help with motor options

yay

yay

Well-Known Member
my question is, where, how does one go about discerning classes of motor sizes?

eg. my first quad has 7mm motors, which are celebrated as bigger than the ones on the previous model.. sounds logical.. but i'm looking for a larger build, and most of the models i'm loo-king at have 1806 motors.

you go to google, put in 1806 motors, there are 1806 motors from various manufs.

so apparently, 1806 means something, or i wouldn't be surprised, two pieces of data.


so, you see, right now, i have in my mind, a continuous line, possibly a plane or 3 or higher dimensional space, mapping two points in it: 7mm and 1806. i can conjecture the existence of 6mm et c. fairly easily, and place these on the other side of the 7mm from the 1806.

what i'd like to do is identify real objects in my postulated motormappingspace existant between these points... everything i've seen with 1806 on it tears along. that's great, but less speed and more cost effective/lighter may well work for me.

i hope you can see how i can think these hideous, forbidden things before some lovecraftean horrors consume me. in the modern era, the only method i can think of for deriving categorisation is to browse retailers. i'm on the hobbykink (unintentional honest) page right now, but it's not much faster than banggood and quite honestly, i sort of feel like they'd both like me to die before i find anything i can read because it finally loaded. almost like my prose.

another question i have is about flight controllers - say i fall in love with the fcb on my eachine e011 and want to have its babies. youtube tells me the e011 can handle 7.4v. what if i throw the e011 fcb on a larger frame with some 1806 motors. am i now becoming cheerfully happy? teh funny for sure.

should i jsut follow a gremlin build and speak only when i am spoken to and spends all of my monie faster? or can i become a sleek and powerful knower of some things. stay tuned for mooo
 
yay said:
my question is, where, how does one go about discerning classes of motor sizes?

eg. my first quad has 7mm motors, which are celebrated as bigger than the ones on the previous model.. sounds logical.. but i'm looking for a larger build, and most of the models i'm loo-king at have 1806 motors.

you go to google, put in 1806 motors, there are 1806 motors from various manufs.

so apparently, 1806 means something, or i wouldn't be surprised, two pieces of data.


so, you see, right now, i have in my mind, a continuous line, possibly a plane or 3 or higher dimensional space, mapping two points in it: 7mm and 1806. i can conjecture the existence of 6mm et c. fairly easily, and place these on the other side of the 7mm from the 1806.

what i'd like to do is identify real objects in my postulated motormappingspace existant between these points... everything i've seen with 1806 on it tears along. that's great, but less speed and more cost effective/lighter may well work for me.

i hope you can see how i can think these hideous, forbidden things before some lovecraftean horrors consume me. in the modern era, the only method i can think of for deriving categorisation is to browse retailers. i'm on the hobbykink (unintentional honest) page right now, but it's not much faster than banggood and quite honestly, i sort of feel like they'd both like me to die before i find anything i can read because it finally loaded. almost like my prose.

another question i have is about flight controllers - say i fall in love with the fcb on my eachine e011 and want to have its babies. youtube tells me the e011 can handle 7.4v. what if i throw the e011 fcb on a larger frame with some 1806 motors. am i now becoming cheerfully happy? teh funny for sure.

should i jsut follow a gremlin build and speak only when i am spoken to and spends all of my monie faster? or can i become a sleek and powerful knower of some things. stay tuned for mooo
Click to expand...
Anyone ever told you that you talk to much.
 
cool, west papuan colours!

rockpools moz, you wanta hear some of these video guys "let's go for a flight, oh also.. (ffwd 4 mins) well let's go for a flight" then you realise you've been staring at some dudes hands for about four hours and he hasn't even asked you out yet.

i like the one guy i call him the "as you can see" guy. "as you can see..." (talks for three minutes about the thing you can see) ffwd and right on cue "as you can see..." he's on about some other thing you can just see.

maybe if i destroy the entire universe by talking too much here, enough attention can be given to redact all of these human travesties all over the world and save the entire planet to be safe for life again.

i remember some guy died from laughing for 20 minutes during the goodies kungfu capers and there's a special word for laughing death. i wonder if there's a special word for death from misery having to hear someone tell you about how you can open the tape on a box with your thumbnail more times than there is sand on all the beaches


............
guessing that swapping bigger motors as i mentioned would require a voltage stepper or mucking about in the flight software for amperage or something. im gonna have to get one of them cable thingys so i can do things as well instead of just not know about them.

buy yeah i do talk a lot. i need to check out the betaflight stuff before i talk more tho, but seriously one reason is because you know, there's a lot of people today who are just what they are told to be, like i look at a PID and i think... doesn't anyone know about leaky integrators??? simple IIR audio filters would allow the PID to be tuned spectrally (such as... reducing reaction to long time data or short time data or n time data) and i think, well i guess all of these guys must be engineering studnets and haven't lived long enough to do something daft like that. sound it out "lee" "key" + integrator = you're golden.

