SMALL - espritmodel.com SMALL - Telemetry SMALL - Radio
Reply
Thread Tools
Old Dec 25, 2013, 12:14 AM
Registered User
HELModels's Avatar
Joined Aug 2003
2,631 Posts
Quote:
Originally Posted by aeronaut999 View Post
I'm confused as to how an "offset" would relate to restarting in the air, in an unknown aircraft attitude, and possibly in a turn, where the G-loading vector (sensed by the accelerometer) is not aligned with the actual gravity vector. Sorry that I couldn't put the issues more concisely in my ridiculously long post 2 posts above but unless I am missing something, there are some issues here.

Steve
The offsets are unique to each sensor and when you figure them out through calibration, they can be ready any time it restarts even in some unusual attitude. They can be added to a program. I don't guarantee much, but it's been done before to guard against restarting in air and having an IMU think a 45 degree bank is level.
HELModels is online now Find More Posts by HELModels
Reply With Quote
Sign up now
to remove ads between posts
Old Dec 26, 2013, 08:45 AM
Registered User
The Willamette Valley, Oregon
Joined Dec 2008
1,141 Posts
Quote:
Originally Posted by HELModels View Post
The offsets are unique to each sensor and when you figure them out through calibration, they can be ready any time it restarts even in some unusual attitude. They can be added to a program. I don't guarantee much, but it's been done before to guard against restarting in air and having an IMU think a 45 degree bank is level.
I may be under-appreciating what can be done by sensing rotation rates as well as linear vectors. I was trying to explain a few posts back why just sensing the apparent G-loading or acceleration vector (including gravity), and the direction of the magnetic field vector, isn't going to allow you to know which way is truly "up", nor will it let you compute the heading (unless you are on the equator where the magnetic field has no dip.) Even if your accelerometers are perfect, with no drift. Speaking specifically of the case where we've started up the instrument with the aircraft in an unknown attitude and an unknown (and not necessarily linear) trajectory. But if we look at the direction and rate of rotation of the magnetic field vector (just as a pilot does when observing the Bohli compass), and we sense the direction and rate of the aircraft's rotation around all three axes, maybe there's enough information there to allow us to quickly decide the true direction of "up". (I don't feel like I really have a complete handle on this.) If so, once we know which direction is "up", it's easy to compute heading from the magnetic field vector.

Steve
aeronaut999 is online now Find More Posts by aeronaut999
Last edited by aeronaut999; Dec 26, 2013 at 06:56 PM.
Reply With Quote
Old Dec 26, 2013, 06:50 PM
Registered User
The Willamette Valley, Oregon
Joined Dec 2008
1,141 Posts
ixGyro Glass Cockpit-- smart-phone app for attitude/ heading display

This smart-phone app for artificial horizon and heading information sounds pretty well thought-out and promising:

ixGyro Glass Cockpit

http://www.ixellence.com/ixgyro/help/index.html
http://www.ixellence.com/index.php?o...id=274&lang=en

According to this link it doesn't sound like it's a good candidate for a quick start-up in the midst of a spiral dive, but maybe we can live with that:

http://www.ixellence.com/ixgyro/help/calibration.html

"To calibrate the neutral position of pitch, roll and the slip indicator simply long press the touchscreen. This will set the pitch, roll and slip angles to zero. Alternatively, you can also calibrate ixGyro by selecting the corresponding option in the menu.

You can recalibrate ixGyro any time and as frequently as you want. Make sure that the device is in rest or at least in a non-accelerated state when you calibrate it.

Important note: Besides the manual calibration of the neutral position ixGyro is continously performing a kind of autocalibration to determine the optimum parameters for drift compensation. Immediately after starting ixGyro these parameters can be inaccurate, leading to a temporary offset of the attitude indicator (pitch/roll). This is normal and should be compensated for automatically after a few minutes."

It's intended for phones with rotation rate sensors as well as accelerometers, but they offer a plug-in 3-d rotation rate sensor for phones lacking these:

http://www.ixellence.com/index.php?o...id=274&lang=en

They are charging 100$ for the pro version without the demo messages, but one could try the demo version first.

