A Pattern, as seen by a Gyro

[Tim Wescott]

I don't know if this is material for the engineering board, the advanced tech board, or of enough general interest that it belongs here.

After several months of work, I've got my combination data-logger/timer to the point where I could tape it onto a plane and record data while flying.  We managed three flights.

Flight #1 was on my #2 plane, which is suffering mysterious engine problems which the addition of much nitro could not cure.  It was really soggy above 40 degrees or so, and when it lost all line tension in the vertical 8 I bailed.

For flight #2 we switched planes to my #1 plane.  Aside from hitting my own wake in the square eights it was one of my best, ever.  I wish I could bottle it and bring it out at the Regionals.  Unfortunately, due to poor wire routing on my part, the power plug came undone as soon as the engine started.

Flight #3 went without a hitch technically, but the clouds rolled over the sun just as I took off, and the wind was going every which way.  This disconcerted me to the point where I did one maneuver twice (for extra credit, look at the "whole flight" plot and see which one it is -- it's clearly there).

The idea is that you can look at a flight afterward and see details that may escape you when you're flying.  I, for instance, just learned that I don't fly nearly as well as I thought I did.  I also learned that I really want to put this thing into the plane of a real expert and see how they fly (so Walker, Resinger, Cox, etc. -- if you see me at the Tune Up or the Regionals with a roll of electrical tape and a furtive air, hide your airplanes).

Here's the results from the one good flight. The files are named after the maneuvers shown.  And no, they're not in order.

[Tim Wescott]

If you're at a contest that I'm attending and you're concerned that I may, indeed, have slipped my flight recorder aboard your aircraft, you can use this photograph of today's setup in your search for subtle clues.

[Randy Cuberly]

Very interesting stuff Tim.
Now all you have to do is put it on Walker's airplane for comparison to yours!!!

It looks like it could stand a little more damping and still give a reasonable trace.

No?

Next is obviously a package that will fit inside the airplane...

Randy Cuberly

[Mark Knoepfle]

Randy - I was thinking about the same thing.

Tim - the spikes could be vibration induced so don't go blaming your flying just yet!

[Tim Wescott]

I think pretty much all that you see there is what the airframe is really doing and not engine vibration.

It'll fit inside a plane.  The biggest reason it's just strapped on there is because I just got the electronics working and haven't had time to cut into the plane to mount it.  The second biggest reason is because I wasn't 100% certain that the gyro chip would work given the amount of vibration generated by an engine.  Now that I know that the gyro is actually going to work with an LA 46 vibrating the whole airplane I'm going to mount the thing in the wing.

Most of the visual mass you see there is foam.  The board is not much bigger than an RC receiver, and the battery is, well, a battery.  I'm feeling hopeful that the thing can be hard-mounted to an airframe, now that I've seen the plots.

I'm going to put it in the wing.  On a full fuse plane with a muffler it would probably mount under the tank, and it'd go under the battery in a 'lectric (and you wouldn't need to have a separate battery).  I'm not sure where you could put it in a piped ship -- in a wing, perhaps.  You need access to the USB port so that you can talk to it with a computer, wherever it goes.

[Jason Greer]

Ah, cool stuff. Getting the first data sample is a good feeling. Brings back memories of staring at graphs for hours last year. I ran out of energy on my project and set it aside when I received Igors timer.

Is this purely gyroscope data or is it fused with accelerometer data? Curious to which IMU or gyro you are using?

[Tim Wescott]

It's purely gyro data.  There's a spot for an accelerometer on the board, but it's currently unpopulated.  I'm using the ST Microelectronics L3GD20.

[Douglas Ames]

Randy - I was thinking about the same thing.

Tim - the spikes could be vibration induced so don't go blaming your flying just yet!

The spikes on the "Roll" graph is what made me think the same. You would think that component would be fairly smooth in calm air.
Is there a way to reduce the sensitivity or isolate the gyro from vibration?

