Measuring very small movements accurately.

Could this be empirically inferred via differential pressure?

polymorph:
Differential capacitance.

However, I'm partial to the LVDT. Not easy to design.

We used LVDTs at the refinery I worked at prior to retirement. The largest covered a 12" stroke on a large hydraulically controlled 56" butterfly control value under PID control. They can be made to work very accurately but for long strokes it's difficult/complex/expensive to compensate for ambient thermo changes and effects. But given the proper electronic support circuitry a LVDT can give very good accuracy and reliability, but being a limited/specialty market I've never seen hobbyist affordable offerings.

Lefty

You can buy very high resolution absolute position optical shaft encoders (e.g. 5000 counts/rev) that have extremely low resistance to shaft rotation. Just attach the windvane to the encoder shaft and you are done. Here are some examples: http://www.automationdirect.com/adc/Overview/Catalog/Sensors_-z-_Encoders/Optical_Rotary_Encoders
RESOLUTEā„¢ absolute encoder system with RESA30 rotary (angle) ring

the problem as I understand it with rotary encoders, is that they are good for measuring movement, but not position. As I understand they generate (for example) 10 pulses per degree rotation. So if you count say 5 pulses you know it's moved through half a degree. But what you don't know is has it moved from 5 degrees to 5 and a half degrees, or from 70 degrees to 70 and a half.

also, they're not particularly cheap.

I may investigate the differential pressure technique.

cheers

The mention of rotary encoders made me think of the optical rotary encoder in the trackball I use. I don't know how many pulses per revolution it produces but it is driven from the trackball through a high ratio friction drive. I think the same mechanism is, or used to be in some mice. There is very little friction in the trackball.

...R

The easiest way is to use an cheapo accelerometer (+/-1G range) as a tilt sensor, do some math to filter the data and remove other noise affecting the measurement.

Not while it's moving/flying.

Robin2:
The mention of rotary encoders made me think of the optical rotary encoder in the trackball I use. I don't know how many pulses per revolution it produces but it is driven from the trackball through a high ratio friction drive. I think the same mechanism is, or used to be in some mice. There is very little friction in the trackball.

...R

I think you're on to something but maybe a joystick instead of a trackball.

Not while it's moving/flying.

BTW the aircraft in question is technically a 'model' aircraft, but it has a 30 meter wingspan (about 90ft if you're american) and weighs 35kg

30m wingspan aircraft's dynamics/vibrations cannot be compared with a small heli or quadcopter model..

You get a lot of atmosphere induced accelerations in a Boeing 747. Imagine what it's like in a 35kg airplane.

...R

You get a lot of atmosphere induced accelerations in a Boeing 747

Yea, at 950km/h speed and against 300km/h jet stream..

We used LVDTs at the refinery I worked at prior to retirement. The largest covered a 12" stroke on a large hydraulically controlled 56" butterfly control value under PID control. They can be made to work very accurately but for long strokes it's difficult/complex/expensive to compensate for ambient thermo changes and effects. But given the proper electronic support circuitry a LVDT can give very good accuracy and reliability, but being a limited/specialty market I've never seen hobbyist affordable offerings.

And therein lies the rub. Or, if you prefer, the devil is in the details. I've been thinking about an LVDT shield for a while, but I haven't been confident of enough of a market. And the specialty ICs for it seem to be rather expensive.

OhMyCod:
But what you don't know is has it moved from 5 degrees to 5 and a half degrees, or from 70 degrees to 70 and a half.
also, they're not particularly cheap.

Absolute angle encoders don't suffer from that problem, because they report the absolute angle. They are available to measure to 1 arc minute accuracy. Some relative angle (incremental) encoders have a "zero" output which can be used to reset the origin, so with proper use they can also serve as absolute indicators. An encoder that would be suitable for your application may cost a few hundred US$ but what is that in relation to the cost of a 30 m airframe? Here is one on ebay that gives absolute angles to one part in 1024 (0.35 degree) PCA Absolute Shaft Encoder ANSL 12BS MBAM/01024 | eBay

Also, if you want accurate measurements, it would seem that the design of the wind vane is critical, in order to avoid turbulence and oscillation.

You can't measure the angle with respect to gravity using only an Acelerometer while the plane is flying. Even if you could, then you would also need the exact orientation of the plane.

It's much worse at 150 knots landing speed.

...R

pito:

You get a lot of atmosphere induced accelerations in a Boeing 747

Yea, at 950km/h speed and against 300km/h jet stream..

Another idea for a high accuracy low friction method would be to use something like this or this they both use non contact measurement so should be low friction and I don't think there is much magnetic drag.
EDIT: Another plus point is you can read off position from power on, compared to a simple A-B encoder that only counts pulses and direction but not where in the turn you are.

Riva:
Another idea for a high accuracy low friction method would be to use something like this or this they both use non contact measurement so should be low friction and I don't think there is much magnetic drag.

Perfect!

Speaking of magnetic drag, a (fixed) aluminium plate on the other side of the magnet would provide some damping.

How about a load cell or stress sensor?

OhMyCod:
What's the best way to measure very small mechanical movements very accurately?

I'd like to make an instrument to measure the angle of attack that a planes wing makes with the air. Mechanically it's very simple, I attach a freely rotating vane to the front of the aircraft and measure the angle this vanes makes relative to the airflow.

relative to the airframe, surely?

Option three - same as option two but using hall switches instead. Disadvantages - all the same disadvantages as for option 2, and costs a lot more.

No, not hall switches, a linear hall sensor (analog output). Arrange the range of movement to bring the magnet
pole from one side of the hall sensor to the other and the detected field strength will have a good range of values.

Calibrate by hand (be prepared to reposition the sensor a bit to get a good response curve).

You'll only need a very small neodymium magnet so very little weight. If you can arrange the magnet to counter
balance the vane's weight even better.