Cheap Motor Torque Measurement / DIY Dynamometer Revisited

Technical Motors
1

Hello All,

I am writing this as a follow-up question to Mike Betts post on this thread:
https://www.chiefdelphi.com/forums/showthread.php?t=38539

Which Was:

(1) Make both the motor and generator the same type of motor.

(2) Wire a power resistor across the generator’s electrical “outputs”. Be careful to size the resistor properly.

(3) Measure the voltage and current at the “input” and “output”. Also measure the speed at the spindle/coupler.

(4) Adjust the input voltage to a known value (in our case it’s almost always 12.0V). Note that you do not need a precision high current power supply… Just use an Exide SLA and let it slowly discharge as you run the MG and take your readings at 12V.

Note also that you can ignore (4) depending on exactly the type of data you are going for…

For a given load resistance, you will take your readings:

Input Power (W) = Input Voltage (V) * Input Current (A)
Output Power (W) = Output Voltage (V) * Output Current (A)
Mechanical Power at coupler (W) = [Input Power (W) + Output Power (W)] / 2
Torque at coupler (N*m) = Mechanical Power (W) / Spindle Speed (radians/sec)
Power Lost in Motor (W) = Input Power (W) - Mechanical Power at Coupler (W)

This sounded like a great way to DIY at home so I mimicked this type of setup with a adjustable DC Load as my power resistor and I collected a bunch of data. Now I am going over the data I collected and I wanted to check back about the testing methodology I used as far as how well it should work out. That said I am not looking for something super accurate, but I was hoping to get reasonable numbers using this method.

In my tests I used 2 sets of different 550 series DC motors rated for 18v. Set 1 was identical and set 2 were identical to each other. These were motors approximately the type you may find in a 18v drill.

From my research to similar motors I expected these motors were rated for approximately 200-250w maximum with maximum efficiency to be at 6-10amps.

Below is some of the data I collected on the two motors. Does it look like I did this right? Knowing that the stall torque should be just under 500(mN-m) does the operating torque I measured seem right?

Any help, suggestions, and feedback would be truly appreciated!

Thank you!


Motor 1										
Input			Output			Coupler				
In V	In A	In W	Out V	Out A	Out W	M Power	RPMs 	Rads	Torq	P Lost
17.2	13.0	222	6.8	11.0	74	148	13100	1370	0.110	74
17.2	14.0	239	5.0	12.0	61	150	14200	1490	0.100	89
17.4	12.0	209	7.0	10.0	71	140	13300	1390	0.100	69
14.4	12.4	176	4.4	10.0	43	109	10600	1115	0.100	67
14.0	12.4	172	4.4	10.0	44	108	10400	1090	0.100	64
11.0	12.0	134	2.2	9.8	21	78	7400	775	0.100	57
10.0	11.6	118	1.4	10.0	13	66	6800	710	0.095	52
12.0	11.8	142	3.0	9.6	28	85	8500	885	0.095	57
17.4	11.2	195	7.2	9.0	65	130	13900	1455	0.090	65
										
Motor 2										
Input			Output			Coupler				
In V	In A	In W	Out V	Out A	Out W	M Power	RPMs 	Rads	Torq	P Lost
15.0	16.2	242	7.6	10.0	77	159	13800	1440	0.110	82
15.0	15.4	228	8.0	10.0	80	154	13900	1455	0.105	74
13.2	14.4	189	6.2	10.0	61	125	11800	1240	0.100	64
12.6	14.2	176	6.2	10.0	63	119	11500	1205	0.100	57
9.0	13.0	116	3.8	9.8	37	76	7500	785	0.100	40
12.0	13.6	164	6.2	9.8	60	112	11300	1180	0.095	52
11.0	13.6	149	4.8	9.8	48	98	9600	1010	0.095	50
10.0	13.4	133	3.8	9.8	38	86	8700	910	0.095	47
8.2	12.2	99	3.4	9.6	32	66	6600	690	0.095	33
2

I was going to build something along these lines a month or two ago, but changed tack because I realized that as this system is operated, the motor coils and resistors heat, increasing the resistance. My current plan is to spin up a rather heavy flywheel with an encoder to measure angular acceleration. This project has been on the back burner the past two weeks; I plan to get back to it next week. (I’ve been prepping arts and crafts for Vacation Bible School which ends tomorrow - this year’s projects include a tree made mostly of pool noodles, and yes, I bought extras to become bumpers.)

3

Gus,

Thanks for the input.

The one thing that may of been good in my test is I used a digital electronic load that is self cooled. Which clearly doesn’t effect the motor coil resistance, but may have helped with the resistor load. The electronic load may have been bad due to the fact I tested the output at many different load levels (5a - 10a mostly) and being new at torque testing I am not wondering if that itself made my test have bad results. I am not sure if I should be performing all tests with one set load.

What do you think?

Also anyone elses input would be appreciated! I just do not know if my test data is all invalid/useless due to the way I tested.

4

You could also consider lifting a weight rather than using a flywheel. The weight will provide constant force opposing the motor. I did that back in high school to provide constant acceleration when I was trying to model rolling friction of RC car wheels.

Another option is to gear the dyno motor sufficiently so that it requires very low current when running. That would minimize the heating effects on the motor. Active cooling (pulling air through the motor with a shop vac) would help too I imagine. At some point you can probably lower the effect enough that you can effectively ignore it.