[Tim Wescott]

The air today was calm on average, but it had all sorts of small, violent spots of turbulence.  I was watching both planes jump, rattle & roll through all three flights.

[Mark Scarborough]

Tim, my new Impact is electric, and I am pretty sure there will be room inside the cowling for something like this ,, how much does it weigh?
I would let you strap it in, the space will probably allow for easy mounting,, of course,, I dont fly like Walker,,

[Randy Cuberly]

I suggested Walker as a Baseline of what to look for!   

Randy Cuberly

[Doug Moon]

Tim,

What is it about the graphs that tell you that you dont fly as good as you thought you did?

Thanks

Doug

[Robert-Jan]

Place the sensor on earplugs to damp the vibrations.
We also used them for the key cam on the quadcopter the difference was huge.

Greetings Robert-Jan

[Randy Ryan]

I'm assuming the traces are gyro displacements, certainly not actual airframe. Not completely sure what I see yet but it seems accelerometer data would be more helpful. Don't misunderstand, I really like what you're doing and maybe I'm missing something. It just seems that angular rates are really high in yaw and even roll at some points. Is it possible your scaling is off some?

[Tim Wescott]

What is it about the graphs that tell you that you don't fly as good as you thought you did?

There were three things that really stood out for me, one was airplane and the other two were me:

The first "me" thing can be seen in all the round maneuvers: a perfect round maneuver should have a rotation rate in pitch that's a perfect square pulse: it should go instantly from zero to some constant value, stay exactly at that constant value for the whole maneuver, and then drop instantly to zero.  Even if the plane is speeding up and slowing down (which mine does: you can tell by looking at the yaw axis rate) the pitch rate should undulate evenly.  Mine is all jitters and corrections.

The second "me" thing can be seen in the square maneuvers: a perfect square maneuver would rotate, then stop completely or rotate slowly*, then rotate fast, then slow, etc., until the maneuver is done.  Mine never reach a constant rotation speed during the corners, and -- more telling to me -- never go to a low constant (or zero) rate in the "straight" sections.  Indeed, after my first flight and seeing this, I noticed on subsequent flights that my "square" loops are more like "pincushion" loops, with tight corners that are held together by bowed-out sections.

The last thing I noticed, which I think is more a limitation of my (and I suspect, everyone's) plane, is that the rotation rate in the square maneuvers is way smaller than I expected.  I expected a rotation rate that would be up to five times faster than the rotation rate in the round maneuvers -- instead, my thumbnail estimate of the peak the rotation rate is less than twice the average rotation rate in a round loop.  I tried to correct for this in my second, unrecorded flight by really hitting the corners hard on the up-to-level transition, and all I could do was make the plane sort of flop over to level and lose all flying speed.

* When you're at the top of a square maneuver you're actually doing a loop around the center of the circle; this shows up not only in how much elevator you need to hold, but in the pitch axis of a gyro.  It's hard to see here because its muddied by my lack of precision, but if you look carefully at the entrance and exit to the outside squares, and the entrance to the clover, you can see it.

[Tim Wescott]

I'm assuming the traces are gyro displacements, certainly not actual airframe. Not completely sure what I see yet but it seems accelerometer data would be more helpful. Don't misunderstand, I really like what you're doing and maybe I'm missing something. It just seems that angular rates are really high in yaw and even roll at some points. Is it possible your scaling is off some?

There's an empty spot on the board (and in the software) for an accelerometer.  It just needs to be filled and the software written.

The plane does about 5-second laps, so in level flight you'd expect to see 72 degrees/second of rotation -- that's about what I see in the "whole flight" graph.

The short-duration wiggling in yaw and roll is, I think, really there.  I think you're seeing the aircraft roll and yaw in response to maneuvers, and then exhibit an underdamped recovery.  Keep in mind that its the rotation rate, so anything that happens quickly isn't going to have much time to build up into an angular change.

And my scaling is off -- it's about 10% low, due to me not reading the gyro data sheet correctly.  After doing some checks (like seeing that 3-1/2 inside loops from level left the plane flying upside down) I took another look and found my error.

