Hi Kim,

I have some engineering questions for you.

1) What sensor device do you use for bellcrank position? I did a brief search and didn't see any hall effect angular sensors as small as yours and with as few wires (pins).

2) I see you have ball bearings for the bellcrank shaft, which must smooth out any sensor position noise. How come the hall effect sensor is not rigidly assembled into the bellcrank shaft passage way? That must add to noise and inaccuracy.

3) Do you have to calibrate the bellcrank range of motion? (seems like an obvious yes). Do you do that just once at home?

4) What is your sampling rate on bellcrank position?

Thanks
Chris

Good info Pat

Why not have the sensor assembly tube attached to a slotted adjustable plate on the top of the wing?

I do have a question, may have been asked before but, if you loose line tension, does the bell-crank return to neutral?

Mike

Okay, so it works where to now?

4khz?  Perhaps you mean industrial servos?
Normal R/C servo updates are every 20msec, so only 50hz.
Digital servos can operate at a max of about 300hz.
This might seem too slow, but is in line with how fast you can move your hand and the servos speed. With a fast servo the elevator keeps right up with control inputs.
This was certainly the case with the indoor model, to me it responded just like the regular control version.


In this system the update rate is about 160hz, based on having enough time to output the three 1-2msec servo position information.
Ultimately the goal would be to go to 250hz, but that would require a more complex interrupt handler to be able to output all the servo signals overlapping each other.
Either that or use a micro with multiple hardware PWM outputs.

The bellcrank position is sensed and all the calculations done as part of the main program loop. This takes ~1.5 msec, faster than the servo positions can be updated.

The sensor used is not a rotary one per-Se. It only becomes one when installed between the magnetic field that rotates with the bell crank.
 Might be why you had trouble finding one in your search 

Part number is SS494B, chosen for its high sensitivity and ratiometric output. 
High sensitivity translates to good voltage swing and the ratiometric output means it behaves in circuit just like a pot would.
This, combined with using supply voltage as the A/D reference, gives supply noise/variation rejection to the crank position.
Spec sheet is here:
http://sccatalog.honeywell.com/imc/printfriendly.asp?FAM=solidstate&PN=SS494B

It measures flux density through the face of the part. At zero flux the output is Vcc/2.
It and the magnets are set up so neutral bellcrank has it at right angles to the flux lines, so no flux through the face.
Rotating the crank puts the flux lines at an angle to the face, so flux density is proportional to the bellcrank deflection.

By the highly scientific method of trying the magnets we had on hand and the spacing between them
 we settled on ones that gave full sensor output at the crank deflection we wanted to have.

Pat MacKenzie

So, On the shelves by Christmas?

We used an Arduino UNO to develop the code, but to save weight also permit a perf board style prototype used a DIP version of the ATMEGA328 flashed with the Arduino bootloader.
This meant we could use the UNO to flash the code.

Pro mini might have enough program space.

ESC was not driven by the flight controller. Didn't want to slow down the frame rates to put out another signal, and we had timer/processors already on hand for the motor.
Ultimately a single processor for both functions would be better, should be possible with a better pulse generator.

Yes.
The two magnets produce a strong area of flux with reasonably straight lines in the centre where the sensor is located.
In theory at least this will improve the linearity of the sensor, and makes its location a bit less critical.

It should also increase the rejection of external magnetic interference.

Pat MacKenzie

Not that I know anything about this but did you consider a potentiometer?

  Pots are not very reliable in any sort of vibration environment, even electric motor vibration. A wire-wound would work OK for a while, a carbon track, probably not all that long.

    The Hall effect sensor has no wear elements. A resolver would be as accurate and also doesn't wear out, but requires more processing and is probably heavier.

    Brett

This brings up a couple of theoretical questions:

1) Is mechanical motion really necessary at all?  The primary function of the control handle is to create a difference in tension on a closed loop of steel cable.  Would it be possible to just sense the differential line tension with a couple of load cells without any actual motion?

2) With FBW does the bellcrank need to be mounted in the middle of the wing?  Is there any advantage to mounting it in either the inboard or outboard wingtip?

Paul

How about optical encoders?