One concern-- they are using GPS data to correct the attitude indicator. I suspect that if the aircraft were enveloped in a strong updraft or downdraft, so that the velocity vector were not horizontal, this would cause the attitude indicator to drift to indicate a nose-up or nose-down pitch attitude when the aircraft was actually not in a level attitude. I would much prefer not to have this issue-- but again, maybe we can live with it... I'm interested to give it a try...

But I still think we could and should do better. Would prefer to use high-quality sensors that need minimal correction for drift. Would prefer not to use the GPS-derived velocity vector for correction of attitude indicator, especially in the pitch axis, due to issues around updrafts/ downdrafts, changes in power setting, and the fact that even in a glider in still air there is not a simple linear relationship between pitch attitude and glide angle. If we make optimum use of a magnetometer I think we could dispense with the need to show GPS data at all-- that might be good. Or maybe it is better to make some use of GPS data to correct the attitude indicator in the roll axis-- so we aren't so bothered by magnetic fields in the aircraft?

Hmmm..

Steve
aeronaut999 is online now Find More Posts by aeronaut999
Last edited by aeronaut999; Dec 26, 2013 at 07:11 PM.
Reply With Quote
Old Dec 27, 2013, 03:35 AM
Registered User
HELModels's Avatar
Joined Aug 2003
2,631 Posts
That app sounds decent.

Here's a project that uses a ground station to display IMU and external compass. It's crude, but over on DIYdrones the designer says he plans to improve it.

http://www.youtube.com/watch?feature...&v=1JbAVCclQcY

Here's what I'm working on now. I got that example running on an android emulator. My own version won't run because it uses serial input and that needs a library or something. I'll keep working on it.
HELModels is online now Find More Posts by HELModels
Last edited by HELModels; Dec 27, 2013 at 11:45 PM.
Reply With Quote
Old Dec 29, 2013, 11:24 PM
Registered User
HELModels's Avatar
Joined Aug 2003
2,631 Posts
I found this android based ground control station that uses a wireless connection to an Ardupilot. The source is available for free and seems like a good candidate for Android.

http://www.diydrones.com/profiles/bl...age=9#comments
HELModels is online now Find More Posts by HELModels
Reply With Quote
Old Jan 06, 2014, 04:53 PM
Registered User
The Willamette Valley, Oregon
Joined Dec 2008
1,141 Posts
Bohli compass notes

Thinking about how a stand-alone artificial horizon and heading indicator might be able to be quickly fired up with no wings-level calibration phase, even if the initial bank angle and pitch attitude and degree of curvature to the flight path are unknown--

Specifically regarding the earth's magnetic field vector, my thoughts go to some sort of electronic analog of the Bohli compass, which has a lightweight, physical needle in full 3-dimensional alignment with the earth's magnetic field.

Check out this manual for the Bohli compass-- http://hkavionics.com/Bohli_man/ba_kompi_e4.pdf

If the local magnetic dip angle is known, the bank angle can be easily determined when the aircraft is in a level pitch attitude and flying on a magnetic heading of 90 or 270, while the pitch attitude can be easily determined when the aircraft is in a level bank attitude and flying on a magnetic heading of 180 or 360. But it gets better than that-- Fig. 7 in the manual shows that if the needle is observed through a complete circle, and the pitch attitude and bank angle are roughly constant through the circle, then both the bank angle and pitch attitude can be determined. It seems that an electronic computation that was analogous to observing the needle of the Bohli compass through a whole revolution, would be very helpful to being able to quickly fire up a stand-alone artificial horizon and heading indicator without an initial calibration phase, even if the aircraft's pitch attitude, bank angle, and flight path are initially unknown.

It might be helpful for fast initialization if some estimate of the downward magnetic dip angle for the present location were pre-set into the device before flight.

