MoTheG Show full post »
MoTheG
sigma_v.png 
As I mentioned above the absolute error \ variance depends largely on the turn rate. The up-side of that is, that you can expect good results for high turn rates, that is minimum time (as opposed to minimum radius) turns.
The error for the centripetal (centrifugal) acceleration is simply the error of the attitude + the error of the acceleration sensor.
If it can not fly with NiMH it is not a plane.
Quote 0 0
MoTheG
Quote:
Because the speed is the quotient\ratio between two measurements we are independent of their absolute magnitude.

That did not come out right. What I meant is:
Because the speed is the quotient\ratio between two measurements the relative error of the speed is not a function of the speeds absolute magnitude. Therefore the speed can have little error even if it is low if the turn is tight enough.
If it can not fly with NiMH it is not a plane.
Quote 0 0
quorneng
You can measure straight line speed pretty accurately with two people and two stop watches.
Quote 0 0
MoTheG
For the horizon sensor I have an idea that I do not know if it is practical \ workable:
Instead of going by the relative brightness of detectors covering continues sectors one might use an encoding scheme.
Using multiple detectors for the same sector one could place lenses in front of them in such a way that the sector is subdivided with each lens looking at multiple subsectors at the same time, but each detector looking at a different set of subsectors.
Assuming we live on a barren flat glowing surface with a dark sky, one could use five detectors to place the horizon into one of 32 subsectors, the one that is changing the most and neither among the uniformly bright ones nor the dark ones.
The encoding scheme could be overlapping single continues sectors per detector, non overlapping non continues multiple sectors per detector, gray code.
Nowadays it is easier to use redundancy and processing than expensive detectors and calibration. Even with the detector on the other side of the aircraft blinded by the (setting) sun, one of the described devices should give a reading.
If it can not fly with NiMH it is not a plane.
Quote 0 0
MoTheG
Here is the most obvious way to do it with ten diodes and 5 comparators:
Align five pairs of diodes along the horizon, have one pair look at the top and bottom half, one pair looking at quadrants, one at octants and so forth. Compare the pairs and take them as digital digits, done.
If it can not fly with NiMH it is not a plane.
Quote 0 0
MoTheG
I drew a segment of the lens in front of the sensor line:
Linse.png 
There are e.g. triangular pits in front of every sensor for those sectors that are supposed to be blind spots.
The refractive index should be low.
( prior art )
If it can not fly with NiMH it is not a plane.
Quote 0 0
MoTheG
Quote:
It takes the aircraft 13s to complete a 260 m [s]diameter[/s]. 41.4 m radius circle.

That is not diameter but of cause circumference.
If it can not fly with NiMH it is not a plane.
Quote 0 0
MoTheG
Method to attain a correction vector for acceleration based estimate of attitude and velocity.
To continuously estimate the attitude and velocity based on 3 axis turn rate and acceleration sensors, it is necessary to separate / distinguish gravity acceleration from change of movement acceleration based on current measurements and the estimated attitude and speed. This can not be done for any prolonged period without additional data or plausibility checks.
With all estimates being correct the gravitational acceleration should always point down in a terrestrial reference system.
To check this, the estimated gravity vector is transformed to the terrestrial reference system and integrated over time until the next correction.
At the end of the chosen correction interval the resulting vector is compared to the expected vector, the difference is the error that now may be used to correct the current estimate. The attitude is easily corrected because its length does not exist, for the velocity the interval duration must be taken into account.
The big problem with all this is that "gravity vector is transformed to the terrestrial reference system" as well as the separation of measured / sensed acceleration into gravity and change of motion both prerequisite knowledge of speed. It may happen that you get into a confirming circle that will lead to mathematically sound yet wrong estimates.
To make this work I would always use a 3D magnetic compass, it costs little extra and raises precision, even though it does not provide one of two and a half attitude angles (attitude consists of one unity vector and one angle or two and a half angles) at any one time it will provide that extra absolute heading information. Today they are easy to come by. Types for under 1mT field strength are: Bosch BMM150, NXP MAG3110, MEMSIC MMC246xMT and there are certainly more.
In the context of a fixed wing aircraft it can be assumed that the yaw is limited. Large yaw angles or yaw for prolonged duration may thus be used to check plausibility, but that is going into more complex modeling that requires the control inputs.
a_{mov}=a_{sens}-g=a_{tran}+a_{rot}(v)
v=\dot{\omega}r=\frac{a_{rot}}{\dot{\omega}}
If it can not fly with NiMH it is not a plane.
Quote 0 0
Flybyknight22
Quote:
You lost me after the words, Full absolute.

“Absolutely” Ian!
And on that note,..
It’s time for a taste of freedom break BB56!

60FF6905-EAFA-499A-8D77-7F823D94FF0F.jpeg 
And on the 8th day God said: "Go NAVY",.. and it was good! Pray like everything depends on God. Prepare like everything depends on you.
Quote 0 0