News:


  • June 15, 2024, 01:11:07 PM

Login with username, password and session length

Author Topic: Governor Bandwidth/response characteristics?  (Read 3209 times)

Online Brett Buck

  • Trade Count: (0)
  • Admiral
  • ******
  • Posts: 13792
Governor Bandwidth/response characteristics?
« on: October 12, 2012, 03:20:24 PM »
Does anyone have an idea what sort of control response the various governor control laws have? For instance if I apply some load, how long does it take to start jacking up the current to restore it?

    Brett

Offline Tim Wescott

  • 2016 supporter
  • Trade Count: (0)
  • Admiral
  • *
  • Posts: 12832
Re: Governor Bandwidth/response characteristics?
« Reply #1 on: October 12, 2012, 03:50:37 PM »
Failing that, if anyone has some readable plots of RPM and current vs. time from a logging ESC, we can at least get an estimate.  The best place to look would probably be the last corner of a square maneuver, or a triangle, or the hourglass.

If someone knows how to coax actual text files of numbers or Excel spreadsheets out of Castle's logging software, I'll run it through my analysis software here and see if I can get anywhere with it.

It would be interesting to compare a Castle ESC in governor mode vs. a KR timer.
AMA 64232

The problem with electric is that once you get the smoke generator and sound system installed, the plane is too heavy.

Offline Jason Greer

  • Trade Count: (0)
  • Captain
  • *****
  • Posts: 486
Re: Governor Bandwidth/response characteristics?
« Reply #2 on: October 12, 2012, 08:30:39 PM »
Interesting question Brett.  A few days ago Peter Germann posted a couple of RPM graphs showing the RPM deviation of both a high gain and low gain setup.  Of course the high gain held the RPM much closer to the target RPM. 

Would the gain have an effect on the response time or does this only serve as a parameter indicating when the ESC will react to the delta between the measured RPM and target RPM?  For example, with a low gain the ESC might only regulate when the RPM delta is 500 RPM and a high gain might regulate when the RPM delta is 100 RPM.  In both cases the actual response time would be identical once the regulation actually happens, but the delay would be in the time it takes for the RPM to change from the target to +- 100 and the target to +-500.  I suppose it would also take longer for the RPM to return back to the target if the difference is 500 RPM versus 100 RPM.

Would the actual response time, regardless of the gain, be determined by the PWM frequency?  For example, a 16khz PWM would react twice as fast as an 8khz PWM.

Obviously I have a lot to learn, but these topics are very intersting.   

Thanks,
Jason
El Dorado, AR
AMA 518858

Offline Andrew Borgogna

  • Andy
  • Trade Count: (0)
  • Admiral
  • ******
  • Posts: 1188
Re: Governor Bandwidth/response characteristics?
« Reply #3 on: October 12, 2012, 08:49:32 PM »
I will send Keith Renecle at KR timer/governor and ask him to join in.
Andy
Andrew B. Borgogna

Offline schuang

  • Trade Count: (0)
  • Ensign
  • **
  • Posts: 43
Re: Governor Bandwidth/response characteristics?
« Reply #4 on: October 12, 2012, 08:52:01 PM »
Failing that, if anyone has some readable plots of RPM and current vs. time from a logging ESC, we can at least get an estimate.  The best place to look would probably be the last corner of a square maneuver, or a triangle, or the hourglass.

If someone knows how to coax actual text files of numbers or Excel spreadsheets out of Castle's logging software, I'll run it through my analysis software here and see if I can get anywhere with it.

It would be interesting to compare a Castle ESC in governor mode vs. a KR timer.

It is hard to compare between two since esc portion is different.  However, I suspect cc is faster since the feedback loop is shorter.

Regarding to the settling, I believe the motor (electrical to mechanical) plays a big role.  The dominant pole is the motor for this feedback system.

Regards,

Sean

Offline Tim Wescott

  • 2016 supporter
  • Trade Count: (0)
  • Admiral
  • *
  • Posts: 12832
Re: Governor Bandwidth/response characteristics?
« Reply #5 on: October 12, 2012, 09:56:02 PM »
Would the gain have an effect on the response time or does this only serve as a parameter indicating when the ESC will react to the delta between the measured RPM and target RPM?  For example, with a low gain the ESC might only regulate when the RPM delta is 500 RPM and a high gain might regulate when the RPM delta is 100 RPM.  In both cases the actual response time would be identical once the regulation actually happens, but the delay would be in the time it takes for the RPM to change from the target to +- 100 and the target to +-500.  I suppose it would also take longer for the RPM to return back to the target if the difference is 500 RPM versus 100 RPM.

In reverse order:

If the gain is done in the usual way it is a multiplier from the speed error to the response -- so the governor always responds to a speed error no matter how small, but when the gain is high it responds more strongly.

Because the motor takes a while to respond, a higher gain will not only hold the RPM error to a smaller value, but it'll make the motor respond quicker.

Would the actual response time, regardless of the gain, be determined by the PWM frequency?  For example, a 16khz PWM would react twice as fast as an 8khz PWM.

Only when the PWM gets really slow.  Until then the limiting factor is the motor, or (in the case of a KR timer) the maximum frame rate that you can pump out of the timer and still have the speed control understand what the heck is going on.
AMA 64232

The problem with electric is that once you get the smoke generator and sound system installed, the plane is too heavy.