Likewise, if GPS data is also utilized, it might be helpful for fast estimate of bank angle (via turn radius) if some estimate of the aircraft's airspeed were pre-set into the device before flight-- especially if the device is to work in everything from say, hang gliders to fast GA aircraft!

A careful comparison of Figs 2 and 7 in the Bohli compass manual should make it clear that if the pilot just glances at the needle display without knowing anything about his pitch attitude, bank angle, or the local magnetic dip angle, he can't know his precise heading. In circling flight, shifting the circle described by the compass "pipper" to the left or to the right (by flying at a different bank angle than the tilt knob has been set for) or shifting the circle described by the compass pipper forward or aft (by flying in a different pitch attitude than the one that the compass mount has been designed to accomodate) will change the position of the pipper on the heading reference grid, at any moment in time. Similarly, we can't simply take a quick glance at a 3-d magnetic compass to get an accurate heading reference. Rather, we have to do some computations involving the pitch and bank angle. But the Bohli compass analogy suggests that those things can be computed or estimated by looking at the way the earth's 3-d magnetic field vector is rotating over time, with respect to the sensors.

Just some random thoughts...

Steve
aeronaut999 is online now Find More Posts by aeronaut999
Reply With Quote
Old Jan 17, 2014, 02:53 PM
Registered User
The Willamette Valley, Oregon
Joined Dec 2008
1,141 Posts
expensive accessory

Oh dear-- I didn't realize the plug-in device required to give lower-end smart phones the needed gyro sensors is a) provided by a different company that has nothing to do with the ixGyro app, and b) costs 800$! If your phone is on the list of phones that has all the needed sensors built in, you don't need this accessory, but...

http://www.aviation.levil.com/AHRS_mini.htm

The link mentions zero drift rate in the sensors. The whole point of the ixGyro app is to provide a workable app even for phones without such super-duper sensors, by means of an optimized calibration system to overcome sensor drift. But, then they suggest going and buying this expensive thing, if your phone has only accelerometers and not rotation sensors. Much cheaper to buy a different phone with rotation sensors (see the list of tested phones on the ixGyro link).

If you did own the expensive plug-in AHRS accessory, then perhaps the ixGyro app wouldn't be the best way to make use of it anyway...


Quote:
Originally Posted by aeronaut999 View Post
This smart-phone app for artificial horizon and heading information sounds pretty well thought-out and promising:

ixGyro Glass Cockpit

http://www.ixellence.com/ixgyro/help/index.html
http://www.ixellence.com/index.php?o...id=274&lang=en

According to this link it doesn't sound like it's a good candidate for a quick start-up in the midst of a spiral dive, but maybe we can live with that:

http://www.ixellence.com/ixgyro/help/calibration.html

"To calibrate the neutral position of pitch, roll and the slip indicator simply long press the touchscreen. This will set the pitch, roll and slip angles to zero. Alternatively, you can also calibrate ixGyro by selecting the corresponding option in the menu.

You can recalibrate ixGyro any time and as frequently as you want. Make sure that the device is in rest or at least in a non-accelerated state when you calibrate it.

Important note: Besides the manual calibration of the neutral position ixGyro is continously performing a kind of autocalibration to determine the optimum parameters for drift compensation. Immediately after starting ixGyro these parameters can be inaccurate, leading to a temporary offset of the attitude indicator (pitch/roll). This is normal and should be compensated for automatically after a few minutes."

It's intended for phones with rotation rate sensors as well as accelerometers, but they offer a plug-in 3-d rotation rate sensor for phones lacking these:

http://www.ixellence.com/index.php?o...id=274&lang=en

They are charging 100$ for the pro version without the demo messages, but one could try the demo version first.

One concern-- they are using GPS data to correct the attitude indicator. I suspect that if the aircraft were enveloped in a strong updraft or downdraft, so that the velocity vector were not horizontal, this would cause the attitude indicator to drift to indicate a nose-up or nose-down pitch attitude when the aircraft was actually not in a level attitude. I would much prefer not to have this issue-- but again, maybe we can live with it... I'm interested to give it a try...

