Saturday, September 26, 2020

EAA Webinar: Founder's Innovation Prize Grand Championship Preview

On September 24, the EAA hosted a webinar where a bunch of characters from the Founder's Innovation prize competition got to discuss their progress. My friends are all awesome and the Airball project is proud to be featured among them. The following is a video of the event, and it may require EAA access, but the EAA is also awesome and membership is quite inexpensive, so you owe it to yourself to join.

The slides from the Airball portion of the presentation are available here:

Sensor supply chain and COVID-19 calibration stoppage miseries

We are really close to getting our probes to a "done"-ish stage. Really really bloody close. So close. And yet.

First problem: Calibration is blocked by COVID-19

The first misery we're dealing with, along with everyone else of course, is COVID-19. This has caused all the wind tunnel facilities nearby to close down. And there with it goes any hope of real calibration. Now a good friend (not outing him in case he wishes to remain anon) offered use of his own homebuilt tunnel, but that's far away and would require a lot of setup and preparation, and honestly my day job is overwhelming at the moment. (With a nearby wind tunnel, I can afford to cheese my setup a little bit and, if it does not work, I can always go back the next weekend.) It is possible to imagine complicated devices I could fly on my plane to achieve the effect of a "wind tunnel" -- but that would require even more R&D time.

Second problem: Pressure sensor stock is limited

You will recall that, after much tribulation, we decided to go with 3.3V SMT mount SPI Honeywell TruStability pressure sensors. They are reliable, name brand, and easy to use. But the problem is that the range of pressures in stock is limited. You can get what you want but with terms like, minimum quantity 100 and lead time 7 weeks. Yikes. And this is in turn a problem because my "new" probe design, that eliminates the static probe "stick", tends to amplify the pressures. This is good, right? But this means we need at least one of the sensors to be a wider range. And this is difficult.

Look at this plot of dynamic pressure (q) versus IAS:

For a Cessna 172 and thereabouts, you can figure q < 0.5 psi. For a Cirrus or Mooney, we're getting into the range of q < 1.0 psi, in other words, the dynamic pressure doubles.

In the old probe design, with the static pressure stick, two of the measured pressures are in the range [-q, q], and one of them is in the range [0, q]. So for a Cessna 172, with VNE = 160 kias corresponding to q = 0.57 psi, a +/- 1 psi sensor, which is a very common part, is adequate for everything.

In the new probe design, which "amplifies" the pressures, two of the pressures remain in the range [-q, q] while the third is jacked up to the range [0, 2.3q].  At VNE with a Cessna 172, this would mean [0 psi, 1.3 psi]. So this means that, even for the humble Cessna, a +/- 1 psi sensor is really pushing it. Close but pushing it.

I could go out and order 100 sensors of the appropriate size and wait 7 weeks. That would be between $3,500 and $4,500 of parts, and I would be doing this on a hope and a prayer having done no wind tunnel calibration. That is a bit -- um -- let's say, gutsy?
In making the next round of probes, I have to use the 1 psi sensors that G-d has graciously provided us. There is no getting around that. Until I do a serious calibration, I'm not going out and buying expensive crap.

Right now, I have a probe board sitting on the bench that has three +/- 60 millibar (+/- 0.87 psi) sensors. These were purchased when I was using the "old" probe nose design. Might as well use them and, if they saturate, they saturate. We will at least get some partial data out of them.

Moving forward, depending on how things go with the new probe in flight, I might decide to go back to the old probe with the static "stick", or standardize on +/- 1 psi sensors and hope for the best.

In this case, "hoping for the best" means that, for the pressure sensor reading that is in the range [0, 2.3q], we would saturate at q = (1 psi / 2.3) = 0.43 psi. This corresponds to an indicated airspeed of 138 kias. This is a reasonable compared to the VNE for our tiny little LSA.

Saturday, September 5, 2020

Airball probe v8 update

It's been a while since there's been an blog post! Sorry for this. I've been completely slammed at work, and just putting in tiny bits of work on the project here and there. But we've been making progress!

This is the newest design for the probe, the "v8" mechanical and electronic form factor:

We are trying out the new scheme of having a 5th hole at 90 degrees at the bottom of the probe nose, to avoid having a bulky static probe. If that doesn't work out, we can always return to the old static probe without much change to the sensors (there's one sensor that would need to be swapped out for one of a different range, but that's minor).

This is built around a 1.5" x 3.5" circuit board that integrates an ESP32-WROOM-32U module, a Bosch BMP388 barometer, 3 slots for Honeywell SPI HSC pressure sensors, a battery charger, and a USB interface.

The temperature sensor is a TI TMP102, which is supposed to be on a tiny daughterboard. For now we don't have that and are instead jury-rigging a TMP102 breakout board stuck to the outside of the probe. The result looks like this:

The circuit board has two known bugs; everything else seems to work:
  • Missing a pullup resistor for one of the USB comms signals
  • The pressure sensors are too close together and so the plumbing is very poor
The mechanical parts need some more development to make them suitable for being a "kit", but they are getting there. Check out this album to see the process of assembly.

Link to photo album: Airball probe v8

Stay tuned while we get this working. Note that this signals a departure from the XBee comms we had previously: This one is just plain Wi-Fi. The plan is to have this thing set up a Wi-Fi base station with a unique name, and you can jump on that and telnet to a known host and port and get a stream of data messages. We tried UDP but we were getting a lot of losses; we might try that again.