Jump to content
Sign in to follow this  
Alex Villa

Artificial Horizon OSD and more

Recommended Posts

Hello, Found this forum goggling about FPV, although I never wrote a post before I have been reading a lot.

My grammar sucks, I am sorry but english is not my native language.

The first reason I came here was searching info about autopilots, and as a start point for that I started a project to make an attitude indicator OSD. Since I have some experience with video signals and PICS ill try to fulfill some requirements that people make to existing OSD's.

For attitude sensor I choose an mma7260Q three axis accelerometer it cost only 6$ at mouser. Way cheaper and better than thermopiles.

I bought a cheap MPR in EBAY (Lyra X2400A) and it is very sensitive to signal lost. To make it work, OSD have to generate synch pulses in absence of video, instead of that I choose to generate synchs all the time and overlay the video part of the incoming signal on the generated synchs.

The hardware is composed of two boards, one board implements the video and on screen display (OSD) process based on a DSPIC30F3013. It generates a complete NTSC signal and overlays characters and graphics in the absence of video input. When video is present, it is used to synchronize with the complete set of synchs generated and video is displayed also, the way that it is implemented allow the restoration of distorted synchs aside from better tolerance to video link problems.

The Other board (Sensor Board) has the sensors (pressure, accelerometer and voltage).

The project is in progress, now i m trying to use the vertical blanking interval, or the first hidden lines in the video signal to send the data, since OSD will be added at ground, and dont want to use a separate modem.

Actual project status is in this sample video .

I will appreciate a lot, if people here give me some thoughs about this, because i am a complete newbie to rc models and my first RC fly is still for comming.

Share this post


Link to post
Share on other sites

Hmm, pretty nice :)

The only problem is that you can't get attitude information on an aircraft using accelerometers only, which is why the others use thermopiles in the abscence of a complex IMU.

The dynamics of flight are such that an airplane in a correct turn, even at 60° bank, will have its acceleration directed to the bottom of the plane. That means that even if you're banking that much, your AHI will show your plane being perfectly level. In practice you won't be able to make a perfect turn, so it will show you readings all over the place. An accelerometer alone will give you a turn coordinator (turn is good.. when the marker is flat). If your turn is badly coordinated, your current device could show your plane banking right when it's actually banking left...

You should go and have a flight in a full-scale airplane to feel that ;)

Share this post


Link to post
Share on other sites

Yep, you must have a reference to the horizon. The only things we can use for this are thermopiles or magnetometers. Gyro's at our needed size have lots of drift and need one of the other 2 metheods to keep them in check.

Thermopiles are by far the cheapest and easiest to use but magnetometers give better results.

Terry

Share this post


Link to post
Share on other sites
Hmm, pretty nice :)

The only problem is that you can't get attitude information on an aircraft using accelerometers only, which is why the others use thermopiles in the abscence of a complex IMU.

The dynamics of flight are such that an airplane in a correct turn, even at 60° bank, will have its acceleration directed to the bottom of the plane. That means that even if you're banking that much, your AHI will show your plane being perfectly level. In practice you won't be able to make a perfect turn, so it will show you readings all over the place. An accelerometer alone will give you a turn coordinator (turn is good.. when the marker is flat). If your turn is badly coordinated, your current device could show your plane banking right when it's actually banking left...

You should go and have a flight in a full-scale airplane to feel that ;)

Well, actually i found a way to go around that.

I konw that linear accelerations can introduce tilt errors, due to centripetal acceleration in addition to gravity, this can result in a false tilt correction when component is not tilted but, instead, linearly accelerated. That’s why software calculates the total acceleration, which should remain close to 1g during all tilt operations. If a tilt guard band were placed around the total acceleration, representing normal tilt operation, any values outside that guard band would indicate that heading errors are present and need to be discarded. Because sensor readings r taken very often (like 10 000 times / sec) and finite response filters are used the result is a very stable readings, Already tested it shaking the sensor board and looking the AHI.

Thanx for the reply

Share this post


Link to post
Share on other sites

...Thats the beauty of the three axis accelerometer. It can measure accelerations in all directions, in all the tilt situations the total acceleration (or euclidean distance) should remain close to 1g anything outside that plus or minus is equal to centripetal force or inertial force in any direction and thus may be discarded. Since those forces usually acts for a short time a good filter may get rid of that.

Share this post


Link to post
Share on other sites
in all the tilt situations the total acceleration (or euclidean distance) should remain close to 1g anything outside that plus or minus is equal to centripetal force or inertial force in any direction and thus may be discarded.

No. A correct turn will see a total acceleration proportional to the bank angle, and even worse, only on the Z axis.

The problem is that when flying, anytime you tilt the plane, if you want to keep your altitude you will have to pull the elevator, and thus apply a superimposed linear acceleration. Unless you specifically fly your plane in a certain way to allow the sensor to work you wont' be able to separate the 2 and tell what is what.

Let's consider you tilt the plane with the ailerons on the left, and that the plane is perfect, it will start diving as well, supposedly being "passive" and mostly affected by gravity. That would allow you to guesstimate a tilt angle. Then, as you don't want to fall and some time has already passed, you need to pull up quite hard to get back to your original altitude, which means a strong acceleration you could filter out. But then add some rudder in the process, and that will be ignored too while the plane will have changed attitude.

But then, even with that pretty uncomfortable approach the plane is never neutral and usually an action on one axis has some effect on the others as well.

Now with the "ideal" approach of a well-coordinated turn, you will linearly pull the elevator proportionally to your bank angle to maintain your altitude. In the process, the only acceleration felt is on the Z axis, and proportional to the bank angle. The X and Y would read a perfect zero during the entire process, from level flight before the turn to level flight after the turn. With any kind of filtering, that wouldn't allow you to even say whether the plane is turning left or right.

Now put both together - the extreme where the maneuver is purposely done in order to get something the sensor can use, and the other where the maneuver is perfectly executed and the sensor doesn't see a thing, there's the entire panel inbetween (99% of the time) where you will have a not-so-good execution of either of them. That means that depending on the way you enter a same angle turn, it would give you different angle readings each time.

We are pretty good at sensing accelerations, but our "gyros" are very bad. If you put someone in an airplane, have him close his eyes and fly a perfect barrel roll, the person will not notice anything. The accelerations he will feel during that time are identical to a level flight, even if the plane has done a complete roll. The problem is the same in your case.

Edited by Kilrah

Share this post


Link to post
Share on other sites

That sounds discouraging, though my aproach was right but certainly in that situations readings may be out of guard band to much time. I think its time to learn how to fly my plane and test the whole thing in real life. And maybe add a gyro. I saw a 2 axis gyro (IDG300) at sparkfun (55 to 75 $) :( way beyond the 6 $ for the accelerometer.

Share this post


Link to post
Share on other sites

your horizon video looks promissing great job,

will it also work on a plane, please make a test video for us ?

the centrifugal force when turning a plane might jam it ?

what about putting it in your car, and make a video clip I like to see that.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this  

×