Friday, November 27, 2020

Thinking about road testing probes

I have a new version of the probe PCB out for production at the moment, and I think this might be good enough for being the core of a "kit", so now we move on to other things....

Calibration of the probe is of course a huge issue. I've been trying to get wind tunnel time -- and I have a generous offer from a friend, but he lives rather far away. So I've been trying out various ideas on the side.

Remember the road tests of the probe? Well so I took out the probe on the road, mounted on a pole in front of my Prius, and tried smoothing the data, and it looks like I could get a reasonable reading if I average for, say, 1 second or so per reading. This might be enough for a "useable" calibration, while I wait for a real wind-tunnel run.

The problem of course is how to establish a "zero" for the wind direction. At the moment I don't know, but I've been thinking of making a "kite" mounted on a universal joint:

This is an articulating head boom, with the 3-sided fin geometry inspired by Andrew Angelotti's Model 23 probe. The parts are 3D printed, with 3 carbon arrow shafts and balsa fins. The fin supports have little compartments for mass balance, if needed (unlikely -- I expect this to balance tail heavy unless nose weighted):

On the end, I show a Robotis 2XL430-W250-T servo (2 axes in one, and one of the cheapest you can get that has the more accurate contactless magnetic encoders) with a pair of FR12-H101K hinge frames.

The idea is that, if the vanes of the "kite" are large enough, then the mis-alignment caused by the asymmetry of moving the probe side to side does not move the "kite" very much. This in turn means that I can rely on the inaccuracy of the relative wind being somewhat bounded.

The alternative is to mount the servo base rigidly, and try to "calibrate" it for the specifics of the car somehow. That seems error-prone and more difficult to set up, but easier to build.

As a final though: The action of the passive vanes to center the "kite" into the wind can be thought of as a form of PD (proportional + derivative) control -- there is a correcting factor proportional to the error, plus some damping. What is missing is the "I" in PID control -- an integral term. This is what would drive the long-term error to zero. And the way to do that would be to:

  1. Measure the wind direction relative to the probe base; and
  2. Use a second set of servos to redirect it into the wind.
The measurement need not be accurate -- it just has to read "zero" when the probe is directly into the wind. Typically you can do this with another air data probe, and just drive the left/right and up/down delta-P values to zero. And this is basically a nulling probe. But that's adding an actual feedback loop and a second servomechanism....

I don't know what I'm going to do yet. For now I'm just tooling around with this. Stay tuned!

Sunday, November 15, 2020

Probe V8 flight ... of sorts

Today was the first flight of Probe V8 with an actual Airball display. Things worked but were not as reliable as I would like. Some notes:

1. There were short periods of "loss of data" which might be caused by buffering in the OS, or might be genuine signal loss. I found no evidence of signal loss when I tried things earlier so I suspect this is some bursty behavior in the OS. Action item: Characterize data reception interval under real-world conditions.

2. The new probe shape is even less well-calibrated than the previous one! The data is there; it's just not reasonable. The airspeed is way off. Action item: Probe calibration.

In general, this is nothing that can't be ironed out with a little bit of work, and we are that much closer to a "kit" for people to build.