3d airspeed sensor?

[Jonathan Challinger]

Lets talk about designs for a 3d airspeed sensor. I've talked to Paul about this a bit (in the context of designing a barometer probe that is not sensitive to air movement)

I'm thinking something the size of a pingpong ball, covered in pinholes, with sealed baffles inside that form 6 separate chambers. The chambers would separate the surface into -X, X, -Y, Y, -Z, Z. Each axis would connect to a relative pressure sensor. Getting 3d airspeed is then just a matter of a bunch of math.

In addition, a second pingpong ball with no baffles and an absolute pressure sensor inside would provide the average pressure over a sphere.

Then, we'd do some math to correct the absolute pressure sensor for the lowered pressure during fast forward flight, eliminating the drop in altitude. Plus, we'd have 3d airspeed, so we could fuse that into the EKF and become more robust to GPS failure. That would allow wind estimation on a copter, which is valuable when you want to drop something very precisely.

So, can such a probe be designed so that it doesn't look like spaghetti and meatballs? Can it be cheap? Does it have to be fragile? Does it come in black?

[Andrew Chapman]

One challenge would be how to mount it to be unaffected by the rotor wash on a copter. I imagine it would need to be on something like a 2ft tall mast in the centre in order to be clear of the rotor inflow effects on your average size multi-rotor?

AC.

[Jonathan Challinger]

It doesn't really NEED to be unaffected by rotor wash to improve height measurement.

For the 3d airspeed sensing aspect of it, yeah, rotorwash is a big concern and it would probably need to be placed very carefully.

--
You received this message because you are subscribed to the Google Groups "drones-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to drones-discus...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Randy Mackay]

 

     Excuse my slight lack of imagination but when do you think it would be useful?  Some possible ideas below to get things started:

          When GPS isn’t available (i.e. indoors)

          Adds another piece of information on the environment which will make EKF more robust (bus also possibly adds one more thing to go wrong)

          Allows users to set a desired air speed instead of a ground speed (although I can’t immediately imagine why they would want this)

-Randy

--

[Evan Slatyer]

I can see this being very useful for planes.

- For autonomous takeoffs, it could detect the crosswind while sitting on the runway
- Knowing sideslip angle allows you to drive that towards zero for better efficiency
- Knowing angle of attack allows you to fly just fast enough to avoid a stall


On an implementation note, what about using absolute pressure sensors everywhere? The current PX4 airspeed sensor (MS4525DO) has a range of 2 PSI (from -1 PSI to +1 PSI) with 14 bit resolution, giving around 0.84 Pa/count. The MS5611 (altitude sensor on the PX4) claims 0.065mBar = 0.65 Pa/count in its fastest mode (256x oversampling) and down to 0.12 Pa/count in slowest mode (4096x oversampling, which is still quick enough for 100Hz output).

Advantages:
- Much cheaper. Each MS5611 sensor is well under a quarter of the price of the MS4525DO and the MPL3115A2 (not quite as nice) is even cheaper.
- You get absolute pressure measurements. Should be able to get the static pressure just by averaging all the sensor outputs so now only one ball is required
- More flexibility. I can't think of any reason why you might want to compare, say, +X to -Y, but with absolute sensors it's easy to do in software. With differential sensors it means adding one more sensor.
- Sensors are physically smaller.
- If I've done the maths right you actually get more range. Standard atmospheric pressure is 103.325 kPa. The MS5611 will measure up to 120kPa, or 18.675kPa over atmospheric, which is about 2.7 PSI. In the other direction it's got much more space available since it goes down to 1kPa (ie over 100kPa below atmospheric). Obviously there are limitations when you're operating at extremely high or extremely low altitudes, but I suspect these would be way outside the realistic range for any 'normal' APM-powered aircraft. Might be an issue for super-high-altitude balloons, I guess.

Disadvantages:
- Twice as many sensors required, one per chamber instead of one per chamber pair. On the other hand, you can have one static sensor, one +X chamber, one +Y chamber, and one +Z chamber - no need for all those negative chambers.
- Absolute pressure sensors don't have barbs. This is a pain with a pitot tube where you run hoses back to the sensor, but I think in the 3D sensor design it'd be fine (you'd mount the sensor in the cavity on a tiny PCB and just run wires out).

[benni...@gmail.com]

A good friend of mine prototyped a (1-D) airspeed sensor using a pair of MS5611 parts for a hang-gliding/ sailplane instrument called the Bluefly.  The problem he ran in to was zero-drift of the different sensors, something that doesn’t tend to be an issue in differential parts.  Not insurmountable with the right filters and sensor characterization, but not trivial either.

You can read about his exploits at http://blueflyvario.blogspot.com.au/2013/12/airspeed-experiments.html

--Ben.

--

[Jonathan Challinger]

Ok, how about 6 MS5611 sensors on one board (for simplicity) with a microcontroller onboard to talk to all of them and consolidate everything onto one i2c port. Then put two 3d-printed half-spheres over it that have holes and ducts in the correct places.

--

[Evan Slatyer]

Ben, good to know. Do you know if he looked at using the on-die temperature sensor to correct the data? The datasheet provides information on doing that, but as far as I can tell the pressure output is not internally compensated - it's left for the end user to implement. The MS4525DO differential sensor is internally compensated.

Cheers,
Evan

[Robert Lefebvre]

I think airspeed instead of groundspeed becomes very important as we move towards using helicopter to fly instead of just hovering.  It would allow us to use target waypoint speeds much closer to terminal speed, without running into problems like loss of altitude control, etc.  For example, think about Dave's 30 m/s run at AVC last year.  If there had been a 15 m/s wind blowing (totally plausible) would he have been able to set a 30 m/s ground speed? I don't think so.  If there had been a 15 m/s wind, and the copter was only capable of 20 m/s airspeed, you'd have to set a 5 m/s ground speed to successfully complete the run.

Jonathan: as for rotor downwash affecting it, I wonder if some kinda of algorithm analogous to "CompassMot" would be possible?  So... "AirspeedMot".  ???

[Jonathan Challinger]

No, much more complicated than compassmot.

[Paul Riseborough]

Anything that changes air velocity also changes static pressure and therefore sensed height. The best place on a multirotor would be on an extension pole above the rotors in the centre.

[benni...@gmail.com]

Yeah those plots are all residuals after the “standard” datasheet compensation algorithms.

[Jonathan Challinger]

The point of the 3d airspeed sensor is to measure the effect of airflows on the barometer and compensate it. For that purpose, I don't think it really matters where you mount the thing.

[Paul Riseborough]

Not in the rotor wash please - power setting will change sensed static pressure and different flight speeds and directions will put it in more or less of the rotor wash leading to erratic readings

[Randy Mackay]

 

    Ok, I see.  So if this would help us get rid of the drop in altitude we see when the vehicle flattens out after fast-forward flight then I suddenly love the idea!

 

-Randy