|Feb 24, 2009, 03:29 AM|
Test Fixture Evaluates Motors/Props/ESCs for Quadcopter Performance
Iíve spent the last several months developing a test fixture and test procedures which I hope will contribute to the quadcopter community. Rusty (13brv3)
has also built a similar test fixture and has helped in the evaluation of motor, props and ESCs as well as the evolution of the design.
This page will be updated as additional motors, props and ESC combinations are tested. Rusty and I have been testing a lot of combinations. Here is what we have so far:
-10 motors and a few custom wound motors
-7 props of different sizes and pitch
As a start, here is the summary for the very popular EPP1045 prop with various motors:
Chart #8 (Updated May 10, 2009)
I hope the color scheme is evident. Note that the best performance number in a column is the deep blue and the worst is the deep orange. I tried to make the color ranges match a regular partition of data distribution.
|Feb 24, 2009, 03:30 AM|
Performance Summaries of Motor, Props and ESC combinations
The test fixture consists of a lever arm with the motor/prop mounted at one end and a "load cell" sensor which measures thrust at the other. There are
weights on the lever arm which allows the system to be balanced before starting the test. During the test, a PIC device records the following data at a rate
of 512 samples per second:
1) Thrust in ounces
2) Current in amps
3) Voltage in volts
4) Throttle in ms
The tests consist of the following:
1) Sweep Test This is a slow sweep of throttle from min (1.000ms) to max (2.000ms), a hold at max for 1/3 second and then a sweep down from
max to min throttle.
2) Wobble Tests Three wobble tests are done to measure how fast the motor/prop/esc can respond to progressively faster throttle swings.
The three wobble starting points are:
a) 1.600 ms Throttle
b) 1.650 ms Throttle
c) 1.700 ms Throttle
Each wobble tests has four groups of throttle swings:
a) Three 390 ms cycles of +/- 50 us throttle steps
b) Three 200 ms cycles of +/- 50 us throttle steps
c) Three 80 ms cycles of +/- 50 us throttle steps
d) Three 40 ms cycles of +/- 50 us throttle steps
Note that the three wobble test center points were chosen so that most motor/prop combinations will have a wobble test close to the typical thrust
for a hover (at least for the quads flying higher end video cameras). For example, a "low wobble" center of 13.7 oz which would be 54.8 oz for a quad.
My typical quad platform with HD camera, FPV system and GPS is 55 oz so this is very close to what the motor would be doing during a hover.
A correlation match is performed for the first three samples to establish a Wobble Quality Factor (WQF).
3) Vibration Test A vibration measurement is made during the stable period between wobble cycles quantifying the thrust variation.
The PIC processor sends the data in real time to a PC where it is stored and then imported to Excel. Here is a video of what the test run looks like:
An Excel spread sheet is used for detail analysis and summary of the data collected. Here are some example screen shots of the graphs produced from the tests:
Sweep test. Left axis is Power (black trace) in watts and right axis is thrust (red trace) in oz. Horz axis is time in ms.
1) The resonance of the motor arm which occurs at a low throttle. If you expand this region in the spread sheet, you can see the classic resonance response.
2) Key metrics from the run are shown shaded green at the top of the chart.
Second Wobble test. Left axis is throttle (blue trace) in us , right axis is thrust (red trace) in oz and current (green trace) in oz. Horz axis is time in ms:
Note that the response time of the system is easily seen in the first cycle group and that the motor/prop/esc cannot keep up as the swings get faster.
A set of metrics were established which allows a comparison of performance for each motor, prop and ESC. Here are the metrics calculated after each test run:
A subset of the above metrics are used in the summary chart in the next post.
Finally, the load cell can measure vibration noise over a very wide spectrum. Here is a sample spectrogram (available at http://www.dewresearch.com/fftp-main.html)
- At the center plateau of the motor sweep you can see that the strongest vibration is at aprox 1400 Hz. For this 12 pole motor, that would mean that the
RPM was 1400/12 * 60 = 7000 RPM. Brashly has a nice thrust vs RPM chart for the EPP1045 prop which shows that 7000 RPM corresponds to 30 oz of thrust.