But I still think we could and should do better. Would prefer to use high-quality sensors that need minimal correction for drift. Would prefer not to use the GPS-derived velocity vector for correction of attitude indicator, especially in the pitch axis, due to issues around updrafts/ downdrafts, changes in power setting, and the fact that even in a glider in still air there is not a simple linear relationship between pitch attitude and glide angle. If we make optimum use of a magnetometer I think we could dispense with the need to show GPS data at all-- that might be good. Or maybe it is better to make some use of GPS data to correct the attitude indicator in the roll axis-- so we aren't so bothered by magnetic fields in the aircraft?

Hmmm..

Steve
aeronaut999 is online now Find More Posts by aeronaut999
Last edited by aeronaut999; Jan 17, 2014 at 03:07 PM.
Reply With Quote
Old Jan 18, 2014, 04:30 AM
Registered User
Joined Apr 2013
70 Posts
I've recently written another blog post, building on the AHRS maths blog post that I previously posted in this thread. It's about correcting accelerometer readings for acceleration based on GPS data. I've come up with the algorithm myself and it should not be affected by updrafts or whatever. The reason I'm posting it in this thread is mostly because I think it might be helpful to the cause, but also because the clever minds here might be able to pick holes in my thinkings.

http://www.camelsoftware.com/firetai...rifugal-force/
SuperCamel is offline Find More Posts by SuperCamel
Reply With Quote
Old Jan 21, 2014, 01:00 PM
Registered User
The Willamette Valley, Oregon
Joined Dec 2008
1,141 Posts
centrifugal accelerations

A short version of a reply:

Re centrifugal accelerations-- it seems to me that these should not be a huge problem, because if they are applied long enough to drive the flight path in a complete circle, the error will have been applied equally in every direction and the cumulative error will be zero. Some error will be detectable after 180 degrees of turn.

That doesn't mean you might not get BETTER data if you DO correct for centrifugal accelerations based on GPS data. But the fact that this is NOT done in a "conventional" mechanical attitude indicator seems significant-- you can circle round and round indefinitely without accumulating a net error, because after every 360 degrees of turn, the accumulated net error is back to zero. In contrast, a prolonged acceleration in a straight line does cause a significant tilt of the gyro, in the pitch axis, for the duration of the acceleration.

Stay tuned for a bit more...

Steve

Quote:
Originally Posted by SuperCamel View Post
I've recently written another blog post, building on the AHRS maths blog post that I previously posted in this thread. It's about correcting accelerometer readings for acceleration based on GPS data. I've come up with the algorithm myself and it should not be affected by updrafts or whatever. The reason I'm posting it in this thread is mostly because I think it might be helpful to the cause, but also because the clever minds here might be able to pick holes in my thinkings.

http://www.camelsoftware.com/firetai...rifugal-force/
aeronaut999 is online now Find More Posts by aeronaut999
Reply With Quote
Old Jan 21, 2014, 01:01 PM
Registered User
The Willamette Valley, Oregon
Joined Dec 2008
1,141 Posts
centrifugal accelerations

Here are some thoughts on AHRS systems, correcting for accelerations in various directions, etc. This is a (much) longer version of my post immediately above...

Thanks for posting re correcting for centrifugal force..

I would like to better understand how the Levil AHRS and other AHRS systems work.

Obviously if there were zero drift of any kind (sensor drift, drift from numerical processing), there would be no need for "correcting" the sensors and no need to sense the direction of gravity. You could just start it up with the aircraft in a level attitude and you would be set for the duration of the flight. (Assuming that you don't cover a great deal of distance over the surface of earth, or fly so long that the earth's rotation is a significant factor.)

A conventional (mechanical) attitude indicator is subject to precession, because the gimbal bearings are not perfectly frictionless and thus any pitching or rolling of the aircraft imposes a small force on the gyro.

How does the device correct for this? By constantly sensing the apparent direction of gravity and applying a small corrective force to the gyro. This is done through the use of four "pendulous vanes" that partly cover the air exit ports.

