Tail PID tuning??
Posted by johnahamelv
|
Tail PID tuning?? June 12, 2024 10:47PM |
How to tune Proportional, Integral, and Differential values? (I know that D stands for "derivative", but VBar calls it "differential"
Default values for Proportional, Integral, and Differential are "80", "60", and "0" respectively for 450 and larger size helicopters. Is there a systematic way to tune these values? Are there any particular maneuvers accentuated by each value? How do each parameter manifest when the values are too high and/or too low?
Default values for Proportional, Integral, and Differential are "80", "60", and "0" respectively for 450 and larger size helicopters. Is there a systematic way to tune these values? Are there any particular maneuvers accentuated by each value? How do each parameter manifest when the values are too high and/or too low?
|
June 13, 2024 09:40AM |
Hi,
basically, you can look up PID control loop, on wikipedia, for better understanding (if necessary).
The term differential (from differentiator) is also not uncommon, in different (derivative?) languages.
Long term monitoring and learning brought up the default values für P and I, as a known-good and broadband starting point, and also a good ratio to maintain, if you make changes.
Interestingly, this works on almost all tails, flawlessly.
One benefit of the VBar, you don't have to learn control loops and dig into it, and (also because of it's adaptive components), it works from small to big, with all kinds of servos, tail blades etc. But of course you can.
Proportional is obviously a very quick component, which gives you quick and solid control (feels connected).
Too high, and the tail will start to oscillate quickly in some situations (P-oscillation). Way too high, it can even be destructive.
Too low (also in relation to I), control response becomes disconnected, sluggish. Only I would mean, you make a suggestion (rotate the course marker on a compass or auto pilot, and the I will eventually rotate the model there).
I is obviously the heading hold, which tries to iron out rate deviations (in controlled pirouettes) as well as it tries to maintain the direction ((almost) true heading hold, e.g. in sideward loops).
Too high, and you will notice a slow tail oscillation (I-wag), or particularly on scale ships with high inertia and lesser tail authority, an inconsistent pirouette or spiraling rate.
Too low (also in relation to P) will lead to inconsistent piro rate, or the model will weather-vane more easily.
D makes everything crisper, this can be used (and be useful) if the actuator is very quick.
Else, it can quickly lead to an increasing high frequency oscillation which, if no plan B, can even be destructive.
D can make the model feel even more locked-in/connected/crisp, to the point where the mechanics are overstressed.
Normally, you hear that oscillation coming up, it before you see it.
Then, there are boundary conditions, like stop behaviour, stop gain, which accentuate only that. Or the optimizer, which—if possible at all—equalizes the effect of stops with/against torque. I control can be limited, and I can be discharged (fed back to zero, over time), to accommodate weaker tails (also fenestrons), or to allow weather-vaning while still having the benefits of heading hold.
All that can be understood as an equation which is amplified or relaxed by the overall gain.
And it all kind of depends on the actuator, read: servo (specs, pulse, refresh rate), control rod, bell crank, tail rotor control, tail blades, authority by means of tail rpm, maxima for tail pitch angles (a stall is a stall, too little authority is what it is, both can't be overcome with software).
Technically, the control loop and a quick and strong servo are easily capable of destroying a tail, since because of inertia and other effects, the tail can never be infinitely quick.
Adaptive programming helps with that, in a way so we can detect e.g. the infamous overspeed maneuvers (with help of the VBar Governor, amongst others), and make other adjustments 'on the fly'.
If you are curious, I guess the best thing to wrap your head around it is, try a couple of things, and always have one safe bank to switch back to, just in case.
First, do a default setup including trim flight (to find the center position for the integrator, mainly), and an optimizer flight (to equalize stops with/against torque), adjust the overall gain to your liking, all until you feel comfortable.
Copy these settings to all banks.
Then, lower P 10-20-30 points (use two banks), leave everything else, do your thing in all three banks. Notice the difference?
P back to default, lower I 10-20-30 points, leave everything else, do your thing. Notice the difference?
All back to default, increase D 5-10 points, do your thing. And?
Should anything weird happen, disconnected feeling, lack of control, P-oscillation (e.g. in fast forward flight), D-oscillation (can happen any time, and quickly, once the control loop starts struggling with itself), switch to your safe bank.
Last but not least, vibration has also a hand in it: if the sensor readings and thus it's constant feedback had interference, all kinds of weird things could also happen, and—worst case—also be amplified beyond what the mechanics can take, by maxed-out parameters, or the tail may feel less locked in, disconnected at times.
Have fun
—Eddi
Born to fly ...
forced to work.
Edited 1 time(s). Last edit at 06/13/2024 09:41AM by Eddi E. aus G..
basically, you can look up PID control loop, on wikipedia, for better understanding (if necessary).
The term differential (from differentiator) is also not uncommon, in different (derivative?) languages.
Long term monitoring and learning brought up the default values für P and I, as a known-good and broadband starting point, and also a good ratio to maintain, if you make changes.
Interestingly, this works on almost all tails, flawlessly.
One benefit of the VBar, you don't have to learn control loops and dig into it, and (also because of it's adaptive components), it works from small to big, with all kinds of servos, tail blades etc. But of course you can.