This is just what was measured as max thrust in this particular test.
- The strong vibration corresponding to the number of poles times the RPM suggests that a lower vibration motor/esc combination might be possible if a
smoother transition was made between each pole "push/pull" impulse. Then again, that would probably slow the system response speed.
There is a lot of analysis work that could be done with just spectrogram study. But for now I will concentrate on the basic data collection of the various motors,
props and ESCs.
As with my other Kquad thread, I plan to maintain an Index in this first post so that key future posts in the thead can be quickly located:
High Cogging vs Low Cogging : http://www.rcgroups.com/forums/showp...7&postcount=66
|Feb 24, 2009, 03:32 AM|
Test Fixture Design Details
As noted in the introduction, the test fixture consists of a lever arm with the motor/prop mounted at one end and a "load cell" sensor which measures thrust at the other. My platform was made from scrap aluminum and some bearings used for sliding doors available at most hardware stores.
(More details supplied here as time permits)
The data collection is done with a modified OMM RX converter board (available from Rusty). A few resistors, capacitors and a 4.7v zener diode was added to the board to allow interfacing to the load cell. The load cell and current sensor used were first identified to me by Rusty:
A small handwired board was built for the current sensor so that it could sit in series with power and interface with the PIC board via standard 3 wire servo connectors. Here is the schematic and picture:
The Load Cell is interfaced to the PIC board as follows:
In addition, this RS232 interface was used to allow a real time data download from the PIC card to a PC:
Here is a system overview of the interfaces and some detail closeups of the PWB interface described above:
The firmware I wrote for the PIC to perform the testing is available here:
The program uses the on-board jumpers to select either special programming of the ESCs or run the normal test:
LED side jumper: Program ESC throttle calibration for Spectrolutions Quad
Center jumper: Run Test
Remaining jumper: Program ESC throttle calibration for max range (1.000ms low, 2.000ms high)
(More details showing how everything interfaces supplied here as time permits)
The Excel Spreadsheets and all data runs are available for ftp download from the quad directory at:
(No password is required)
|Feb 24, 2009, 09:16 AM|
Someone has been busy little bees. Beautiful and amazing work. I, for one will be reading and re-reading this thread a lot. Much information to learn, and I thought I was through school 45 years ago. Just to prove, one never stops learning...Keep up the good work, OMM and Rusty...Tony
|Feb 24, 2009, 10:49 AM|
United States, TX, San Antonio
Joined Feb 2007
Great work OMM and Rusty! So it appears the 2410-09 series are still the ones to beat for efficiency and cost per weight/thrust/vibration. Now we need to come up with at way to balance the motors themselves and provide good vibration isolation from the motors.
Just wondering, did you use new motors and correct the motors end play and allen screw tightness on the shafts and bases before the tests, or just use them as they came new? Also, how did you make the props consistently dynamically balanced for each motor, in order to eliminate them as a cause for vibrations?? What do the white spaces in the vibration fields represent? No data or no vibrations?
|Feb 24, 2009, 11:59 AM|
Very nicely written Mike. Looks like the beginning of another classic OMM thread
I'd also like to add that it's a real pleasure to work with someone as talented as Mike is, though for about a month, I thought I had been enrolled in the OMM school of engineering! I spent many, MANY hours on Internet just trying to keep up with the old guy. When he started pulling out Spectrograms, I knew I was in trouble
To answer a few questions, Mike mentioned WQF (wobble quality factor) in the initial post. He'd have to comment on exactly how that's measured if anyone really wants to know (or is having trouble sleeping <g>). He fine tuned this several times, and we went through lots, and lots of motor runs to gain confidence in the figures. Basically, higher is better, where 100% would mean that the measured thrust exactly matches the commanded thrust.