Offline Jason Greer

  • Trade Count: (0)
  • Captain
  • *****
  • Posts: 486
Re: Governor Bandwidth/response characteristics?
« Reply #6 on: October 12, 2012, 10:38:05 PM »
Thanks for that Tim.  A multiplier makes sense.

E stunt has been working great for me for the last two years.  I've learned what numbers to plug into the software to get the desired results, but I've never taken the time to really learn how those numbers are actually applied to the controller logic. 
El Dorado, AR
AMA 518858

Offline Keith Renecle

  • Trade Count: (0)
  • Captain
  • *****
  • Posts: 894
Re: Governor Bandwidth/response characteristics?
« Reply #7 on: October 12, 2012, 11:20:03 PM »
Interesting subject. I'm more of a "suck-it-and-see" gadget person than a proper engineer. When I started out with the idea, Igor mentioned that he thought that an external governor may just be a bit too slow but that I should try it anyway........so I did. The charts from my Eagle Tree data logger seemed to show that it worked well enough and from flying it and comparing it to my Jeti Spin at the time, it felt similar, so I carried on with the project. I'm pretty sure that it must be slower than an internal system, but the question is....how much difference does it make in practice?? That I don't know.

The Eagle Tree brushless rpm sensor gets rather spiky so it is difficult to work with. I need to try feeding the logger from my timer with a cleaned-up signal. I can most likely do this with the retract output on my pc board so I'll give this a go sometime soon. To answer Tim on my sampling rate, I use the 20 ms frame rate. I've been told that a governor loop at 50 times per second is pushing my luck, but so far it is working o.k. I'm playing around with this continually so Brett's question (as always) is something I'd like to find out.

Keith R
Keith R

Offline Peter Germann

  • Trade Count: (0)
  • Captain
  • *****
  • Posts: 401
Re: Governor Bandwidth/response characteristics?
« Reply #8 on: October 13, 2012, 04:11:07 AM »
Brett, attached is a number of logs recorded recently. Shown is RPM during five laps of level flight (left) and at time of the wingover manoeuvre (right) with Governor Gain set to min. respective max. values.

Airplane:  Max Bee  64 oz
Motor: AXI 2826/12
ESC: Castle Phoenix ICE lite 75A
Timer: Hubin FM-9
Timer signal: steady at 1.433 ms (for 9'624 RPM)
Battery: 5S2600 60C
Wheather:  light wind with low gain and calm with max. gain
Propeller: APC 13 x 5.5 EP
RPM set:  9'624
Lap (64ft x 0.015): 5.25"
ICE lite 75 A Governor Gain : Low (15) (Minimum) or Custom (50) (Maximum)  
Head Speed Change Rate: Custom (20)
Motor Timing: Low (0) with min. gain and Custom (2) with max. gain

The throttle signal was recorded as a flat and steady line at 1.433 ms coming from a Hubin FM-9 timer operating in New Phoenix High mode.

It's quite interesting to note that the only significant change resulting from switching from lowest to highest gain is the degree of RPM fluctuation in level flight (+/- 160 RPM vs  +/- 70 RPM) Recovery time seems, within the limits of the inaccurate measuring method used, to be almost the same.

Peter
Peter Germann

Offline Tim Wescott

  • 2016 supporter
  • Trade Count: (0)
  • Admiral
  • *
  • Posts: 12832
Re: Governor Bandwidth/response characteristics?
« Reply #9 on: October 13, 2012, 09:55:54 AM »
Peter: you must live in a rural state -- it looks like your governor is hunting.  (At least in high gain)

Keith: I wouldn't say that a low frame rate is pushing your luck, unless you're trying to achieve some specific loop bandwidth.  It may end up being a limiting factor in the bandwidth (and hence the speed and accuracy that you can achieve), but as long as you always tune things for stability then nothing worse than disappointment is going to happen.

The rule of thumb for bandwidth in a sampled-time system is that you should design for sampling at least ten times faster than your desired bandwidth (or a sample interval at least 1/10th your desired settling time).  If your back is to the wall, and if you get lucky, and if you accept some bad behavior, you can sometimes get better than this.  It's probably safe to assume, though, that Keith's bandwidth is no more than 5-10Hz.
AMA 64232

The problem with electric is that once you get the smoke generator and sound system installed, the plane is too heavy.

Offline Keith Renecle

  • Trade Count: (0)
  • Captain
  • *****
  • Posts: 894
Re: Governor Bandwidth/response characteristics?
« Reply #10 on: October 13, 2012, 10:19:18 PM »
Hi Tim, thanks for that info. A bandwidth of 5~10 sounds o.k. and in practice it seems to be doing a good job. So much to learn and I'm having fun along the way, thanks to you guys on this forum! Those charts from Peter are most interesting as well. Maybe someone can do some comparisons to engines as well. I think that it was Alan Hahn that posted some charts on engines. I really miss Alan's posts on this forum...........time to come back please Alan if you're reading this!

keith R
Keith R

Offline Peter Germann

  • Trade Count: (0)
  • Captain
  • *****
  • Posts: 401
Re: Governor Bandwidth/response characteristics?
« Reply #11 on: October 14, 2012, 04:13:32 AM »
Peter: you must live in a rural state -- it looks like your governor is hunting.  (At least in high gain)

Tim, as you can imagine from the attached picture, our circle here is not always free from turbulence. With both flights recorded having been done when there was wind, the "hunting" shown was perhaps caused by turbulence induced, rapidly changing prop load.