[Brett Buck]

I'm assuming the traces are gyro displacements, certainly not actual airframe. Not completely sure what I see yet but it seems accelerometer data would be more helpful. Don't misunderstand, I really like what you're doing and maybe I'm missing something. It just seems that angular rates are really high in yaw and even roll at some points. Is it possible your scaling is off some?

      This seems about right to me, and I have looked at *a lot* of data over the years. They aren't the displacements, they are the rotational rate.

     One check on how accurate it is would be to take the data, integrate it into the attitude including the kinematics, and then see how fast it is diverging. I am sure Tim already thought of that.

   Even as it is, you could answer some important questions about trim and probably pick the line and aerodynamic natural frequencies right out of it.

    Brett

    

[Tim Wescott]

    One check on how accurate it is would be to take the data, integrate it into the attitude including the kinematics, and then see how fast it is diverging. I am sure Tim already thought of that.

I have, and that's how I discovered that I was reading the data sheet wrong on the scaling.

  Even as it is, you could answer some important questions about trim

I need to post some example graphs from flight #1, with the slow and much more poorly trimmed Banshee.  You can see severe hinging on the squares, as well as yaw that's much worse than the Twister/Sister Jenny.

and probably pick the line and aerodynamic natural frequencies right out of it.

Not to mention the fact that you can see that the natural frequencies drop when the aircraft is going slow.  Both planes are severely underdamped in both roll and yaw (I suspect they all are), and on both of them you can see the frequencies drop significantly overhead and to some extent right after square maneuvers.  This has to be a consequence of the outward acceleration dropping, and changing the period of the "pendulum" formed by the plane and lines.

[Mark Scarborough]

This is fascinating Tim
I have a question though,, this measures the force, or the actual track of the airplane?
It would seem to me that wind variations would impart force activity on the air-frame regardless of the actual track,,
I don't have the correct verbiage to express what I am envisioning I am afraid,,
so I guess the easiest way is to ask, is this recording the actual track of the airframe through space, or the loads and direction changes on it,, because a round loop is not symetrical with respect to the loads on the airframe,, nor would control inputs be symetrical,,, but the expectation is that the track would be symetrical ( of course not in OUR patterns Tim,, but in "their" pattern) lol

[Tim Wescott]

Mark:

Good question.

This is measuring the actual track of the aircraft.  Specifically, all it that it is measuring (well, all that it should be measuring) is the speed at which it is rotating, in all three axes (pitch, roll, and yaw).

So what you see is the result of control inputs, wind, the pilot yanking on the lines, bird strikes, and any other forces that affect the aircraft.  What you're not seeing, at least not directly, is line tension, airspeed, ground speed, control inputs, etc.

(Control inputs -- I should log control inputs!)

[Tim Wescott]

( of course not in OUR patterns Tim,, but in "their" pattern) lol

This is why Paul & co need to mount an armed guard around their airplanes at the next contest where I'm present!  I really want to get a baseline!

[Mark Scarborough]

Thanks Tim,,
again, fascinating stuff

as with all things like logging the electrical data from a flight,, its usefullness is yet to be fully realized,, but its interesting none the less,,

and my offer stands,, maybe we could both win,, you can get data, and I can get paul or someone to fly my plane,, hmmmm

[Igor Burger]

etc.

exactly :- )))))

[Randy Ryan]

     This seems about right to me, and I have looked at *a lot* of data over the years. They aren't the displacements, they are the rotational rate.

     One check on how accurate it is would be to take the data, integrate it into the attitude including the kinematics, and then see how fast it is diverging. I am sure Tim already thought of that.

   Even as it is, you could answer some important questions about trim and probably pick the line and aerodynamic natural frequencies right out of it.

