Fly Electric!

Balancing Motors

It is feasible to make your own brushless motors but it is not without its challenges. The one thing that makes a motor worthless is if it resonnates. This is such an unpleasant sound that you will not want to use your motor. High speed motors will not tollerate even small vibrations. My machining tips page describes how I avoid most of these problems. However, this page describes my experiments in dynamic balancing your pride and joy when it does 'rattle and roll' a bit!

This page is still evolving as I explore this topic but I think I have some meaningful results. The following photos show my setup. As you can see I am using an oscilloscope to measure AC vibration signals generated by an accelerometer (from Dimension Engineering). You can use a free PC/sound card based scope (example below) but these often 'clip' at lower voltages than you want (you should be able to get round this with a resister voltage divider).


The motor mount comprises a thick wooden base. The motor itself is mounted on the smaller 4mm thick ply (there is a hole in the base to clear the motor). This is attached to the base with flexi mounts to allow slight movement. The accelerometer is installed in a metal box with 'blue-tac'. This is then mounted on the small ply board at the top (again with blue-tac). This small board is mounted rigidly to the 4mm motor board so they move together. The main base is weighed down with four 1.5kg weights (the more rigid this is the better to improve sensitivity).

Electrical noise is a problem which is why the accelo is in the tin box. The lid and base of the tin are connected to the accelo's ground. Screened cable connects the accelo to the scope. I have a small 100pF disc ceramic cap between ground and signal on both leads (inside the plugs) to filter out higher frequencies (above 500hz). An IR 'reflective sensor' is used to (i) determine RPM, (ii) the start and finish of each cycle on the screen and (iii) to establish a position for determining where to add the counter-balance. You can see this in the photo on the right (again held in position with (white) blue-tac). One axis of the accelo should be aligned with the reflective sensor. In my setup both the accelo and IR circuit use the same 5v supply. It is battery powered to avoid ground loops.

Measurement technique

Now on to balancing. Here is my 5x43mm 12T/14P motor. It has more than double the vibration of my 10mm Ditto motor (based on the amplitude of the scope signal 1v/div). This increases with RPM although there are some rough bits at intermediate speeds which I will explore separately.

The shape of the curve appears a bit different but proved to be easy to make the repeatable measurements shown below. However, what is actually happening is that the accelerometer is cropping the signal. With a 5v supply, 0.3v represents 1G. The 2v observed represents over 6G which exceeds the device's 5G rating.

The motor has one trigger line marked on it and the reflective sensor shows that it spins at about 14,600 rpm at full throttle (0.5sec/div). I have positioned the trigger line (through trial and error) to match a low point on the trace.

The amplitude of the trace should reduce when vibration is reduced. However, due to the cropping described, the effect with this motor was to improve the trace's stability. The photos are taken at 1/4 or 1/5th sec exposures. These reveal that the stability improves progressively until it falls away again. The most stable is the third photo taken with a 10mm length of lead solder.

No counter-balance

5mm counter-balance

10mm counter-balance

15mm counter-balance
Too much

Experiments revealed that the most stable trace was obtained at the 180' position (between the trigger marks on the scope screen). In the following photos I demonstrate how the optimum position can be found. In the first three photos the weight was moved by ~25 degrees (1 magnet width - 360'/14 magnets) to find the best position. The last photo is fully opposite the optimum to prove I was getting it right.

25' left of 180'
Worse than 180'

at 180'

25' right of 180'
Worse than 180'

At 0'

By reducing the voltage I have slowed the motor down in the left photo to 10,900 rpm to keep it within the accelo's 5G range. I have also taken a five second exposure to reveal the average trace (and moved the trigger to the 180' position for convenience). Although the scales are different from the photo on the right you can see there are three imbalance points in each cycle which is why the higher speed (clipped data) had such a flat top.

10,900 RPM

14,600 RPM


The above tests were taken at full throttle which on this motor reveals the imbalance positions the best. There are two intermediate speeds at which this motor resonnates which I explore in this section. These cause the trace to become extremely 'agitated' and have a very small amplitude, presumably because it is 'changing direction' so rapidly.

Because the accelo's full scale deflection is 2v, I have changed the scope's vertical scale to reflect this (0.5v/div). Also, to fit a slower speed onto the screen the horizontal scale is on 1ms/div.

The first resonnance point is at about 6,600 rpm in the photo on the left. To illustrate what you may see at intermediate speeds, the next photo is at about 6,800 rpm - a larger but unstable trace. The third photo shows resonnance again at 8,900 and then a normal more jumbled trace in the last photo at ~9,100. The resonnance bands are very narrow but are invariably at the speed you want to fly!

6,600 RPM

6,800 RPM

8,900 RPM



Repeating the test a day later, the motor still resonnates badly at the same two speeds. Although the trace is still quite 'concentrated' at the resonnant rpm, it is less so in the first photo on the left compared to the versions above. I then added weights progressively at the position determined above. As you can see the counter-balance 'softens'/spreads the trace and although it has not entirely removed it, the resonnance is also less audible. It may not be very obvious from these small photos but they all have fine 'sawtooth' traces; these become less pronounced as the resonnance decreases. I suspect that the 5 or 10mm weights would be best for this motor. These photos were taken at 6,600 rpm; the results were similar at the 8,900 resonnant speed.

No counter-balance

5mm counter-balance

10mm counter-balance

15mm counter-balance

Other considerations

The two channels on my accello look very different. However, they both appear to be revealing the same waveform as these two different tests show. They should be 90 degrees out of sync which I think they are. I tend to use the channel with the smaller trace. Using just one channel is all you need and allows you to observe the refective sensor at the same time as most of the shots show.

