanks!

@Brian Martin That is interesting. Testing the motor without load was mostly useful to obtain the kV now. The motor speed is function of the input and the load torque from the propeller. We could change the load on the motor by changing the pitch of the helicopter head. It would also be possible to test different load with fixed pitch propellers by just changing them. We are definitely working on an improved version now, and we will post an update in a few weeks hopefully. We want to build this database you are talking about.


You need to be a member of DIY Drones to add comments!

Join DIY Drones

© 2019   Created by Chris Anderson.   Powered by

Badges  |  Report an Issue  |  Terms of Service