I will be happy to provide those wishing to do further analysis with the excel data of both min. and max. gain flights. Just send me a pm...

Regards,
« Last Edit: October 15, 2012, 03:25:57 AM by Peter Germann »
Peter Germann

Offline Robert-Jan

  • Trade Count: (0)
  • Lieutenant
  • ***
  • Posts: 93
Re: Governor Bandwidth/response characteristics?
« Reply #12 on: October 14, 2012, 05:11:59 AM »
« Last Edit: October 14, 2012, 10:34:52 AM by Robert-Jan »

Offline Igor Burger

  • Trade Count: (0)
  • Admiral
  • ******
  • Posts: 2166
Re: Governor Bandwidth/response characteristics?
« Reply #13 on: October 14, 2012, 09:30:50 AM »
Brett, I think ESC starts with regulation almost immediatelly (it measures rpm several times per one rotation) but important is how quickly really regulates, means how long it takes until RPM come to some at least aproximately wanted rpm. I think a logger working with 0.1s period is too slow to see it. When I play with motor running without prop and I try to brake it by fingers, the reaction is certainly quicker than 0.1s, at leas I cannot recognize any visible time. Backward regulation (after unload) is slower and it tends to overrun, because they typically cannot brake.

ESCs use 2 different syetms of regulation, either PID where P and D gain makes immediate regulation and "I" is little slower and others use slope regulation which can be typically set to minimum of 0.1s, means "some" regulation can be done also in fraction of 0.1s

And regarding Keith system - I think the speed is not only question of his device, it is question of ESC. The regulation loop is longer than heli ESC with governor which is "inside", but he still measure and controlls 50 times per second, so it is quicker then we need, but if ESC has some low pass filter, then the speed of regulation is detemined by that filter. So his times not comes shorter with chip ESC, it even benefits on simple ESC without too clever software with smooothing low pass filters :- ))))

Offline bob branch

  • AMA Member
  • Trade Count: (0)
  • Admiral
  • *
  • Posts: 941
Re: Governor Bandwidth/response characteristics?
« Reply #14 on: October 14, 2012, 12:52:21 PM »
I'd like to point to another area in this study that no one seems to be addressing. Rick Sawicki and I did testing on this several years ago. I found I could feel the motor hunting in rpm when I started to ramp the gain to higher numbers. This pulsing sensation coming thru the lines changed in frequency upwind and downwind. When I lowered the gain it smoothed out to smooth consistent power with no surging. Rick found this to be true on his planes as well. We were able to notice this on planes ranging in size from nobler to T-Rex / Sv-11 sizes and were flying axi 2826/12, castle creation, hubin setups. 

When we went to higher gain settings we did not hear as much motor change as at lower gain levels in maneuvers. Both of us felt less power in climb with higher gain settings as well. When the logging CC esc's became available we were able to confirm this. Peter's plots show it as well. In the high gain plots the rpm spikes on maneuvers are lower both in upline (power increase) and downline (breaking) than in the low gain settings. This is noticeable at the handle as well. We confirmed this on multiple platforms as well. We did these tests in governor hi mode, not fixed rpm. We found the same to happen when comparing governor hi to fixed rpm mode, governor hi mode giving larger power spikes.

The question becomes, what are we looking for, constant rpm, or constant airspeed thru maneuvers. What we look for in the flying part seems to be more power going upline and less power going downline. If you listen to a well tuned piped motor you will hear a considerable lag in the position of the airplane thru the maneuver as to where the power change has happened. I have watched a number of piped flights to evaluate where the power comes on and where in the maneuver the power comes off.  Its an instructional exercise I would recommend. With electric the sound change happens right when the maneuver happens. We can obviously control that change with gain and props. For my money and I think Rick is of the same feeling, I will take the greater amount of power and the greater amount of decreased power. The difference of a few micro seconds I do not think is a determinate to what the airplane is doing. I'm too old to respond in microseconds anyway. Regardless, the results have shown that the planes can be set up to work very well in either situation.

bob branch

Offline Tim Wescott

  • 2016 supporter
  • Trade Count: (0)
  • Admiral
  • *
  • Posts: 12832
Re: Governor Bandwidth/response characteristics?
« Reply #15 on: October 14, 2012, 01:40:04 PM »
Bob: the hunting with high gain is a well-known phenomenon in servo loops.  It is, in fact, one of the biggest challenges a control engineer faces: to get the disturbance rejection (speed changes in our case) low you need a high-bandwidth loop; to get high bandwidth you need high gain, and when you crank the gain up too much you get instability -- which exhibits itself as hunting.