sod PID, custom algorithm esse. the one and only there is no other holmes.
 
just for fun

dsp is real easy. your 8 year old can handle it kinda stuff.

in brief, integration is a calculus term. fortunately, it means what it sounds like. it *integrates* somethnig into something else... what???? like.. if you integrate a student into a class, you add the student to the class...

so an integrator just adds stuff.

so if we have:
x = x + y;

x is the integrator, y is some new stuff.

you can see that if you do this loads of times, x will eventually be a big old giant thing unless the incoming y values are well balanced between + and - (unlikely, especially for real world applications).



a *leaky* integrator leaks. it looks like this:

x = x + y;
x = x * 0.99;


so if the input (y) is 0 for a while, x will eventually shrink down a whole bunch right.


functionally, it would bias the PID towards "only considering more recent input".






you can calculate the shrinky rate and get exactly the gain value you need. like in reverb algorithms, we talk about a room's T60, which is the amount of time a reverberating signal takes to drop -60dB. -60dB happens to be a value of 0.001....... say we are doing digital audio with a 44100 sample rate and want a T60 of 2 seconds .....

gain_value = pow(.001, 1/88200);
gives us a gain value of 0.999921683846...

..example, if you set z at 1.0, then multiply z by 0.999921683846... 88200 times, you will get 0.001 ;)
 
:rolleyes: .... Apparently, Mr. Chatty, you don't fully understand how PIDs work.
It is, of course, an acronym for Proportional Integral Derivative.
The "short time data vs long time data" factors you refer to are a function of the integral aspect, as sampled over a given time (no calculus needed and yes, I DO know calculus too)
4adf00a8a26fd6b9d052e81607ea15c2af9d6122


With the integral values determined, the derivative is called upon to invoke the correction of that sample while effectively limiting any overcorrection.
9315f1516ee5847107808697e43693d91abfc6e8


Upon which the correction is applied via the proportional aspect, the action of which is relatively self explanatory (Note that the derivative is somewhat predictive in it's sampling indications)
2fc702045b95b36a9bd694c6395cdaf1f60ff570


Which makes for a highly reactive, very adjustable (and yet still stable) control loop feedback mechanism which can be summed up as:
cd581e5c8539ce46453574d1188bd9d52a610fe0
:eek:

What does all this math mean to the average quadcopter pilot ? o_O

Simply put:
"P" is the most important value to worry about (in fact both "I" and "D" can be zero and the quad will still fly).
It represents the REACTION to inputs (your's and the PID's).
The higher a P value is the sharper the control is, while a lower value yields softer control (think of it like "twitchy" versus "lazy").
Too high of a P setting can lead to overcorrection and wobbling though.

"I" is the self correcting against outside forces, like wind for instance.
It tries to keep the quad as it is.
If you find that your quad is kinda drifty then up your I value.

"D" tries to soften the abruptness of P's actions.
If you notice some minor oscillations, especially when performing acro maneuvers, you might want to bump up D a little bit.
Too much D makes the quad feel sluggish though and can also amplify noise in the control system which will then reintroduce oscillations !

Turn in your homework and school is out. :cool:
 
OMG they will all be at it.
Soon I won't understand anyone.
Yay sounded like he knew what he was on about till that last paragraph. But I know he's wrong.
 
PID offers a parametric (meaning, "it has parameters") method that can be modified for serviceable results by an end user, obviously we know it's a viable method.

the basic PID algorithm can be refined. it is an IIR filter, and the topology ("shape" or how we are doing this) can be modified for improved suitability. when i describe the signal as spectral, this means eg. if we take a fourier transform we can observe changes in the signal that are slower and faster.

right now, the only spectral availability to the end user is the balance of P to (I and D).

using spectral methods we can refine the responsivity of each element of the PID separately, or with as much complexity as we can conceive and put on a fcb, so that eg. "if it is up too high and gets oscillations" we can refine the threshold and then eg. derive spectral components pertinent to the build.

for instance, frame resonances produced by material m of dimension n. maybe not a pragmatic example but you can see.

air is a substance, propellors et al have substance, and transmissiveness has frequency dependent qualities. again, this may not be practical when dealing with an item that only performs several dozen rotations per second but practically, the use of simple lowpass or highpass filtering of PID parameters may serve to increase performance boundaries ;)


PID is simple, PID works fine. why can't PID be improved, just because many people find it adequate?

in digital audio, the biquad filter topology was dominant because it offers a cost effective and flexible solution, however it introduces group delay (another word for that frequency dependent quality) so that eg. signals may have up to a 90 degree phase offset. 90 degrees of 10kHz is 0.000025s, maybe not a worry, but 90 degrees of 10Hz is 0.025s, which may be more applicable.

now most people, "Concert A" being say 440 Hz, can't be bothered with a 1/4th of a single wavecycle. but the audio industry *can* because people with very good attention to sound can hear the difference in the response of *resonant filters* so that eg. if you modulate the frequency of a resonant filter on a moog vs. a biquad digital moog filter emulation, that 90 degree delay will produce a chirp which only people who spend all their time comparing analog vs. digital filters will care about.