"Vib-O" and "Vib-R" are vibration measurements taken on out test rigs (O for Mike's, and R for mine). While both rigs work the same, I got just a bit carried away building mine, and as a result, it tends to have a higher natural vibration frequency than Mike's rig. Fortunately, the thrust, and wobble results track well between the two rigs, but getting the vibration to match has been somewhat elusive. For now, it's not fair to average the results of the vibration together. Ideally, we'd want to mimic the vibration properties of an actual frame, but that's still a works in progress, and Richard (Brashley) has done lots of work in this area already.
Jim, I haven't noticed any difference at all between having end play or not in the motors. There should be a slight difference, since the magnets in the bell should be centered on the stator, but it must be too small to measure.
Props are not precisely made, so there will be variations. Motors are not balanced either, but I've yet to see any significant different when attempting to balance the actual motors. I tend to use the same prop for all the tests, but Mike kept a prop mounted on each of his test motors. When he sent them to me, I got the whole assembly, so the only difference was our rigs.
Nuts vs prop-saver (o-rings) mounting was tested, and I was surprised to see no difference in the wobble performance. I thought if there was ever going to be a good reason to use nuts, that would be it. Fortunately, it proves that there's no good reason to use nuts
Enough for now,
|Feb 24, 2009, 12:30 PM|
United States, TX, San Antonio
Joined Feb 2007
...or a reason to use prop-savors either...
Again...great testing and data Gents!
|Feb 24, 2009, 12:47 PM|
United States, TX, Fort Worth
Joined Jan 2009
Bit of a scientist myself (hence the -lab in arbilab). Career, until sent to China.
I have jaw-dropping admiration for what you've assembled and accomplished. True repeatable, quantifiable data, where we're used to seeing "a lot better than stock" as an evaluation.
I bet the 'professionals' making and selling these things don't have that degree of elegance in their development programs. Outstanding work.
|Feb 24, 2009, 01:50 PM|
United States, NH
Joined Jun 2006
Great work as usual. I am using Himax HC2812-0650 with EPP1245.
looks lot of thrust at low power, I had to add gromets for vibration with the EPP1245.
Have you tested these motors.I am curious to see were they stand on vibration.
|Feb 24, 2009, 04:02 PM|
|Feb 24, 2009, 07:32 PM|
[QUOTE=13brv3] While both rigs work the same, I got just a bit carried away building mine, and as a result, it tends to have a higher natural vibration frequency than Mike's rig. Fortunately, the thrust, and wobble results track well between the two rigs, but getting the vibration to match has been somewhat elusive.
Outstanding! Nice work boys. I love the test fixture Rusty. Just how much different are your respective natural frequencies? Considering you guys both used the same Load Cell, the pivot to cell distance will primarily drive the system stiffness. Next comes the rocking inertia of the arms on either side of the pivot. After that, the method of coupling the forces usually dominate. Are the test jig resonances outside the operating range?
Any pictures of the jig Mike?
The wobble tests are most interesting. were the ramp profiles based on data gathered from Eagletree Recorder flight data?
|Feb 25, 2009, 02:09 AM|
Australia, VIC, Ringwood
Joined Feb 2008
So I am seeing a rise time on the example of around 100mS is that correct? What is the PWM pulse repetition rate and is there an offset from the trailing edge of the pulse and the motor response? Ditto if I2C control.
Questions relate to control loop considerations of course.
|Category||Thread||Thread Starter||Forum||Replies||Last Post|
|Yippee!||motors/props for 1/200th amato||sgonzo||Scale Boats||15||Aug 06, 2011 10:12 AM|
|Sold||Brushless motors/CC ESCs for sale!!!||turbotodd||Aircraft - Electric - Power Systems (FS/W)||0||Aug 22, 2007 09:30 AM|
|Brushed motors and ESCs for sale||topforce||Aircraft - Electric - Airplanes (FS/W)||1||Nov 21, 2003 11:10 PM|
|FS - Props, Motors, Gearboxes, ESCs and Misc.||Rooster287||Aircraft - Electric - Airplanes (FS/W)||14||Nov 18, 2002 07:39 AM|
|Sold||Bunch of stuff for sale, Motors, props, receivers||Steve H.||Aircraft - Electric - Airplanes (FS/W)||10||Oct 16, 2002 07:09 PM|