I couldn't tell you exactly why the hunting frequency was different upwind than down.  In a perfectly linear systems (control theorists obsess about linearity a lot) the hunting frequency would be constant with load -- but in a perfectly linear system, the amplitude of the oscillation would just grow and grow.  So something is acting to limit the amplitude of the hunting (possibly a current limit, or the natural voltage limit of the batteries).  Whatever that is also affects frequency.  If you noticed the frequency or other characteristics of the hunting changing as the battery ran down, that would certainly point to voltage limiting.

I'm not quite wrapping my control-loop-oriented head around your comments about airplane speed.  Normally, looser regulation would mean the motor follows its own inclinations more, which in turn would mean it would go slower on the up-line of a maneuver and faster on the down.  So I'm not sure why a more loosely regulated system would feel better in this regard -- unless the hunting was so bad that it was otherwise messing up the regulation: I've certainly seen that happen in a number of control loops, up to and including motors that were supposed to be holding positions bursting into song and effectively going limp, because the oscillation took all the available power with none left over for the motor to do its job.
AMA 64232

The problem with electric is that once you get the smoke generator and sound system installed, the plane is too heavy.

Online Howard Rush

  • 22 supporter
  • Trade Count: (0)
  • Admiral
  • *
  • Posts: 7818
Re: Governor Bandwidth/response characteristics?
« Reply #16 on: October 14, 2012, 02:25:15 PM »
Tim, as you can imagine from the attached picture, our circle here is not always free from turbulence.

In the interest of Science, I suggest the application of a chainsaw.
The Jive Combat Team
Making combat and stunt great again

Offline bob branch

  • AMA Member
  • Trade Count: (0)
  • Admiral
  • *
  • Posts: 941
Re: Governor Bandwidth/response characteristics?
« Reply #17 on: October 14, 2012, 03:16:04 PM »
Tim

My interpretation is that the lower gain results in the governor not correcting as frequently. When it corrects it is seeing a greater drop in rpm and gives a bigger spike, for instance in a maneuver. The corrections are happening more quickly than we need. Going into the wind the plane is going to see more load, thus less rpm, thus an increase in power. Its just what it is supposed to do. Downwind its going to see less load, rpm tries to go up so it decreases power. I guess the issue may be that there will be harmonics in the system and at some gain setting for the entire system, you can sense them thru the handle.

In the tests we did we were not dealing with voltage issues at all. We were using large batteries, 2 minute flights, and the tests were run at beginning of the flight.

bob


Offline Tim Wescott

  • 2016 supporter
  • Trade Count: (0)
  • Admiral
  • *
  • Posts: 12832
Re: Governor Bandwidth/response characteristics?
« Reply #18 on: October 14, 2012, 03:19:29 PM »
Well, if the loop is implemented correctly the governor will always correct as frequently, but with lower gain the system will react more slowly.

(The whole bandwidth/gain thing works the same in a continuous-time system that never samples at all, but is done with amplifiers and capacitors and stuff -- no processors or sampling or anything).
AMA 64232

The problem with electric is that once you get the smoke generator and sound system installed, the plane is too heavy.

Online Crist Rigotti

  • 23 supporter
  • Trade Count: (0)
  • Admiral
  • *
  • Posts: 3864
  • Electric - The future of Old Time Stunt
Re: Governor Bandwidth/response characteristics?
« Reply #19 on: October 14, 2012, 03:34:32 PM »
In the interest of Science, I suggest the application of a chainsaw.

I thought the same thing!
Crist
AMA 482497
Waxahachie, TX
Electric - The Future of Old Time Stunt

Offline Igor Burger

  • Trade Count: (0)
  • Admiral
  • ******
  • Posts: 2166
Re: Governor Bandwidth/response characteristics?
« Reply #20 on: October 14, 2012, 03:48:37 PM »
this is not answer for Brett question, but it can show reality little bit

this graph is from ESC which we tried to use for quicker rpm changes for active regulators, it shows how quickly can ESC CHANGE rpm ... it is clear that the rpm change is short fraction of second ... and it HAS low pass filter, constant rpm regulation of governor is quicker than change of rpm

Online Howard Rush

  • 22 supporter
  • Trade Count: (0)
  • Admiral
  • *
  • Posts: 7818
Re: Governor Bandwidth/response characteristics?
« Reply #21 on: October 14, 2012, 05:16:16 PM »
... constant rpm regulation of governor is quicker than change of rpm

If it weren't, I would like to have one of those motors.  

Is the green line the actual motor step response, or the RPM command output from the controller?    If it is the motor response, does it have a propeller?

Is the red line the recorder's aliasing of the current pulses caused by the motor demand, or is it a result of sensor noise?
The Jive Combat Team
Making combat and stunt great again

Offline Igor Burger

  • Trade Count: (0)
  • Admiral
  • ******
  • Posts: 2166
Re: Governor Bandwidth/response characteristics?
« Reply #22 on: October 15, 2012, 01:17:57 AM »
it was motor without prop, the inertia is only from magnetsand rotor

motor runs in PWM mode, so the red line should be filtered to hide 8kHz chopping

that slope on begin of curve is artifically slowed spin-up 2s (setting of ESC) and governor is in that time off and current is relatively flat, so it is clear that governor regulates relatively hard ... that current is measured inside ESC so it is not filtered by its capacitors, it is feeded from capacitors

but the point which is important on that picture is that the current is off in 4th second with only very small lagg after command to change rpm ... then motor changes rpm as free wheel (you can see the slope is the same like after complete off in 6.5s) and when it gets proper rpm, then the current is almost immediatelly again on and rpm curve is relatively flat again without overregulation and also without slow convergence ... it all means that it is not the ESC what makes regulation slow

unfotunatelly I do not have opposite example and also not reaction to load, but it is difficult to make so quick and repetitive load, this method is much easier :- ))

Offline Peter Germann

  • Trade Count: (0)
  • Captain
  • *****
  • Posts: 401
Re: Governor Bandwidth/response characteristics?
« Reply #23 on: October 15, 2012, 03:20:37 AM »
In the interest of Science, I suggest the application of a chainsaw.

