I've watched the debate over the "speed" and performance of ESCs here for a while. I also see a great deal of effort being spent re-flashing ESCs to improve factory firmware capabilities. The general consensus has been that how fast the propeller can react to a flight controllers commands is directly related to the performance of a multi-rotor. This is true somewhat, but the blame for poor performance has been placed on the ESC mostly. This is incorrect simply because the ESC, Motor, Propeller combination sub-system that actuates the system needs to be considered as a whole. I do not believe the ESC is the bottleneck but rather the drive sub-system.

So about a year ago I set out to measure the Transfer Function of the ESC, Motor, Propeller subsystem since as a flight controller designer this is what is important to me. In engineering, Transfer Function is one of the most common figures-of-merit which is used to compare and measure a systems performance.

What is Transfer Function? Simple.. it is a measurement of a systems response to a known input. I've seen tests where the motor is hit with a step change in throttle. This is a great measurement. Transfer Function is then the derivative of the time measurement which is then Fourier Transformed to yield the Transfer Function (actually this is the magnitude of the previous). The easiest and best way to measure Transfer Function is to drive a system with a sinusoidal input and measure the time varying sinusoidal output. The modulation depth is then measured using the amplitude of the output waveform.

So my test stand optically measures the RPM (as my output value) of a common RC brushless motor (BL-2217/9) 16 times per revolution. I used a APC 10x3.8 propeller. A microprocessor drives several different ESC models with a sinusoidal signal at 200Hz. This may be confusing...the signal is the throttle input to the ESC which varies in time sinusoidally. I chose to use an average throttle value at 50% (hover in most multi-rotors) and vary the throttle 25% above and below this point. Using a larger or smaller value doesn't actually change the result much.

The resulting Transfer Function is a curve showing Relative Attenuation vs. Input sinusoidal frequency. To compare performance of the ESCs you can pick a frequency, say 30Hz and observe the attenuation seen for each ESC. A lower value means that the sub-system was unable to output a sine wave as large in amplitude as another ESC with a higher value.

The results are attached. Let me point out a few things.

1. Turnigy had the best overall response with an almost identical curve for the re-flashed Simonk Turnigy. Does this mean the re-flash is a waste of time? You decide...I am just presenting the data.

2. The Castle Creations Multi-Rotor firmware is cutting off at too low a frequency. They are working on this now and I fully expect them to be able to match the Turnigy curve. I'll post new results when I get them.

3. The Maytec 400Hz performs as well as the Turnigy but I wouldn't use it. At around 65Hz it begins to exhibit aliasing. This is when higher input frequencies are folded around a sampling frequency and result in a lower frequency being output. At 65Hz you will see about 20Hz oscillation which is BAD for a multirotor.

4. Smaller Props and faster motors will provide better performance. However, it will simply raise up the curves maybe 5%. Their relative shapes will remain the same.

5. The older Castle Creations firmware (data not shown) exhibits extreme aliasing after 50Hz. This is why the older firmware did not work on multi-rotors. The flight controllers where sending out commands with changes above 50Hz which resulted in very low frequency changes in RPM. Not what your basic flight controller expects and as a result FAIL.

6. Faster I2C communication is not needed based on this test because the actually ability to alter a motor/propellers speed is around 30Hz.

7. If your contention is that your multi-rotor performs better with one ESC over another it may be the flight controller or another part of the system.

8. RCTimer is the lowest performer assuming CC updates their firmware.

Additional Test

I have just completed two additional tests.

The first is a simple ramp up in throttle slowly to measure the RPM to Pulse Width values.

The second is a fast step up from about 5000 RPM to 6000 RPM.

I do not see any difference in the performance. Maybe a different test pair would yield different results. Suggest two models to test and I will repeat the test.

Added Transfer Function test for new Castle Creations firmware 4.01.

Looks like Turnigy is the best and cheapest for multis. Glad I didnt waste my time with the reflashing stuff. There are some here who swear that reflashed w/Simonk is way better than stock Turnigy.

I've been using Turnigy Plush for a long time and this data solidifies my positive experiences with them. Thanks for the efforts.

How do you explain the stability and response increase going from a stock plush to a re flashed one?

Modified fast pwm firmware Towerpro 18A ESC (1 min 48 sec)

Originally Posted by redridinghood

How do you explain the stability and response increase going from a stock plush to a re flashed one?

http://www.youtube.com/watch?v=c-Y6dyR8YFQ

I would say for your test to be accurate you would need to start out with atleast a 400hz ESC. then test stock 400hz to flashed 400hz.
I think what Als test proves is if you start with a good (fast)ESC then flashing it wont make much of a diff. Not the same as flashing an underperforming slow ESC to see if there us a difference.

Originally Posted by Z06 Tony

I would say for your test to be accurate you would need to start out with atleast a 400hz ESC. then test stock 400hz to flashed 400hz.
I think what Als test proves is if you start with a good (fast)ESC then flashing it wont make much of a diff. Not the same as flashing an underperforming slow ESC to see if there us a difference.

So give me the best two to compare. I will buy, modify how you specify and then test them.

These types of videos only give you qualitative data.

it is the same result like in the video with the original turnigy plush esc and flashed ones quad just runs smoother and motors respons quicer

I would like to prove that the motors run quicker. I just don't see it in the data.

Finally an engineering approach to this important topic. Congrats.

this is indeed very interesting. On one hand you measure that there is no difference between a normal turnigy and a reflashed trunigy (i assume both use the atmel chips), but people show videos where the response can be clearly seem to be better when using reflashed ESC's.

Also, I thought that the ESC's were setup to smooth input throttle response rather than provide direct changes - is what you have done proving that this is not actually the case?

ice.

What about this, seems to prove it really does make a big difference.

SimonK_ESC_reflash (1 min 4 sec)

Originally Posted by Z06 Tony

Looks like Turnigy is the best and cheapest for multis. Glad I didnt waste my time with the reflashing stuff. There are some here who swear that reflashed w/Simonk is way better than stock Turnigy.

i'm running stock turnigy on a few quads.. they work ok... but not as well as i have seen some esc's with simon's firmware.

i wish i could flash my turnigy... but they are silab

i think the numbers we are looking at on paper doesn't necessarily directly translate to flight performance..

Transfer function is the measure of small signal response where nothing should saturate or limit. These tests used quite large signal changes (25%) which may have taken them outside the small signal region. We often operate ESC's on 'copters outside of the small signal response and move into the large signal area when the input signal slews at a high rate. This can give a completely different and non-linear response.

Some ESC's have digital input filters to limit the rate of change of input demand which when removed does improve response.

I would suggest further testing to evaluate ESC performance over different speed change ranges and under step change and other high slew rate inputs.

Peter

Originally Posted by hugy

What about this, seems to prove it really does make a big difference.

http://www.youtube.com/watch?v=FLKIO...eature=related

Not sure the video proves anything really. I've watched it in the past. It's pretty obvious that the throttle range is different on each test and one seems to be balanced well.

Originally Posted by Peter Seddon

Transfer function is the measure of small signal response where nothing should saturate or limit. These tests used quite large signal changes (25%) which may have taken them outside the small signal region. We often operate ESC's on 'copters outside of the small signal response and move into the large signal area when the input signal slews at a high rate. This can give a completely different and non-linear response.

Some ESC's have digital input filters to limit the rate of change of input demand which when removed does improve response.

I would suggest further testing to evaluate ESC performance over different speed change ranges and under step change and other high slew rate inputs.

Peter

I've measured it using all different values and ranges.

Transfer function is not defined as you suggest.