I'll have to go through my circle of friends... :)

Back on that subject, I do think that the RC Benchmark one seems to be one of the most promising, in terms of quality, and being adaptable. I'm really hoping that a solution to determine the damned optimal timing advance curve for my motors, once and for all, can be found! :lol:

It has another advantage that would only be recognized by someone (me) who has attained the lofty status of "Old Codger". Judging by his photo, Jebarus looks young and eager, which is perfect for other guys (me) who just want to kick back and let somebody else do the hard work. :)

This is more or less what I was thinking too. :)

Also, I feel like some of the existing dynos were built by people with more practical experience than theoretical experience, which ends up "mostly right", but not quite? Slave motors? Slave motors [I]and[I/] flywheel (some of the McPappy setups)?

Now I'm getting a nice whiff of engineering degrees, and hear mentions of Matlab, which I'll take as good signs. ;)

The G-Force 1S 120A, a common ESCs used for my class of racing, is rated for 120A continuous, 760A burst: http://www.gforce-hobby.jp/products-en/G0014.html

The Reedy Blackbox 410R that I use is rated for 150A continuous: http://www.reedypower.com/products/details/27000/

My charger tells me that it puts in up to 2500mAh after a 5 minute run, so averages up to 30A? This means the continuous current isn't close to the 150A rating of my ESC, but it's hard to know if this is 5A continuous with big peaks, or if it's more like 20A with 40A bursts, etc. But it should give you an idea?

Some motors (such as the LRP X12 and X20, I believe) have built-in temperature sensor, in the form of a 10k thermistor (but not all motors have them). See section 8.5.1.3 of the ROAR (the US governing body, one of the biggest, so this is as close to a standard we get) rules:

http://www.roarracing.com/downloads/..._Rule_Book.pdf

For me, this motor temperature sensor would only useful mainly as a safeguard, to stop an automated test sequence if it exceeds a threshold?

I was just dreaming up of nice features, really, I think being able to script and automated most of the data gathering, even if there's a few pauses where the script tells me to do something to the ESC, would already be quite nice. :)

Maximum torque at very low RPM (right at the limit of zero, for example) sounds very difficult to get with a propeller, though? And even at some other RPMs, I'd expect I would need a heck of a propeller to get that information!

For example, free running RPM on this 10.5T motor seems to go to nearly 30,000 RPM, to get the maximum torque reading at 2,500, 5,000, or 7,500 RPM, I would need a propeller big and pitched enough to keep the RPM down at those speeds while having the ESC at full-throttle, 100% PWN! This could get dangerous, and possibly change the climate inside my apartment. ;)

I think the main difference from quadcopters and such is that we're concerned by the torque curve over RPM, which implies that it's the maximum torque the motor can put out, which then implies that the ESC is at full-throttle. Imagine a slow hairpin leading into the straight line, the motor might be revving as low as 4,000 RPM as you come out, then you mash the throttle, trying to get to the top speed on the straight as quickly as possible. Now, how much timing (phase) advance do I want at each RPM? Here's a (theoretical) example of full-throttle torque curves at three timing advance settings:

http://pphaneuf.github.io/rccars/res...ing-torque.svg

If I have the high timing setting kick in before 10,000 RPM, I'm losing out. If I'm still on the low timing setting past 7,000 RPM, not so good. Past 12,000 RPM, disastrous! So I need to program my ESC properly, to ramp up the phase advance properly, but right now, it's all educated guesses and poking around in the dark, from looking at the car on the track...

What I'd like to do is gather the torque curves for a number of static timing advance settings, say from 0˚ to 80˚, every 5˚, find a timing advance programming that will follow the optimal timing advance as closely as possible, and validate the programming by getting the torque curve with the dynamic timing settings. If those 18 timing setting runs need to be multiplied by X maximum torque readings (with different propellers), this is quickly becoming a big project! With a flywheel, I just need one spin-up at each static timing settings (more realistically, a few spin-ups, to have a few samples and make this more statistically significant), calculating the torque from the curve of the acceleration (since torque is MOI * acceleration, for a fixed flywheel/MOI, it's basically equal to acceleration, since I don't really care much about proper units).

Quadcopters probably only rarely need the full-throttle information, and would be more concerned by efficiency (thrust per watt, say?) at various partial throttle positions, over the RPM range, reflecting how it would be used? For that, you might be more concerned by the efficiency of the combination of motor and propeller, at various RPMs (achieved through ramping of the PWM/throttle)? Timing advance might be of some interest, but considering how outrigger motors seem to not need much timing advance, and how most gains are in power at the cost of efficiency, I'm guessing it's of little interest to the flyers?