Well, considering the fact that birdlife preservation groups have originally a) withdrawn objections to the building of the circle and b.) paid for what what you suggest to cut, makes us think twice...

rgds. Peter
Peter Germann

Online Howard Rush

  • 22 supporter
  • Trade Count: (0)
  • Admiral
  • *
  • Posts: 7818
Re: Governor Bandwidth/response characteristics?
« Reply #24 on: October 15, 2012, 11:15:32 AM »
I presume that model-bird accidents are handled discreetly.
The Jive Combat Team
Making combat and stunt great again

Online Brett Buck

  • Trade Count: (0)
  • Admiral
  • ******
  • Posts: 13792
Re: Governor Bandwidth/response characteristics?
« Reply #25 on: October 15, 2012, 01:10:58 PM »
Bob: the hunting with high gain is a well-known phenomenon in servo loops.  It is, in fact, one of the biggest challenges a control engineer faces: to get the disturbance rejection (speed changes in our case) low you need a high-bandwidth loop; to get high bandwidth you need high gain, and when you crank the gain up too much you get instability -- which exhibits itself as hunting.

   i.e. instability limited by nonlinearities that have the effect of altering the gain (up or down), i.e. a limit cycle.

    Brett

Online Brett Buck

  • Trade Count: (0)
  • Admiral
  • ******
  • Posts: 13792
Re: Governor Bandwidth/response characteristics?
« Reply #26 on: October 15, 2012, 01:22:03 PM »
this is not answer for Brett question, but it can show reality little bit

this graph is from ESC which we tried to use for quicker rpm changes for active regulators, it shows how quickly can ESC CHANGE rpm ... it is clear that the rpm change is short fraction of second ... and it HAS low pass filter, constant rpm regulation of governor is quicker than change of rpm

   In the case you show, its probably not the low-pass filter, but saturation of the control signal. But yes, the low-pass characteristic is also present and that's what I was asking about. Although I hadn't considered that the signal would saturate in flight - it might.

   Peter's data shows what appears to be the sort of thing I was looking for. Looks like the low-gain system has a characteristic response time of around a second, and the high gain of maybe 5x that. Treating it as if it were a second-order system it looks like the low-gain system would not do much in the middle of a corner, and the high-gain system would. Not enough samples for a complete system identification, but not far off what I had expected. I think from the responses I can guess the answer to the next question, but does anyone have an description of the control system or control algorithm?

 

   Brett

Online Brett Buck

  • Trade Count: (0)
  • Admiral
  • ******
  • Posts: 13792
Re: Governor Bandwidth/response characteristics?
« Reply #27 on: October 15, 2012, 01:31:08 PM »
Brett, attached is a number of logs recorded recently. Shown is RPM during five laps of level flight (left) and at time of the wingover manoeuvre (right) with Governor Gain set to min. respective max. values.
It's quite interesting to note that the only significant change resulting from switching from lowest to highest gain is the degree of RPM fluctuation in level flight (+/- 160 RPM vs  +/- 70 RPM) Recovery time seems, within the limits of the inaccurate measuring method used, to be almost the same.


   This is very good data that goes a long way to anwering the question.  I think it's pretty clear that the earlier perturbations were recovered far faster with the high-bandwidth system. When I get home I will mark it up with what I think I am seeing. The big humps in power are just it tracking the load, and that is controlled by the time it takes to do the maneuvers. One thing I don't see is large power spikes in the corners, which tells me it isn't responding fast enough to have any material effect on the airspeed. The sample rate appears to be fast enough to show it.

   Here's a second question - was there a qualitative difference in the feel of the two flights? Particularly in the square maneuvers?   Call me Quixotic but I am still trying to figure out if I can relate the things we feel at the handle to engineering parameters that can be measured.

    Brett

p.s. I am just guessing, of course, but I think we will find that it is filtering the RPM data and that that filtering doesn't change with the gain setting. Then the filtered RPM reading it then run through a control system with adjustable gain. That would have a time response of the two items in cascade, where ultimately the speed of the response is bounded by the response of the RPM filter, regardless of how high the gain gets turned up. That's probably why the RPM variation is only a factor of two different despite the fact that the bandwidth looks like a factor of 5 or more different.