Whenever the vanes are operating, they are slowly tilting the rotor into alignment with the apparent direction of gravity. See for example 1:40 to 3:05 here:
The Attitude Indicator (3 min 44 sec)
.

The video gives the incorrect impression that the precession forces on the gyro are strong, but are quickly corrected by the pendulous vanes, which exert a strong corrective force on the gyro to keep it aligned with the apparent direction of gravity. Obviously, a more accurate description of the situation would be that the since the gimbal bearings have only a little friction, the torques transmitted through the bearings to the gyro are weak, and so is the resulting precession effect. The corrective action of the pendulous vanes is also weak-- weak enough not to cause huge errors during normal turning flight, when the apparent direction of gravity is NOT aligned with the true direction of gravity, but the amount of time spent flying in any one particular direction is small.

I believe I've read that the operation of these vanes is disabled during high-G turns and other high-G maneuvers, but I'm not sure if that's really true- I'm not finding a reference in support of that now.

Whenever the vanes are operating, they are slowly tilting the rotor into alignment with the apparent direction of gravity. This creates a small error in turning flight, when the apparent direction of gravity is not the same as the actual direction of gravity. The resulting cumulative error in the gyro's orientation is greatest after the aircraft has turned through a heading change of 180 degrees. After a full 360 degrees of turn, the erroneous tilting effect from the vanes has operated equally in every direction and the net error is back to zero.

For example, see the "Errors" section of this webpage: http://www.pilotfriend.com/training/...g/attitude.htm

You can see how this error would tend not to be a problem at very high turn rates, because there isn't much time for an error to accumulate.

With an electronic attitude indicator, bearing friction is no longer an issue, but sensor drift and numerical processing cumulative errors are issues. Again we need some kind of way to correct the gyro, for example by sensing the apparent direction of gravity. (In theory, we could also use the earth's magnetic field to correct our gyro, after the manner of the Bohli compass-- see my post #36-- without paying any attention to the apparent direction of gravity at all.)

If we are sensing the apparent direction of gravity, we will be affected by acceleration errors. Interestingly, some versions of the Levil AHRS G mini connect to the aircraft pitot-static system and try to compensate for errors relating to accelerations along the flight path (increases or decreases in forward speed):

"During flight, the instrument utilizes the indicated airspeed from the pitot-static system for roll and pitch calculations. However, SS and SW models that are not connected to the pitot-static system will not have access to speed and might display a small error during take-off, often seen as a pitch-up error on Jets and fast airplanes. " ( http://www.aviation.levil.com/How_it_works.pdf )

I don't understand the reference to "roll" calculations-- it seems to me that this would be a pitch issue only-- but perhaps not, if the aircraft is banked while accelerating?

The Levil AHRS G mini apparently does not use GPS data at all for correcting the AHRS system.

The IxGyro (cell phone app) apparently does similar corrections for accelerations along the flight path, but based solely on the GPS track data. The corrections are supposed to be good enough to work even with the lower-quality rotation rate sensors found in smart phones.

"ixGyro is the first true-attitude indicating glass cockpit app for Android smartphones. The reliable and robust artificial horizon is created by processing the current data of the smartphone sensors (accelerometer sensor, GPS signal and the gyroscope). Even trajectorial accelerations do not influence the true-attitude indicator. In addition to the existing aircraft instruments ixGyro is a useful tool for pilots of small aircrafts." ( http://www.ixellence.com/index.php?o...id=274&lang=en )

And:

"Although the attitude can be determined also with the acceleration sensor if the device is in rest or in linear motion, the accelerometer becomes unusable as soon as the device is subject to trajectorial accelerations because of their interference with gravity. Thus a sensor like the gyroscope is needed that is able to measure rotation independent of acceleration. However, the problem with a gyroscope is that it cannot measure the attitude directly but only the rotation speed. Pitch and roll have to be calculated from that by integration which leads to numerical drift. Moreover, there is a physical drift due to the large bias of the low quality gyroscopes in today's smartphones. We have developed a new method to compensate for these drift effects, so our attitude indication is very stable and reliable if a gyroscope is available besides GPS and acceleration sensors." ( http://www.ixellence.com/index.php?o...id=274&lang=en )

Now to steer the conversation back around to the issue of centrifugal accelerations-- it seems to me that these should not be a huge problem, because if they are applied long enough to drive the flight path in a complete circle, the error will have been applied equally in every direction and the cumulative error will be zero. Some error will be detectable after 180 degrees of turn.