Old Tests - 19-Mar-06

Spinning just the rotor with another motor (see 18mar) has proved to be difficult. Even with the 'drive' motor (Speed 480) mounted on a separate table alongside the rotor under test, the vibration of the S480 is transferred via the drive belt and makes that approach impractical.

So here are three photos with a 10mm Ditto motor under its own steam at full throttle (~11,300 rpm - highest vibration point for this motor). This is the same motor as in the 5Mar results with a free PC sound card based scope. The results are essentially the same with near identical amplitude at full throttle (the real scope is set to 1v/div).

The photo in the middle is the motor without any counter-balance. The amplitude reduces with a 3mm length of thin solder taped to the rotor near the reflective sensor trigger mark (right photo). The amplitude is greater when this is positioned 180 degrees out of sync (ie: more vibration). While it is easy to increase the amplitude it is hard to reduce it on this motor. Increasing the weight at the 0' position tends to push the imbalance sideways and tends to increase the amplitude.

Counter-balance at 180'

No counter-balance
Baseline measure

Counter-balance slightly to right of 0'
Slightly better


Some new experiments and equipment today:
1. Proper analogue scope (100mHz Tex 465) to eliminate potential inconsistencies from PC sound card.
2. Same rotor but on a bearing holder without any stator to eliminate effect of magnets/ESC and any electo/magnetic driving forces.
3. Rotor is turned with a 7.2v Speed 480 from (James' old) 7v variable power supply to regulate voltage/speed.
4. The S480 and rotor are connected with a thin rubber band whose vibration could influence results but is unlikely to be in sync with the rotor imbalance (I may test this with a simple shaft and no rotor).
5. The photos are taken at slow speeds so they pick up some 'duplicate' traces as the trace bounces about (and a little screen 'memory').

In the following shot (1/40th sec) you are looking at one cycle (1 revolution). Using the reflective sensor I have set the rotor's speed at ~6,000 rpm to fill the screen at 1ms/div. I am displaying both accelo channels to compare them. One channel has a consistently larger amplitude and cleaner signal. Both reveal the same basic waveform. They are both triggered on the same channel and I interpret these as being out of phase by 90 degrees as one would expect from the accelo's X and Y axis. What also now appears to be quite dominant at this RPM are THREE imbalance points (while true this proved to be from the S480 not the motor under test).

Taking the cleaner channel and including the reflective sensor trace at the bottom (to confirm the position of each cycle and to calcualte RPM), here are a couple of one second exposures. The first shows the rotor turning at ~6,000 again (10ms) and the second at ~8,570 (7ms). As you can see the traces jump around a bit but do have some clear concentrations. Interesting is that there appears to be more than 3 imbalances but that 3 are more dominat than the others. Oh dear, its getting complicated!


Getting closer. I've made more changes and this is where I am at:
1. Test stand is on kitchen work surface. Solid surfaces are important to remove 'sympathetic' vibration.
2. Stator is horizontal. This eliminates side gravitational forces on the rotor and the accelerometer.
3. The motor is mounted on rubber supports to allow slight movement. The Accelerometer is mounted rigidly to the motor directly above the motors shaft (ie: centered). X and Y axis essentially now give the same readings so only one is needed.
4. Cables have 'flex' near motor to allow a little movement and reduce the risk of 'tugging' on the accelo unit.
5. The whole accelerometer unit is in a grounded metal case (small sweet tin). This removed a great deal of interferance.
6. Cables to the PC are shielded. Metal plugs to further guard against interference.
7. 100pF disc ceramic capacitors between signal and ground on both leads at ends of cables near PC (inside plugs). This filters out more noise.
8. The Accelo unit is held in the tin can with 'blue-tac'. The tin is also held to the test stand with blue-tac. Wax would be better but the join breaks too often. Epoxy would probably be too permanent.
9. The accelo unit appears to be getting only a little feedback from the ESC and motor now. Tested by mounting the accelo unit in its tin box on an independant structure and suspended a few millimeters above the test stand. Scope monitored while motor speed changed. Small 'blip' corresponding with motor rpm and another at about 4.5kHz.

The main test (see screen print below):
1. Different motor again. Motor is usable but has noticable vibration.
2. Reflective sensor on Right channel (green). The beam is triggered by a 2mm black stripe painted with matt acrylic paint on the rotor. The image actually shows four 'blips' because I have four black stripes. These are unevenly spaced so I can tell which is which.
3. One accelo axis on Left channel, triggered by the Right channel. The larger the curve, the greater the vibration. The signal is still not very clean and the peaks on this motor often split into two. Nevertheless, I think I have a usable signal now.
4. By moving the marks on the rotor until they correspond with the peaks from the accelo, I hope I now know where the 'heavy' side is (it is either where the marks are or 180 degrees out on the other side of the rotor). Hopefully I don't have any phase shifts to throw me out!
5. You can calculate the motor's speed from the refective sensors time period (a little under 5.5ms). However, the 'multimeter' on the right gives an accurate frequency (PC time-base dependant of course) of 184.2487hz (x60 = 11055 rpm).
6. The three major peaks on the Spectrogram represent the motor's speed (left one) and two harmonics to the right).


Trying to understand what I am doing by using the OscilloMeter 'sound card' PC scope. First a configuration test with a 0.9v RMS AC signal from 50hz mains supply. Why 0.9V? Because my sound card seems to clip just short of 1v.

More information on DIY motors and machining can be found here:
* Summary of my DIY motors
* CD-Rom motors (easy)
* Crocodile motors (high efficiency)
* Machining tips (accurate machining)
* Winding Density (advanced advice)
* Milling machine (DIY mill from a drill)

Top of page
Click on the links above for more info.
Click on the links below for other pages.

Home Art of the Possible Absence of Matter Links Updates Email