Offline schuang

  • Trade Count: (0)
  • Ensign
  • **
  • Posts: 43
Re: Governor Bandwidth/response characteristics?
« Reply #28 on: October 15, 2012, 03:54:40 PM »
  This is very good data that goes a long way to anwering the question.  I think it's pretty clear that the earlier perturbations were recovered far faster with the high-bandwidth system. When I get home I will mark it up with what I think I am seeing. The big humps in power are just it tracking the load, and that is controlled by the time it takes to do the maneuvers. One thing I don't see is large power spikes in the corners, which tells me it isn't responding fast enough to have any material effect on the airspeed. The sample rate appears to be fast enough to show it.

   Here's a second question - was there a qualitative difference in the feel of the two flights? Particularly in the square maneuvers?   Call me Quixotic but I am still trying to figure out if I can relate the things we feel at the handle to engineering parameters that can be measured.

    Brett


p.s. I am just guessing, of course, but I think we will find that it is filtering the RPM data and that that filtering doesn't change with the gain setting. Then the filtered RPM reading it then run through a control system with adjustable gain. That would have a time response of the two items in cascade, where ultimately the speed of the response is bounded by the response of the RPM filter, regardless of how high the gain gets turned up. That's probably why the RPM variation is only a factor of two different despite the fact that the bandwidth looks like a factor of 5 or more different.


Brett,

The transient response of the governed motor control system from the command (timer) is not quite the same from the output (prop load change) - at least differs regarding to the slew rate.  The transient response of a second order system consists of the slewing + settling behavior of the system (overdamped, underdamped and critically damped that kind of stuff).  I suspect the slow setting from the plot Igor provided is due to the slew rate settling of the ESC.  Since the governor method for ESC was designed for the Heli guys so I think they are intentionally slow down the throttle response to prevent the tail kick during the initial head spin.

BTW, If you do a Google search on Matlab/Simulink with a motor control example -- you should be able to play with it. 

Regards,

Sean

Online Brett Buck

  • Trade Count: (0)
  • Admiral
  • ******
  • Posts: 13792
Re: Governor Bandwidth/response characteristics?
« Reply #29 on: October 15, 2012, 05:23:49 PM »

Brett,

The transient response of the governed motor control system from the command (timer) is not quite the same from the output (prop load change) - at least differs regarding to the slew rate.  The transient response of a second order system consists of the slewing + settling behavior of the system (overdamped, underdamped and critically damped that kind of stuff).  I suspect the slow setting from the plot Igor provided is due to the slew rate settling of the ESC.  Since the governor method for ESC was designed for the Heli guys so I think they are intentionally slow down the throttle response to prevent the tail kick during the initial head spin.

BTW, If you do a Google search on Matlab/Simulink with a motor control example -- you should be able to play with it. 

Regards,

Sean

   I understand the concept, having reviewed/approved at least a few motor controllers over the years. What we are missing is the actual control parameters/topology used by our controllers specifically.

   Brett

Offline Tim Wescott

  • 2016 supporter
  • Trade Count: (0)
  • Admiral
  • *
  • Posts: 12832
Re: Governor Bandwidth/response characteristics?
« Reply #30 on: October 15, 2012, 07:01:07 PM »
   I understand the concept, having reviewed/approved at least a few motor controllers over the years. What we are missing is the actual control parameters/topology used by our controllers specifically.

And I've designed and implemented quite a few (although none of the ones that Brett has reviewed).  I couldn't agree more.

Once you get past the basics of it being some sort of a speed-control loop, probably using a PI or PID controller, there's a whole lot of details that are left for us to guess about.
AMA 64232

The problem with electric is that once you get the smoke generator and sound system installed, the plane is too heavy.

Online Crist Rigotti

  • 23 supporter
  • Trade Count: (0)
  • Admiral
  • *
  • Posts: 3864
  • Electric - The future of Old Time Stunt
Re: Governor Bandwidth/response characteristics?
« Reply #31 on: October 15, 2012, 08:32:54 PM »
  <snip>
   Here's a second question - was there a qualitative difference in the feel of the two flights? Particularly in the square maneuvers?   Call me Quixotic but I am still trying to figure out if I can relate the things we feel at the handle to engineering parameters that can be measured.

    Brett

<snip>

Brett,
2 years ago when I was flying my Resolve EP at the NATS, I was taking to Bob Hunt and he mentioned turning up the gain on the governor.  I turned it up all the way and immediately felt more "connected" to the airplane.  Like going from a cable handle to a hardpoint handle.  Phoenix 45, Scorpion SII3020-780, FM-9 timer, 5S 2700mah 20C batteries.
Crist
AMA 482497
Waxahachie, TX
Electric - The Future of Old Time Stunt

Online Brett Buck

  • Trade Count: (0)
  • Admiral
  • ******
  • Posts: 13792
Re: Governor Bandwidth/response characteristics?
« Reply #32 on: October 15, 2012, 09:25:38 PM »
Brett,
2 years ago when I was flying my Resolve EP at the NATS, I was taking to Bob Hunt and he mentioned turning up the gain on the governor.  I turned it up all the way and immediately felt more "connected" to the airplane.  Like going from a cable handle to a hardpoint handle.  Phoenix 45, Scorpion SII3020-780, FM-9 timer, 5S 2700mah 20C batteries.

   Interesting, any other observations? I don't want to suggest things to you, but I am interested specifically in the feel/feedback/control pressure in the corners.

