This is an amazing contribution from Dave Woods on youtube. I thought it worthy of a discussion here. He adds in the comments that results are different for different flight controllers. I don’t know about you guys but for me this seals the deal on PWM rate - I’m headed back to the configurator right now to ditch 32k.
Ha, I just wanted to create a new topic, but I was NotFastEnuf, lol.
Very interesting what @woodsturning is showing us here.
What do we learn from that?
Looking forward to all your opinions, conclusions and discussions my brushed microquad enthusiasts 'round the world.
Let me know if you think this is a logical interpretation: Attempting to update a small brushed motor 32 thousand times a second is nuts especially when the flight controller is only providing new information according to looptime. Any update beyond the speed of looptime is just resending in the same information. Losses increase with increased pwm due to duty cycle? So a balance of matching looptime at the highest pwm to increase accuracy of updated information vs lower pwm to increase efficiency of the driving signal. Maybe even double the looptime at a maximum to account for motor pwm and looptime being out of sync? ( or is there a brushed setting equivalent of gyro sync that I don’t fully understand yet?)
If i say the same thing 3 times fast, its harder to understand then just speaking it 1 time clearly.
With Brushless, you are sending a value that gets computed into an output.
With Brushed, there is no signal, per se, it’s just ON/OFF.
What I am guessing is that with the PWM rate, the signal must go to zero at least once every cycle? So the more cycles per second, the more time you spend at 0. It doesn’t seem like it would need to be this way… but it must be how it’s implemented. Otherwise if the FC was calculating an update every 32uS, and that output was 2000, aka MAX ON, the FC should just turn that mosfet on and keep it on until the next update. And if the next update was also for MAX_ON there would be no difference in thrust.
Don’t confuse it with Brushless updates, because a value is being transmitted and an ouput computed by the ESC. With brushed PWM IS the output. Unsynced brushless 32khz works for a variety of reasons that are less obvious.
Okay so there is no correlation between a brushed pwm rate and looptime.
This is really interesting. Im curious what more testing will bring.
There can be if you run sync’d. But if my memory is correct you could run PWM rate of 32k, long before BF supported unsynced motor updates. So I’m not exactly sure how the brushed PWM mechanism works.
Ok, so now I have 2 of my (SciSky FC) quads pwm set to 1000. I agree with @NotFastEnuf on his interpretation - Why have faster PWM rates than loop times… no sense updating the motor 32 thousand times a second if the FC is only updating 1000 times a second.
I would increase my loop time frequency, but increasing loop time from 1000 to 2000 results in the CPU duty cycle going from 7% to 35% - this to me would signify more battery draw. I also think that flight times will go down due to the increased current draw (according to the chart at the end of the video) at the lower PWM settings.
Going to the field this morning - will report back.
The PWM rate setting with brushed motors is not about how fast the info to the motors is “updated” - but it’s about the “shape” of the info (signal). I hope it’s understandable what I wrote
I thought I read somewhere that the FETs on brushed boards only “update” around 60x a second, so I do not completely understand why we need the pwm rate set so high. I guess it could depend on the board you are using and how the FETs expect to be driven, but I dont really see a reason why we couldnt use a lower pwm rate.
But if you think about it, the duty cycle still be the same either way whether you update 32 times a second or 32,000 times a second. The sine waves just get shorter? Perhaps the motors have more time to develop more thrust when switched on/off slower?
@thexboxblaster - IMHO the shape of the signal shouldn’t change…It should be a square wave where the only change is in the frequency of the pulses, not the shape. I believe it’s just a series of on and off pulses. The quality of the pulse (slew rate) is dependent on the component quality of the FC. I would suspect that the higher the frequency, the higher the slew rate and possibly higher susceptibility to noise. I don’t have an o-scope anymore, so I can’t verify my comments - maybe someone does.
Having had some time to sleep on this, I now theorize the answer lays in how the fet responds to this pwm rate and that here in describes the different performance of boards with different fets. This behaviour of the fet is then displayed by the motor. Curiosity drives me towards a fet data sheet to attempt to interpret these results there somehow.
Ok having done some brief reading to refresh myself on the behavior of fets, there is a cumulative effect of time to turn on and time to turn off the gate with higher pwm driving rates where essentially nothing happens explaining increasing losses in efficiency. The rise time of the gate is also somewhat dependent on the impedance of the circuit in which it is wired as a function of the capacitance of the gate multiplied by the value of the resistor it is switched through. This gives further opportunity to explain the different performance of different boards based on how their circuits are constructed in addition to the specific selection of fets. What I find so interesting next is the different way this behaviour is displayed over different motors. IE - the bigger spread on 6mm motors.
Although without doing the math on the equations to see if this is all that accounts for the losses, it still also leaves a possible contributing amount of loss from how the physical construction of a motor responds to the signal from the fet.
Think I’ll just accept it at this point and circle back around to thanking @woodsturning for identifying this in a tangible way.
@chime13 - by shape I meant the density of the pulses (frequency), but now I see you understand it well, better than I do Thanks for the note about slew rate - always nice to learn something new
I like a scientific approach such as this to look into effects that can be measured. But, flying a quad is partly about feel including noise, you can hear a well tuned quad as far as pids go, it also applies to pwm rate. Try flying at 500, the lowest that brushed should work ok at, then try 1000, 2000, 3000 etc until you find a spot your quad feels best at, that includes noise as well as reaction to controls. When I tried this approach I stopped at 16000 as I was aiming to stop as low as possible. I know higher pwm rates decrease the efficiency of the system. All the way up to 10000 I’m sure I felt constant, slight improvement. After that it was hard to say there was much difference though I kept going, settling at 16000. All my brushed fleet now get set to 16000 and I feel no need to change my habits now
Just thought I’d share my experiences, I still feel fine recommending 16000 to noobs, would never say 32000.
That’s a very good approach. Seat of the pants feel should follow scientific theory for the best end result. I look forward to doing the same!
4k-12k can be particularly annoying sonically. I usually use > 16k because for some reason at 43 I still have exceptional ability to be annoyed by high-pitched sounds. I would have thought all the trap & skeet shooting and amateur pyrotechnics I did as a kid would have “fixed” that.
Ok - my takeaway so far… I was able to fly my 6.5mm (fast) tuned to 1000pwm and other than the noise, it felt more powerful off the line - until I had motor failure - end of experiment for this one.
My umsmile 110 with 8.5 (Insane) motors was noticeably noiser when spinning up at a low speed - the punch is spectacular but I notice I have a harder time finessing it at low speed as compared to my default setting of 16Kpwm. For S&Gs I may just put it at 8K and split the difference…
More experimenting needed (at least for me)
That said and done - I feel exactly like @Chaotix states - keep adjusting until you like the feel.
Great discussion guys! I have read in some of the other threads here about the PWM rate for brushed motors and this thread here is getting to the meat and potatoes of it. @woodsturning has done a lot of micro motor testing and if you look into his RCG threads you will find that he did many tests of the micro motors when they were first becoming available to the public or I guess I should say when they were first becoming popular about 3 years ago. Dave is always giving to the community. One of the YouTube channels I subscribe to and I always enjoy watching his stuff whether it be just some flight footage or a tutorial. He flies these brushed quads and makes them look like a brushless, that always amazes me.