That doesn't mean you might not get BETTER data if you DO correct for centrifugal accelerations based on GPS data. But the fact that this is NOT done in a mechanical attitude indicator seems significant-- you can circle round and round indefinitely without accumulating a net error, because after every 360 degrees of turn, the accumulated net error is back to zero. In contrast, a prolonged acceleration in a straight line does cause a significant tilt of the gyro, in the pitch axis, for the duration of the acceleration. (Or more precisely, the error continually increases for the duration of the acceleration, until at some point the gyro has come into full allignment with the "apparent" direction of gravity in the accelerating reference frame?)

One more note-- it's interesting to note that from the viewpoint of the earth's reference frame, there's no difference in the gyro errors that tend to accumulate in a skidding turn versus a coordinated turn. In either case, the turn rate will be determined by the centripetal force component. The gyro doesn't care that the net aerodynamic force is "square" to the wingspan in the case of a coordinated turn, and tilted toward the inside wingtip in the case of a skidding turn. The direction of the net aerodynamic force vector, relative to the earth's surface, is the same in either case. In either case, any correction based on the apparent force of gravity will tend to slowly tilt the gyro until--given enough time-- it is fully aligned with the net aerodynamic force vector. For the same turn rate, the gyro's orientation will be the same relative to the earth, whether the turn is skidding or coordinated. But the gyro's orientation relative to the aircraft will be different in each case-- it will tend to indicate that the aircraft is wings-level in the case of the coordinated turn, and will tend to indicate that the aircraft is banked toward the outside of the turn in the case of the skidding turn. These are the theoretical errors that would accumulate given enough time, if the turn rate were very low (i.e. the aircraft were flying very fast.) But again, in practice, we see these errors only manifested to a small degree, peaking when the aircraft has turned through 180 degrees, and disappearing again when the aircraft has turned through a full 360 degrees.

Disclaimer-- I'm just doing my best to figure these things out as I go. For example, I haven't actually tested the effect of a series of prolonged 360-degree wings-level skidding turns, on either a mechanical or electronic attitude indicator!

Steve
aeronaut999 is online now Find More Posts by aeronaut999
Last edited by aeronaut999; Jan 21, 2014 at 02:18 PM.
Reply With Quote
Old Jan 21, 2014, 02:44 PM
Registered User
The Willamette Valley, Oregon
Joined Dec 2008
1,141 Posts
You may find this to be worthwhile reading:

http://myahrs.wordpress.com/

Especially this section: "Turning Error Correction":

http://myahrs.wordpress.com/2012/04/...or-correction/

Steve

Quote:
Originally Posted by SuperCamel View Post
I've recently written another blog post, building on the AHRS maths blog post that I previously posted in this thread. It's about correcting accelerometer readings for acceleration based on GPS data. I've come up with the algorithm myself and it should not be affected by updrafts or whatever. The reason I'm posting it in this thread is mostly because I think it might be helpful to the cause, but also because the clever minds here might be able to pick holes in my thinkings.

http://www.camelsoftware.com/firetai...rifugal-force/
aeronaut999 is online now Find More Posts by aeronaut999
Reply With Quote
Reply


Thread Tools

Similar Threads
Category Thread Thread Starter Forum Replies Last Post
Data How does artificial horizon display work? davensocal Eagle Tree Systems 32 Jun 07, 2012 03:51 AM
Discussion Guys lets do it! Lets make an official rc television show!!!!!!!!!!!!! lightspeeddud Fuel Plane Talk 0 Jan 28, 2009 02:07 PM