    Brett

Online Crist Rigotti

  • 23 supporter
  • Trade Count: (0)
  • Admiral
  • *
  • Posts: 3864
  • Electric - The future of Old Time Stunt
Re: Governor Bandwidth/response characteristics?
« Reply #33 on: October 15, 2012, 10:16:22 PM »
   Interesting, any other observations? I don't want to suggest things to you, but I am interested specifically in the feel/feedback/control pressure in the corners.

    Brett

The airplane felt that I had "tighter" control throughout the whole flight.  It was very noticable....it was an ah ha moment.
Crist
AMA 482497
Waxahachie, TX
Electric - The Future of Old Time Stunt

Offline Peter Germann

  • Trade Count: (0)
  • Captain
  • *****
  • Posts: 401
Re: Governor Bandwidth/response characteristics?
« Reply #34 on: October 16, 2012, 04:15:01 AM »
  was there a qualitative difference in the feel of the two flights? Particularly in the square maneuvers?  

Flight conditions (wind/turbulence) have been such that I have not noticed a great difference. However, having heard ever so often of the high-gain positive effects, my very subjective impression is that I felt tighter connection to the airplane. In corners, that is.

Peter

p.s. are we possibly on the way to make things a bit too complicated for the majority of e-stunt flyers? Should we perhaps, instead of "designing" rather sophisticated automatic power management systems, discuss the use of of-the-shelf manual throttles fpr c/l stunt?
Peter Germann

Offline Tim Wescott

  • 2016 supporter
  • Trade Count: (0)
  • Admiral
  • *
  • Posts: 12832
Re: Governor Bandwidth/response characteristics?
« Reply #35 on: October 16, 2012, 09:38:47 AM »
Igor Burger and Kieth Renicle are both frequent contributers to Stunthanger, and both design stunt timers.  Will Hubin (as far as I know) pays attention and gets on occasionally.  Others might want to design such for CLPA, and would find these discussions useful.

And I think the guys that don't need to know all the engineering details to get good flights are fully capable of not reading these threads when they pop up.  I know that I enjoy the chance to contribute, even if it may never spur Brett into making ESC/timer combinations specifically for CLPA.
AMA 64232

The problem with electric is that once you get the smoke generator and sound system installed, the plane is too heavy.

Online Brett Buck

  • Trade Count: (0)
  • Admiral
  • ******
  • Posts: 13792
Re: Governor Bandwidth/response characteristics?
« Reply #36 on: October 16, 2012, 11:04:14 AM »
Igor Burger and Kieth Renicle are both frequent contributers to Stunthanger, and both design stunt timers.  Will Hubin (as far as I know) pays attention and gets on occasionally.  Others might want to design such for CLPA, and would find these discussions useful.

And I think the guys that don't need to know all the engineering details to get good flights are fully capable of not reading these threads when they pop up.  I know that I enjoy the chance to contribute, even if it may never spur Brett into making ESC/timer combinations specifically for CLPA.

  I can almost guarantee that I am *not* going to do it!  If anything I might (and have been asked in the past) to advise but so far I don't think I know enough about how they work now - and more importantly, what we actually want it to be -  to be of any use.

 
    Brett

Online Brett Buck

  • Trade Count: (0)
  • Admiral
  • ******
  • Posts: 13792
Re: Governor Bandwidth/response characteristics?
« Reply #37 on: October 16, 2012, 11:05:21 AM »

p.s. are we possibly on the way to make things a bit too complicated for the majority of e-stunt flyers? Should we perhaps, instead of "designing" rather sophisticated automatic power management systems, discuss the use of of-the-shelf manual throttles fpr c/l stunt?

 I certainly can't see the harm in discussing the entire range of topics. I can't see why, with multiple national champions and a world champ participating we have to limit ourselves strictly to "how to"/"cookbook" beginner threads. There are already 1000 posts on "how to wire it up".

   BTW, I am looking for something specific that is potentially relevant to ALL stunt propulsion and can explain some fundamental observations we have had for decades.

    Brett
 

Offline Tim Wescott

  • 2016 supporter
  • Trade Count: (0)
  • Admiral
  • *
  • Posts: 12832
Re: Governor Bandwidth/response characteristics?
« Reply #38 on: October 16, 2012, 11:36:21 AM »
  I can almost guarantee that I am *not* going to do it!  If anything I might (and have been asked in the past) to advise but so far I don't think I know enough about how they work now - and more importantly, what we actually want it to be -  to be of any use.

Aww.

I'm actually working on this as a very back-burner project -- but its pretty much dependent on me having time and money both at the same time, which as a contract engineer in this economy is fairly rare.

I do have a 50A, 50V brushless motor controller development board that'll fit on the nose of a Banshee.  At 3 x 5-1/2 inches it's a bit bigger and heavier than a normal ESC, but it has the advantages that it is (a) surplus from a paying customer, and therefore essentially free for this purpose, and (b) laid out with lots of space around each component, which makes it easy to troubleshoot and modify.