5

You’re welcome! Looking in more detail, it appears that keeping the load resistor constant is great for determining motor/generator efficiency, but the change in motor coil temperature still matters. I’ve been thinking about trying to determine coil temperature from measured resistance, but I haven’t figured out a good way to calibrate it - the most practical I’ve come up with so far involves running signal wires through the crack in the door of my [cooking] oven.

Also, as I look in more detail at Mike Betts’ post, I think that the equation for mechanical power at the coupler assuming equal efficiency for the motor and generator modes of the two matched motors should be the geometric mean rather than the arithmetic mean:

Mechanical Power at coupler (W) = sqrt(Input Power (W) * Output Power (W))

I do intend to use a falling weight to measure the moment of inertia of my flywheel. If you lift a weight to measure torque, you need to have a way to keep track of the moment arm (spool radius), which either means a “level wind”, or a relatively flat cord (or tall thin pulley) for which the radius will increase in a known fashion. With a falling weight, it is easier to control the radius as a function of the number of revolutions.

At least for what I’m trying to do, I want to be able to use the motors at “FRC realistic” situations. FRC motors in well designed applications spend most of their operating time in the range between peak efficiency (for those which run nearly continuously) and peak power (for those which run in short bursts). If possible, I’d like to be able to test at or at least reasonably extrapolate to both free running and stalled full-voltage running, because these are the numbers which are most commonly published and will provide a reality check.

6

Hi all,

I’ve been pondering a way to measure our motors, and motors with gear boxes for years now.
Weights, pumps, magnetorestrictive, motors, etc…

What I’ve come up with, and please! feel free to point out deficiencies here, is to use a motor, possibly many motors to accommodate different test motors, as a generator with an electronic load, perhaps just some mosfets on a big heat sink, but it’s variability is key, as a load.
The key is to mount the load motor on a free moving set of bearings, think 2 posts, one on the nose, and one on the tail with bearings that hold the CIM for instance.
Then have a lever arm of a precise known length pushing or pulling on a strain gauge, or even just a commercial force gage.

This setup would allow one to know the rotational torque by actually measuring it, while measuring the input voltage (variable) and current to the device under test.
This could be done while changing the load to the DUT, and thus generate a curve.
Of course, the best way would be to do this all automatically controlled by a computer.

The hard part for me is to come up with a ‘relatively frictionless’ way to mount the load motor as that friction would be an error element.

Any thoughts?
Mike

7

This is the best way IMO.

8

What you are building is called a dynamometer, or dyno. At my company we use bearings to mount the load motor, those are called the trunnion bearings. You need to mount the load motor in a frame that can hold bearings at the front and the back, inline with the shaft. Use open bearings with light oil to reduce the stiction that causes errors.

When you are calibrating you need to have the load motor running to generate a tiny vibration that helps keep the trunnion bearings from sticking. You then need to filter out your load cell signal to get the best results.

9

The frame holding the bearings that allow the load motor to rotate slightly would have to be very stiff. Any flexing would cause the bearings to be misaligned leading to extra friction. Also, flexing of the frame connecting the two motors may exacerbate any mechanical resonances in the motors. Lastly, the shaft of the motor under test would have to be lined up very precisely with the shaft of the load motor unless some sort of compliant coupling is used.

My last workplace had a lab with several such dynamometers with capacities ranging from a few hp up to over 1000 hp. They were all mounted on massive steel blocks that were precision ground to be perfectly flat. The Motor Test Lab Techs would spend about half a day installing each motor under test and getting it aligned perfectly with the load motor. Unfortunately, I never paid much attention to the other details of how they were built since we were mainly there to test motor controllers.

I think the advantage of the OP’s method is that it avoids all the mechanical issues by assuming that the efficiency of the motor converting electrical power to mechanical power is the same as the efficiency of a motor of the same type operating as a generator. The difficulty of this method is that the load motor must be of the same design as the motor under test. This limitation can be acceptable when one is trying to characterize the motor and can get two of any type. It would not be acceptable when one is doing R&D on the motors and tweaking their design.

10

Yep, a dyno, or De Prony Brake… Maybe there’s a difference, but either way.

First let me say, that if I can get the error under 5-10% I’d be a happy camper. As I understand it, the mfg tolerances of these cheap motors probably isn’t even that good, and besides, for FRC .1% just isn’t necessary most of the time.

Having said that, yes, a flexible coupling method is pretty much assumed I think in any setup given that aligning things to .1 mils is difficult to say the least.

I agree that the OPs suggestion takes a bunch of short cuts, and probably gives numbers that are ‘close enough’ maybe?, but that’s why I’d like to have something to refer it to. Something that we actually measured and the kids physics teacher would approve of so to speak.

There are a few principal things I’d like to be able to demonstrate, and accomplish in making something like this.
I’d like, probably mostly, for this to be a nice project for some of our mech, electrical, cadd, and software students. Kind of an advanced project for Jr’s and Sr’s.
I’d like it to be able to show the real power output of the motors under the conditions we run them at, which is to say, never at 12 VDC unless it’s free running, but rather at the 3 to 10 volts that are actually delivered to the motor after going through the whole FRC chain of things.
And finally, if we can swing it, what is the effect of adding a gear box?
Numbers seem to suggest anywhere from 1% to 50% loss depending on the type, but do these numbers really hold up using FRC parts, Vex, AM, BB, etc.?

Anyways, that is my dream as it were, lol, I’ve been pushing it for years now, maybe this will be the year I can get a critical mass of students interested.

As always, thanks for your thoughts,

Mike