    Brett

    

OK in the inside square loop graph I'm looking at the interval between seconds 213 and 214, one second elapsed. I'm seeing a total rate change of about 100degrees within that second, what am I missing? If that is the true rate of yaw, I would expect it to be radically visible, the mean rate being ~50 degrees. The first portion, 500ms, and a rate change of 80degrees/second. So what it the rate the rate of? Rate of gyro reaction?

[Tim Wescott]

It's going fast, but not for long.  So the result looks impressive when you're looking at rate, but it doesn't have time for the position to build up.

Here's a plot.  I took that gyro data and integrated it into angular positions.  The pitch is all over the map, so I left it off the plot.  The short-term wiggles in roll and yaw span maybe three degrees.

[Dave_Trible]

There were three things that really stood out for me, one was airplane and the other two were me:

The first "me" thing can be seen in all the round maneuvers: a perfect round maneuver should have a rotation rate in pitch that's a perfect square pulse: it should go instantly from zero to some constant value, stay exactly at that constant value for the whole maneuver, and then drop instantly to zero.  Even if the plane is speeding up and slowing down (which mine does: you can tell by looking at the yaw axis rate) the pitch rate should undulate evenly.  Mine is all jitters and corrections.

The second "me" thing can be seen in the square maneuvers: a perfect square maneuver would rotate, then stop completely or rotate slowly*, then rotate fast, then slow, etc., until the maneuver is done.  Mine never reach a constant rotation speed during the corners, and -- more telling to me -- never go to a low constant (or zero) rate in the "straight" sections.  Indeed, after my first flight and seeing this, I noticed on subsequent flights that my "square" loops are more like "pincushion" loops, with tight corners that are held together by bowed-out sections.

The last thing I noticed, which I think is more a limitation of my (and I suspect, everyone's) plane, is that the rotation rate in the square maneuvers is way smaller than I expected.  I expected a rotation rate that would be up to five times faster than the rotation rate in the round maneuvers -- instead, my thumbnail estimate of the peak the rotation rate is less than twice the average rotation rate in a round loop.  I tried to correct for this in my second, unrecorded flight by really hitting the corners hard on the up-to-level transition, and all I could do was make the plane sort of flop over to level and lose all flying speed.

* When you're at the top of a square maneuver you're actually doing a loop around the center of the circle; this shows up not only in how much elevator you need to hold, but in the pitch axis of a gyro.  It's hard to see here because its muddied by my lack of precision, but if you look carefully at the entrance and exit to the outside squares, and the entrance to the clover, you can see it.

On watching video of the last team trials in slow motion your description of square loops is accurate.  Most fly them as a continuous loop with different radii in spots,  hard to find any straight section, including yours truly.  Good thing our judges don't get instant replay....'banging' corners might be more about getting straight sections.

Dave

[Igor Burger]

Most fly them as a continuous loop with different radii in spots,  hard to find any straight section, including yours truly.

And pitching rate is still not enough to tell if the CG path was round or straight, model can do nice 90 degrees turn, but AoA (means wing to AIR not to ground or vertical) can (and will) still be at some angle .. it will need also acceleromer and calculate, but for proper calculation and integration we need also exact weight of the model ... and it is changing during flight because of used fuel ... so it is not so easy as it looks :- )))

[Tim Wescott]

And pitching rate is still not enough to tell if the CG path was round or straight, model can do nice 90 degrees turn, but AoA (means wing to AIR not to ground or vertical) can (and will) still be at some angle ..

Hmm.  I hadn't thought that far, but yes, just knowing pitch doesn't necessarily inform us of path.

it will need also acceleromer and calculate, but for proper calculation and integration we need also exact weight of the model ... and it is changing during flight because of used fuel ... so it is not so easy as it looks :- )))

I disagree about needing to know the weight of the model.  Just having the acceleration and angular rates, along with starting position, orientation, and velocity, would be enough.  I don't think that knowing the model's mass would give you useful information.

[Igor Burger]

Theoreticaly should be enough, but I remember (well it is 6 years back when I did this) I found somewhere some exception ... I do not remember exactly ... I will think about it.