So if I ever work myself out from underneath the pile of work I'm currently laboring under, I may actually get some code into the thing and start playing around.  Unfortunately, while I most certainly have all the motor-control smarts needed, at my current level of flying skill the best answer I can give to "how did it feel with tuning A vs. tuning B" is almost always going to be a wide-eyed "well, it didn't crash!".  My ambition is to get an ESC built that basically exposes every possible tuning parameter, and has an IMU input for folks that want to play with gyros and/or accelerometers, and if that's not enough, has code released under GPL and hardware designs released under Creative Commons or whatever works best for that side, so that anyone who wants to can replicate and/or modify at will.  That way, folks who fly well can try a bit of tuning and tweaking on their own, and let the rest of us know what works and what doesn't.
AMA 64232

The problem with electric is that once you get the smoke generator and sound system installed, the plane is too heavy.

Offline Keith Renecle

  • Trade Count: (0)
  • Captain
  • *****
  • Posts: 894
Re: Governor Bandwidth/response characteristics?
« Reply #39 on: October 17, 2012, 12:05:29 AM »
I think what Peter is proposing is something that we have discussed before on the F2B sub committee, and that is the use of remote throttle control in stunt. The AMA rules allowed this previously and Windy Urtnowski used it with some success. The F2B rules do not allow this. I do not personally think that this will be less complicated that what we have now with IC or electric power. Right now the off-the-shelf electric stuff is virtually plug n' play. Pro-stunt engines from Randy Smith and others were the same, BUT still take some intelligent (maybe not the right word) user to get them to work. We always hear...."just do what the top fliers do....build one of their designs.......use the engines they use" etc. Now, you can buy most of the really good designs in kit form, even some in ARF form, and just add one of their chosen power packages. This should mean that the top guys should have been knocked off their respective perches by now. True story?? No ways! The same top fliers are still there with a little shuffling of positions. It all comes down to who is wiggling the handle and it's a total package. Great airplane, power package, coach, correct practice methods, dedication etc. etc. etc. It's quite amazing how no matter what technology comes along, the same guys seem to be able to make the absolute best of it. They all have one common ingredient that cannot be explained rationally or in technical terms, and that is feeling and passion for the sport. They are all like fine musicians playing on the very best instruments. Anyone with enough money can buy a Steinway 9 foot grand piano, but only those with feeling and passion can get the best out of it. So it isn't just the technology, and to answer Peter's question about the throttle, sure we can easily do this.....all you have to do is to get enough votes from the international F2B folks to agree. I think that I will stick to my technical stuff rather than try that one. (we've tried that one before haven't we Peter??)

On Tim's CL esc thing, we have spoken about this in an e-mail a while back. I also have a good friend that is a design engineer with a specialty in motor controls. He has offered me a design, but costing it out has shown that it will be pricey and you need to offer at least a few different sizes. The small quantities will make production really expensive, so for now I will stick to my simple external governor thing and low cost Chinese esc's. I will however play with the idea just to get myself thrown into the deep-end on how these esc's work and for sure, it will improve my programming skills.

Keith R
Keith R

Offline Igor Burger

  • Trade Count: (0)
  • Admiral
  • ******
  • Posts: 2166
Re: Governor Bandwidth/response characteristics?
« Reply #40 on: October 17, 2012, 12:26:31 AM »
My ambition is to get an ESC built that basically exposes every possible tuning parameter, and has an IMU input for folks that want to play with gyros and/or accelerometers

:- ))) ... fortunately I am living on place where we have 2 producers of ESC very close so I am in contact with them ... and I know about such ESC:

http://mgm-compro.com/manuals/en-manual-x-series-071011-a1.pdf

if you go to page 10 you will see over 60 settings :- ))) ... but as I found lately, the less ESC does, the better ... because I found that if I can solve all tricks inside my active timer, it is much easier than to explain someone what I really need :- )))) ... so I still prefere SPin, where I set only gain of  governor and that is all :- )))

Offline Dean Pappas

  • Moderator
  • Trade Count: (0)
  • Admiral
  • *****
  • Posts: 1195
  • Welcome to the Stunt Hanger.
Re: Governor Bandwidth/response characteristics?
« Reply #41 on: October 18, 2012, 09:17:20 PM »
Hello Brett,
Hi Gang, sorry I have been away for a while due to work and it ain't over yet ...

The real short answer is to make the loop gain as high and the control bandwidth as wide as possible without causing hunt or oscillations. That is my opinion, and you should feel free to disagree after playing with your setup.
I will go as far as saying that it is okay, maybe even desirable, if the sudden step changes in load of square corners results in a moderate underdamped overshoot.
This is what I had helped put together for Bobby way back.
See the Norm Whittle Cookbook for very similar figures.

The Castle Phoenix, adjusted as originally set up several versions of software ago and never cleanly re-investigated(!) was set up for a 10% to 90% RPM rise/recovery time to a step input load change of a little over 250 milli-Seconds.
The step was accomplished by partially stalling and unstalling the propeller with a board placed closely in front of it.
Most ... even the outright majority ... are perfectly happy with control bandwidth half or less of what I have just described, and are happier to avoid the occasional hunt or RPM overshoot.

The gain varies by user, but the tightest setups I've seen vary +/- 50 RPM in the schedule, and many are happy with +/- 250.

I hope that answers part of your question. This sure has been a neat discussion, and fun reading.

Regards,
  Dean Pappas
                       P.S. I am still tickled over Igor's win!

Dean Pappas


Advertise Here
Tags:
 


Advertise Here