Proportional is obviously a very quick component, which gives you quick and solid control (feels connected).
Too high, and the tail will start to oscillate quickly in some situations (P-oscillation). Way too high, it can even be destructive.
Too low (also in relation to I), control response becomes disconnected, sluggish. Only I would mean, you make a suggestion (rotate the course marker on a compass or auto pilot, and the I will eventually rotate the model there).
I is obviously the heading hold, which tries to iron out rate deviations (in controlled pirouettes) as well as it tries to maintain the direction ((almost) true heading hold, e.g. in sideward loops).
Too high, and you will notice a slow tail oscillation (I-wag), or particularly on scale ships with high inertia and lesser tail authority, an inconsistent pirouette or spiraling rate.
Too low (also in relation to P) will lead to inconsistent piro rate, or the model will weather-vane more easily.
D makes everything crisper, this can be used (and be useful) if the actuator is very quick.
Else, it can quickly lead to an increasing high frequency oscillation which, if no plan B, can even be destructive.
D can make the model feel even more locked-in/connected/crisp, to the point where the mechanics are overstressed.
Normally, you hear that oscillation coming up, it before you see it.
Then, there are boundary conditions, like stop behaviour, stop gain, which accentuate only that. Or the optimizer, which—if possible at all—equalizes the effect of stops with/against torque. I control can be limited, and I can be discharged (fed back to zero, over time), to accommodate weaker tails (also fenestrons), or to allow weather-vaning while still having the benefits of heading hold.
All that can be understood as an equation which is amplified or relaxed by the overall gain.
And it all kind of depends on the actuator, read: servo (specs, pulse, refresh rate), control rod, bell crank, tail rotor control, tail blades, authority by means of tail rpm, maxima for tail pitch angles (a stall is a stall, too little authority is what it is, both can't be overcome with software).
Technically, the control loop and a quick and strong servo are easily capable of destroying a tail, since because of inertia and other effects, the tail can never be infinitely quick.
Adaptive programming helps with that, in a way so we can detect e.g. the infamous overspeed maneuvers (with help of the VBar Governor, amongst others), and make other adjustments 'on the fly'.
If you are curious, I guess the best thing to wrap your head around it is, try a couple of things, and always have one safe bank to switch back to, just in case.
First, do a default setup including trim flight (to find the center position for the integrator, mainly), and an optimizer flight (to equalize stops with/against torque), adjust the overall gain to your liking, all until you feel comfortable.
Copy these settings to all banks.
Then, lower P 10-20-30 points (use two banks), leave everything else, do your thing in all three banks. Notice the difference?
P back to default, lower I 10-20-30 points, leave everything else, do your thing. Notice the difference?
All back to default, increase D 5-10 points, do your thing. And?
Should anything weird happen, disconnected feeling, lack of control, P-oscillation (e.g. in fast forward flight), D-oscillation (can happen any time, and quickly, once the control loop starts struggling with itself), switch to your safe bank.
Last but not least, vibration has also a hand in it: if the sensor readings and thus it's constant feedback had interference, all kinds of weird things could also happen, and—worst case—also be amplified beyond what the mechanics can take, by maxed-out parameters, or the tail may feel less locked in, disconnected at times.
Have fun
—Eddi
Born to fly ...
forced to work.
Edited 1 time(s). Last edit at 06/13/2024 09:41AM by Eddi E. aus G..
|
Re: Tail PID tuning?? June 15, 2024 06:59AM |
Clear as mud. 
|
June 17, 2024 09:01AM |
Hi,
to put it simple
. make best possible mechanical adjustments (not yet powered), 90° angles at levers, 2-3° tail pitch angle at servo center.
. select proper servo specs (center pulse, refresh rate closest to max. the servo can take)
. powered, adjust sensible and similar limits, rotate servo arm one notch if it helps with similar limits
. pre-adjust center again using tail trim flight value if you like
. adjust gain to sensible value (lower for small helis, higher for larger helis, 60-90, for starters)
. do trim flight
. do optimizer flight
. adjust stop gains to equalize stops (could be necessary to relax stronger side, if weaker side can not do more)
. max out overall gain if you really like, but not necessarily necessary if the tail holds anyway
. this can all be done within one battery's flight time
Works under almost all circumstances.
—Eddi
Born to fly ...
forced to work.
to put it simple
. make best possible mechanical adjustments (not yet powered), 90° angles at levers, 2-3° tail pitch angle at servo center.
. select proper servo specs (center pulse, refresh rate closest to max. the servo can take)
. powered, adjust sensible and similar limits, rotate servo arm one notch if it helps with similar limits
. pre-adjust center again using tail trim flight value if you like
. adjust gain to sensible value (lower for small helis, higher for larger helis, 60-90, for starters)
. do trim flight
. do optimizer flight
. adjust stop gains to equalize stops (could be necessary to relax stronger side, if weaker side can not do more)
. max out overall gain if you really like, but not necessarily necessary if the tail holds anyway
. this can all be done within one battery's flight time
Works under almost all circumstances.
—Eddi
Born to fly ...
forced to work.
Sorry, only registered users may post in this forum.