[Randy Ryan]

It's going fast, but not for long.  So the result looks impressive when you're looking at rate, but it doesn't have time for the position to build up.

Here's a plot.  I took that gyro data and integrated it into angular positions.  The pitch is all over the map, so I left it off the plot.  The short-term wiggles in roll and yaw span maybe three degrees.

Ahhh that's what I would have expected to see, thanks.

So the angular rates in the graphs are gyro precessions?

[Brett Buck]

Ahhh that's what I would have expected to see, thanks.

So the angular rates in the graphs are gyro precessions?

   Not, almost certainly not. There are plenty of things forcing it in roll and yaw that are bigger than precession - like line whip, aerodynamics, lateral imbalance, etc.

    And its not a precession of the gyro, either, since - it's not really a gyro (no spinning wheel). Tim can tell us but its likely a MEMS "tuning fork" "gyro".

     Brett

[Tim Wescott]

   Not, almost certainly not. There are plenty of things forcing it in roll and yaw that are bigger than precession - like line whip, aerodynamics, lateral imbalance, etc.

Before we all get too confused, I think there's a terminology collision here.  Brett's thinking forced precession, as from the propeller.  I'm pretty sure that you're thinking of the rotation rate of the gyro -- which is what is being measured.

    And its not a precession of the gyro, either, since - it's not really a gyro (no spinning wheel). Tim can tell us but its likely a MEMS "tuning fork" "gyro".

Actually I think that these days they use oscillating masses that don't resemble tuning forks at all.  It is a MEMS gyro, however, and almost certainly has several things inside of it that flap back and forth in a way that allows rotational rates to be measured.

In all my years working around gyros, this is the first time I've heard (well, seen) someone say "it's not really a gyro because it has no spinning wheel".  When the fiber-optic gyro guys were trying to sell us fiber-optic gyros they didn't use quote marks.  And when the quartz MEMS gyro guys were trying to sell us quartz MEMS gyros, they didn't use quote marks.  And when the spinning-wheel gyro guys were successfully keeping our business, even they didn't use quote marks around "gyro", and they would have been motivated to do so (they did use quote marks around "good" when they talked about their competitors, though!).

[Brett Buck]

Actually I think that these days they use oscillating masses that don't resemble tuning forks at all.  It is a MEMS gyro, however, and almost certainly has several things inside of it that flap back and forth in a way that allows rotational rates to be measured.

In all my years working around gyros, this is the first time I've heard (well, seen) someone say "it's not really a gyro because it has no spinning wheel". 

     Howard is probably finding this entertaining... If anyone ever says I have no self-control, I will direct them to this thread.

When the fiber-optic gyro guys were trying to sell us fiber-optic gyros they didn't use quote marks.  And when the quartz MEMS gyro guys were trying to sell us quartz MEMS gyros, they didn't use quote marks.  And when the spinning-wheel gyro guys were successfully keeping our business, even they didn't use quote marks around "gyro", and they would have been motivated to do so (they did use quote marks around "good" when they talked about their competitors, though!).

   I mention it because gyroscopic precession is not involved in the sensing of your rates, just like no gyroscopic precession is involved in fiber-optic "gyros", or ring-laser "gyros" or hemispherical resonator "gyros".

     I haven't bothered to look the particular units you describe, because I don't care very much, but it's very likely two blade-like piezo-electric crystals mounted perpendicular to each other, one driven into resonance - you know, like a tuning fork - about one axis, and the oscillations from the driven blade coupled by rate into the undriven blade, which is also driven into resonance - like a tuning fork - and sensed by the piezo-electric effect voltage thus developed. The magnitude of the coupled oscillation is proportional to the rate. Most of this type to not operate in  rebalance mode, so not inertial-quality and not an integrating unit (which is why it doesn't look too good when you integrate pitch), but good enough to get a feel for the motion and potentially stabilize motion over short periods.

    In the vernacular with which you are familiar, of course you are right, all rate sensors tend to be called "gyros" even though they aren't.
     
      Brett