Sorry about the delay in posting, I was checking the thread everyday, but development took a lot of time this week.

I figured out this week that basic motor theory does not exactly fit experimental data, especially when there is an ESC between the battery and the motor. I spent a lot for time trying to figure out why, and what is the best way to fit the model to the data. Also, visualizing and explaining the data is not that simple. It took me quite some time to do in matlab, and I have an engineering background as well as practical experience. We cannot expect that most people will do the same. The good news however is that all the data I will show here could be generated automatically with a script.

You might be wondering why is the motor model important? It allows comparison between motors, simulation, and it is very important if we want to build a database of motor. Motor model will allow in the future to do a search on a database and say: I want a motor with 0.2Nm of torque at 4000RPM that has 70% efficiency and more, and a diameter of less than 40mm. It does not exist right now, and finding a motor is difficult for designers.

Also, how can it help you race faster? Motor theory helps to understand how changing parameters will affect the torque and speed profile of the motor.

I am using the traditional motor model. It is explained on this website (sorry, I cannot post links due to insufficient post count)

ctms.engin.umich.edu/CTMS/index.php?example=MotorPosition&section=SystemMode ling

I was using a NTM 28-30 800KV and a small 18A car ESC.

Here are the graph I obtained:

imgur.com/a/qqvlf

What are those graph? They give you almost all the information on the motor.

Here is an annotated version:

i.imgur.com/SNiSkvc.png

If you look at the bottom right (point A), you see the point of no load, and maximum RPM (100% RPM). At the top left, (above point B), you see the maximum torque theoretical.

Trapezoidal shape is cut at the top, as it represents the maximum current rating for the motor.

When you move along the black line, you are at constant input, in this case 100%. A line parallel to it will also be a constant input. When doing an acceleration test with an inertial wheel, you are doing a sweep along this line.

When testing a propeller, you are doing a sweep along the white line.

For cars, as mentioned earlier, the most interesting points are the maximum torque, as well as maximum power. Ideally, you'd always be in the maximum power region (on the motor power graph). Since the gear ratio is fixed, it is not possible, so you have to compromise.


We implemented a system to allow bursts current in the software. The idea behind it is pretty neat, we are estimating the temperature of the board with a low pass filter of the current. The limit current is set to 50A burst and 40A continuous now, as it is the maximum we tested. However, the device should support at least 80A continuous per design, and much more burst. The current limit we have is due to the precision resistors. Since they are beside two big aluminium connectors, they should be able to take some heat.

Consequently, we will work on testing higher current limits to certify the device. I have a GTB speedspectrum from my B44, rated at 540A per phase. We will buy at flywheel to work on dynamic tests. As Pphaneuf mentionned, the solution will probably to run the test multiple time and average the data for more precision.

I have some more research to do, and I may actually order a custom laser cut aluminum part for the flywheel....I will keep you updated.

I forgot to mention. The circles in the points above are the experimental data points. Here are the graphs from the side, so you can see better the results.

imgur.com/a/nrlyT

The efficiency model has the most error. I think it is partly due to the ESC not properly modelled.

To obtain this model, I spent a lot of time deriving the motor constants. Here they are, with some explanations:

KV: 800 (experimental) The most known motor constant. It came out to exactly 800 surprisingly. I tested a lot of motors, and the manufacturers are usually not that close, it is probably chance. There is often more than 10% of variation. It is simply the max RPM divided my the voltage.

Kt: 0.0119 N.m/Amp This is a ratio of torque to current. In theory, the current is proportional to the torque. In SI units, KV and kt are equal. Some researchers are recommending to calculate Kt with a fit on the current vs torque plot. It is suppose to give a more accurate model. I tried it. The result is a bit different, but the problem is then that the KV you obtain is different from the manufacturer KV, and it will confuse everybody on a database.

b: 7.3795e-008 N.m.s Friction constant on the motor at no load (from bearings, etc...). It has a very small impact on the model, only at very low torque. It is important for modeling efficiency though.

R 0.3116Ohm. Our dynamometer measures the motor resistance as 0.1648 ohm per phase. The resistance constant here includes the ESC. Was calculated with a fit on the data.

I0: 0.33A Current at no load, or the current required for the motor to spin at no load.

Imax 20A Max current due to the motor wires. Could be higher for short periods of time.


I know it is a lot of information, and I want to make a tutorial on our website. I am looking forward to start the dynamic tests and compare the results.

Another note: A problem I encountered is that the battery voltage changes during the tests. It changes the results, but it should also be possible to model. I think it would be great to have a battery database on our website, and rank them by actual, tested C rating.