about a decade ago the "zero delay feedback" or 0df topology started to offer a solution which ditches the frequency dependent/group delay for under twice the amount of computation. you're probably familiar with a 2d rotation matrix... there's a more efficient form that uses half as many computations ..... s0 -= s1 * w; s1 += s0 * w; (initialise s0 = 1, s1 = 0) which will output a signal in quadrature, just like a unit vector rotated by a matrix, but with only two mults (it can be cooked down to one!) ---- however, you'll note that the s0 has a half step delay ... group delay ;) fortunately, you're familiar with the matrix form than this approximation.

quad pilots are likewise the only people who would have the refined discretion to discern how performance can be refined. that place where a single scalar is about as much as the system can take could possibly be improved if we reduce its sensitivity to long time signals and thus refine the temporal sensitivity of the algorithm ;)

we want to do this because the medium we are acting within, also has frequency dependent properties. like cancelling a resonance from beam vibration or motor responsivity i guess perhaps. i've been modeling acoustic air and material vibration for some time, so i'm interested to see what mucking aroudn with it does.


but yeah, i get PID, what i don't know about is what other folks are doing. i did see a note on some firmware where a filter topology was changed, so certainly i would assume flight controllers in current use aren't just rotation matrices and PIDs, and i'm certainly not here to insist on anything.

but no one ever found out from me not talking about something ;)

eg. reading through the PID for arduino library, the algorithm includes parameter modulation..... in audio, control signals are lowpassed when stepping is problematic. simple to do, state += input + w * state;
 
Last edited:
seriously, if someone had an idea about a performance deficit and were able to measure things pertinent to that perception, you could derive a fourier convolution ("massive spectral refinement over temporal responsivity").
 
the simplest way to understand this is,

oscillations ... being frequency specific, can be cancelled out by bandpassing ;)
 
won't let me edit, lp for parameter smoothing is of course state += w * (input - state); with small values for w (angular frequency).
 
moz, you can use vitex agnes-castus seed in moderate amounts to modulate your libido and get back to you other head mate. or a picture of gary glitter might do it. wait - you ever watch the long version of aphex twin windowlicker?
:D nsfw

gotta mention premptively, i got some kinda mental problem (you can tell) wheres i dislike capitalising. every tech field i've ventured into someone sees me as competition and starts making my life miserable. last time i made a little 2d app and got pictures of my home mailed to me. i'm not gonna capitalise or git famous. this is something to pass time on my way to teh grave. i iz only here to make a couple flyers for security and canyoneering and familiarise myself with rc and pfv. gonna take a while on current economy.
 
:eek: ... WTH have you been smoking ?! ... You had better pull that nasty video asap before you get yourself banned from here !!!

Fortunately (or perhaps unfortunately) I am fully cognizant of the "tricks" to which you refer BUT bandpass filtering, phase shifting, vector algebra, matrix manipulation, et al
would also directly impact a quad's flight behavior in an unfavorable way. All that blather you've spewn (while completely accurate) is pretty much besides the point (imho) when
it comes to flying quadcopters. Clean amplifiers or precision controllers like CNC and/or nanotech applications perhaps, but for quadcopters not so much.

While it's true that resonant frequencies (vibrations or oscillations in our application) can be filtered out, what you're overlooking is that any dynamic parametric equalization effect directly affects the performance characteristics of the quad, wherein "dead bands" or at least "muted areas" (for lack of better terms) would be introduced into the
flight envelope (and at arbitrary points no less, dependant upon the virtually innumerable variables associated with said sources) as that method of compensation became engaged. Parameter modulation and fourier transforms are (at least in my understanding) used for more accurate signal analysis, which is not to say such methods should necessarily also be applied to pilot control inputs or FC responses to those inputs.

In other words, and much more simply put, if I'm in a high rate knife edge turn I do NOT want my FC to cut my throttle back to eliminate harmonic vibration encountered due to
frame stresses.... Here's another scenario: I'm inverted then moving nose down as I'm performing an reverse Immelmann only to find my expos have been automatically tweaked
by some frikkin compensation scheme that's engaged ... end result, I hafta dig up my quad to scavenge parts, but hey it didn't wobble one bit as it power dove right into the ground.

I'm an advocate of the "KISS" method (keep it simple stupid) whenever practical and Lord knows tuning (and the understanding of) PIDs is already one of the most complex
aspects of being a well rounded quadcopter pilot that already exists. In fact I'd dare say that most pilots have much less an understanding of them than you've demonstrated,
so I stand corrected in that regard. That being said, I still maintain that the rather static approach (PIDs as they are) would be much more preferable (being predictable) to a more dynamic (unpredictable) scenario as you propose. Keep it simple (and you're certainly not stupid, so I'll just leave it at that).) :)
 
Back
Top