UAV ongoing learning project

[ALtitudeap]

I have been working on a uav that will use a Basic Stamp 2p chip. ( www.parallax.com ) Today i tested out my wing leveling. I am using the BS2p with the PSC Servo controller board. The psc drives the Aileron servo. I am using the Memsic 2125 Accelerometer that is mounted to the wing. I had it all mounted in a GWS E-Starter. It worked remarkably well. it was a very positive move in the right direction.

[Unterhausen]

That's great. Is the accelerometer an I2C device? If not, how are you interfacing to it?

[ALtitudeap]

it provides constant inputs to the basic stamp. Here is a doc on connections. http://www.parallax.com/dl/docs/prod/compshop/SICMemsicTut.pdf This thing works great. I have decided to use the BS2p because it will do everything that i need it to do. I will monitor my altitude. Tilt. GPS (Heading Correction). It will comunicate to the ground via a wireless modem (7 mile range at 9600 buad) it will also control the angle of the onboard camera and control the shutter. I am using a math coprocessor to handle the heavy math.

[kd7ost]

Altitudeap,

Hows this project coming along. This is interesting stuff.

Dan

[ALtitudeap]

very good so far. i am using a MEMSIC 2125 for wing leveling. An oem gps (sandpiper from ebay) for the gps coord. a SN74AHCT157 for my autopilot mux. The MPXA4115A for the pressure sensor (altimeter) Flight battery monitor and electronics battery monitor through a ADC0834. A uF-FPU math co-processor to manage the "big math" and a 9xStream 9600 baud radio (7 mile range) for monitoring.

[andyg]

I don't mean to burst your bubble but if that plane ever turns in a coordinated fashion your accelerometer will read the wings as being level and it will crash to the ground.

[LukeZ]

Altitudeap, I'd been wondering the same thing as Andy brought up. How effective is the accelerometer over an extended period of time? Do you have any way of re-centering it in flight?

Luke

[ALtitudeap]

it has a motorola pressure sensor and i didnt want t linerize the non linear equation. so the coprocessor also handles the floating point math. Altitude=(10^(log(P/P_0)5.255879)-1)/-6.855856*10^-6

[reedchristiansen]

Altitudeap, I'd been wondering the same thing as Andy brought up. How effective is the accelerometer over an extended period of time? Do you have any way of re-centering it in flight?

Luke

An Accelerometer pointed out the right wing to measure roll angle won't give you much data over the long term. It will measure a slight blip as the airplane enters a turn then read level as the aircraft begins coordinating itself. You need gyro information to measure your bank angle.


A cheep and dirty way to do this is to buy the FMA copilot sensor head only. Ditch the rest of the copilot and read the analog voltages from the sensor head. They provide a nice continuous voltage (o-5v) in both the pitch and roll axis. The big problem is that the span of changes day to day, hour to hour based on the weather, cloud coverage, etc. I spent many months of my life trying to figure out a way to autocalibrate that thing in grad school. We made many autonomous flights with it, but each flight involved a few minutes of calibration.

When analog devices came out with the adxrs gyros we started using them for pitch and roll measurements (combined with the adxl accelerometers). A good way to go, and not too much math. If you have a couple extra multiplies and divides on your processor, you can do it.

Reed

[clolson]

A cheep and dirty way to do this is to buy the FMA copilot sensor head only. Ditch the rest of the copilot and read the analog voltages from the sensor head. They provide a nice continuous voltage (o-5v) in both the pitch and roll axis. The big problem is that the span of changes day to day, hour to hour based on the weather, cloud coverage, etc. I spent many months of my life trying to figure out a way to autocalibrate that thing in grad school. We made many autonomous flights with it, but each flight involved a few minutes of calibration.
Reed

I'm sure there has to be more too it, but I assumed that these devices drive an aileron deflection proportional to the temperature differential between the two sensors.

I can understand why the standard copilot would need to be calibrated because there could be parts of the airframe in the sensor view, or the sensor might not be mounted exactly horizontal (i.e. part way out a wing with dihedral.)

But if the sensor is mounted horizontally and there isn't any pieces airframe in the fov of the sensor, I don't see why a calibration would be necessary (unless there is a significant variation in what different sensors read when seeing the same exact bit of sky/earth.

Am I missing something, or is it just really hard to get perfectly matched sensors?

Curt.

[reedchristiansen]

I'm sure there has to be more too it, but I assumed that these devices drive an aileron deflection proportional to the temperature differential between the two sensors.

I can understand why the standard copilot would need to be calibrated because there could be parts of the airframe in the sensor view, or the sensor might not be mounted exactly horizontal (i.e. part way out a wing with dihedral.)

But if the sensor is mounted horizontally and there isn't any pieces airframe in the fov of the sensor, I don't see why a calibration would be necessary (unless there is a significant variation in what different sensors read when seeing the same exact bit of sky/earth.

Am I missing something, or is it just really hard to get perfectly matched sensors?

Curt.

Curt,

The big problem with the sensor is that the full scale reading changes with weather. Basically, the sensor is designed to read the differential in temperature between the sky and the ground. On some days, their is a huge difference in temperature (Noon in the summer when the sun has been heating the ground vs a snow covered field in the winter when the air and ground are close to the same temperature).

On the summer day, the voltage difference between level and a 90 degree bank might be 2.5 volts. In the winter with snow on the ground, etc, the difference might be .1 volts. This makes it impossible to convert sensor output to calibrated roll angle without calibrating before you fly. (Hold level, read voltage, tilt 90, read voltage). It is dificult to set the gain on the roll angle hold loop without converting the sensor voltage to calibrated roll angle.

If you are not intersted in knowing your roll angle, then you could just feed the raw a/d value into your roll angle PID loop. The problem their, is that you will need to re-tune your roll angle gains every flight. (We did this for a while until we figured out how to do a ground calibration of the sensor before each flight)

The other problem with the sensor is that the side of the sensor that is facing the sun, tends to heat up a bit, so you end up with an airplane that always turns away from the sun (More heat = ground = bank towards the sun, so the autopilot corrects with opposite roll)

Reed

[alexcmag]

Don't be a 3-axis magnotemeter better to hold attitude and heading?

[ElectroLawndart]

Hi all,
Use the accelerometer to make sure everything is coordinated and the magnatometer and or rate gyro to view heading changes. Accelerometer says "wings are level", rate gyro says "we are turning" means coodinated turn/bank. Rate gyro says "we're holding heading" means straight and level.
Dart

[ALtitudeap]

thanks dart