PS: After testing it, car dynos definitely have a lot more startup torque than quadcopter dyno.

Edit2: Your timing graphs are really interesting, I did not know about that. They seem to be modifying the Kt value. I wonder how the curves will look like when experimentally tested.

Edit3: My college suggested that on factor explaining why my experimental curves were not following the theoretical one is the changing timing of the ESC. We were also discussing how the accelerating characteristics are pretty much as important as the acceleration characteristics when racing.

I can see how that could be very useful information for a designer! In the car racing world, motors are more standardized (rules, again!), but there's some variation in the number of turns, and the design of specific models, not only between different brands, but sometimes within a model range, such as the Reedy Sonic M3 13.5T and the Reedy Sonic M3 13.5T short stack.

It's my understanding that manufacturing tolerances are just loose enough that you can get a bit of variation between even identical model motors, but it would be good to have some starting point when setting up ESC boost settings!

Exactly. I've got the general shape of torque curves and such, but motor specific constants need to be filled in, so that we can get the best settings.

Here's an example of the graphs I used to base my theoretical graphs in my articles:

http://www.rctech.net/forum/attachme...4-20-09001.jpg

A common pattern of these graphs is low current at the beginning of the spin-up, which some of us think might be due to current limiters ("punch control") in ESCs?

I'm not very familiar with flying ESCs, but I would guess that most of them apply some small amount of dynamic timing, to improve efficiency and/or performance? ESCs used for car racing usually have what is referred to as a "blinky mode", with zero phase advance (fixed timing), to comply with rules. It's called blinky mode because it is identified by a blinking LED on the ESC, so that race scrutineers can easily verify for compliance (and this also explains why there are so few fully programmable, or open source ESCs for racing, because they could too easily be reprogrammed to circumvent this).

Here's an example of the power curve with a very rough dynamic timing advance (only three timing settings, simply switching at set RPM thresholds, while real ESCs allow a linear timing advance based on RPM):

http://pphaneuf.github.io/rccars/res...wer-simple.svg

You can see how a smoother application of the timing advance could still look like a fixed timing advance, but might very well throw off your model's expectations!

I'm not sure I understand what you meant by that? "Accelerating" vs "acceleration" characteristics? Or did you mean to compare flying and car racing?

Sorry, that was a typo, I was talking about deceleration, or braking. The motors' characteristics in a dynamic test should be the same in acceleration and breaking, but the ESC's firmware might handle it differently.

Oh, yeah, you get an increase in braking strength with a more powerful (or better tuned) motor.

And the ESC makes a big difference on that: my LRP Flow and Reedy Blackbox 410R both have strong braking, but my SkyRC Toro 1S120A is very weak, with the same motors!

I'm usually not concerned about braking strength, it's usually more than enough, and I just turn down the endpoint on my transmitter (except for my SkyRC Toro!). ;)

I started the tests with a homemade inertia wheel using a sensorless motor:

http://i.imgur.com/ljEut3h.jpg
http://i.imgur.com/u7qhwnx.jpg
The motor does not want to start, no matter how much I spin it initially. Basically, it means that for a sensorless motor, testing with an inertia wheel is difficult. Loading with a propeller will give all the information necessary as sensorless ESC cannot go over a certain level of torque.

I will now start to test with a 540 sensored motor and a proper racing ESC.

Now you see why we use sensored motors for racing!

The startup algorithms for sensorless motors are goofy, at best. I've written software to do it--probably 30 years ago!-- so I have sympathy for those engineers that attempt it.

My SkyRC Toro 1S ESC works in both modes, but I can tell when the sensor cable detach itself after a crash, starting up gets coggy, and it loses all its smoothness out of slow corners... :)

There is a lot more torque with the sensor now! However, as I expected, the vibration is horrible due to the bad adapter. Consequently, I ordered a cd to propeller mount adapter:

http://i.imgur.com/2Z6IRHd.png

It should come in in two weeks (shapeways). I cannot really do any dynamic tests before that, as I do not have a flywheel. My colleague and I will focus on releasing the motor database. I think cds will be great to prototype as they are balanced, and it will be easy to try different flywheel inertia. I can stack up to 13 central discs, so 15 CDs!

I ordered 3 copies, and we'd need two to test here. If they work, would someone be interested in a copy?

I don't CDs are a good choice. Their normal operating speed is only a few hundred RPM. Our motors can easily exceed 20000 RPM. At that speed, it's very likely a CD will explode!

My 10.5T motor, with 1S and dynamic timing, regularly hits 25,000 RPM under load (from my ESC RPM logging function).

According to Wikipedia, about 40-52x CD-ROM drives: