No plan withstands first contact with the enemy, and so it was with our first attempt to modify an inexpensive pilot/static sensor to add angle of attack capability. A couple of blogs ago, I mentioned a skunk works project to develop this probe. It was based on pitot/static sensor shown in figure 1. This commercial, off-the-shelf sensor is made from two ¼” OD aluminum tubes welded together, with the lower tube capped off with a small bullet-shaped plug and a static hole drilled in the side. The only machined parts are the press in bullet plug for the lower static tube and the upper flange.
Our first iteration was to simply add a third tube to the bottom of the assembly to measure “P45.” Ben fabricated a three-tube flange and plug fitting for the static (middle) port tube. This is shown in Figure 2. Astute readers will note that we were learning to weld aluminum, and it showed! We bravely quit trying after almost burning through the tubes and used some JB Weld to finish up our test article. We fabricated a new inspection port test rig and mounted it to the RV-4 (Figures 3 and 4).
New WiFi Capability
Until now, the second V3 system mounted in the left wing of the RV-4 has served to record data from the secondary AOA probe and the EFIS. These data were used in post-flight analysis only, so there was no need to connect to the wing mounted V3 via WiFi. With the addition of automatic calibration, I can glean a lot of information using an iPhone from the system by running an “auto cal” and we can also reprogram in flight. This is a big increase in test capability and allows me to test multiple curves in a single sortie, greatly increasing productivity. It’s also a nice work-around for not having a flap position signal to the wing-mounted system…now I can configure the airplane as desired, run a quick automatic calibration and then gather test points at that flap setting.
Thus, it became desirable to have two WiFi networks in the RV-4: one for the primary cockpit mounted system and one for the secondary wing mounted system. We knew from previous experience, that we couldn’t see the wing network in the cockpit; so, we made some physical and software modifications. A close look at Figure 3 above will show 2 ¼” holes cut in the V3 box. Additionally, some ½” diameter holes where cut in the inspection plate to serve as WiFi “windows.” The holes are covered on both sides with clear, high-speed, high G packing tape. Additionally, we dialed up the WiFi volume (strength) to 15 dB and renamed the wing network to deconflict with the cockpit network. These modifications resulted in the ability to select either network in flight.
Initial Test: Wow, this thing sucks…
On 10 May, we flew the initial three tube configuration. The screen shot of the first auto calibration run is shown in Figure 5.
After looking at the nice VN-300 curves from previous sorties, this was a bit of a shock. A couple of things stood out immediately: a noisy pressure signal, a “convex” regression curve and a wacked out (technical term for "not quite right") stall speed. The low stall speed meant we had an issue with the quality of the static pressure, i.e., significant static source pressure error. Things didn’t get any better with flaps (Fig 6).
I ran some additional tests to generate a bit of data, but the results of the automatic calibration were sufficient to call this one a bust. You can watch the 10 May sortie here: https://youtu.be/aTvUB-IUo5o .
Plan B: Modify the sensor
After looking at the 10 May data, we determined the noise was coming from the lower, bent “P45” tube. Intuitively, it seemed as though a smaller diameter tube or opening might eliminate some of that variability. We also knew the static pressure error was excessive, so we decided to re-use some parts and cobble together a modified two-tube sensor that eliminated the static pressure sensor. Our “store bought” probe in Figure 1 had a bullet shaped plug in the lower static tube, so we decided to pull that out and see if we could drill a small hole in using a drill press. We fabricated two new tubes and bent the lower one down 45 degrees and offset it 1” behind the upper pitot inlet. We also increased stand-off distance from the lower wing from 5 to 6” with the intent of getting it further into the free stream. The result is shown in Figure 7. A friend proficient in welding aluminum took care of that, so no JB Weld required for this iteration.
We removed the initial three probe sensor and replaced it on the inspection plate test mount with the modified design (Figs 8 and 9).
Testing the Mark II
I was able to fly the “Dirt Cheap MKII” on 11 and 13 May. It was apparent after the first auto calibration run that we fixed the noise issue and had a normal looking curve. What was interesting was the shape of the raw data (Fig 10). Although we got a good regression, with a normal “convex” curve fit; the raw data were not as smooth as we previously observed using the VN-300 IMU platform. We also saw a bit more error with the new sensor using auto calibration (Fig 11).
Turns out, that not all gyros are created equally, and we have never “aligned” the Dynon DY-10A EFIS system since installing the VN-300 reference gyro in the RV-4. Just as we discovered as we’ve made the transition from our original IMU platform to the newer “twenty buck chuck” in the ONSPEED system, the quality of electronic gyros varies. The Dynon DY-10A is a first generation EFIS system, and I’ve observed over the years it’s prone to presession and, occasionally, tumbling (always post stall with excessive yaw present). Figure 12 shows a plot of VN-300 derived alpha, boom alpha and Dynon DY-10A derived alpha. Some of the difference between the plots is due to gyro misalignment (about a degree or so; but we’ll need to confirm that after jacking the airplane up for an alignment check). The important thing to notice is how smooth the boom vane and VN-300 curves are relative to the Dynon curve. Since automatic calibration of the new probe uses the Dynon-derived AOA, which is not as smooth as VN-300 derived AOA, the result is the wavy pattern observed in Figure 10 above. Figure 13 is just a plot of VN-300 vs Dynon pitch during the auto calibration run—here you can observe the misalignment between the two platforms.
Eliminating the underwing static measurement results in good pitot performance with the two-probe configuration. For testing, the V3 in the wing is simply vented to the inside of the wing; but note the stall speed in Figure 10 is close enough for government work. Additionally, since we are measuring AOA relative to the boresite line of the gyro, the auto calibration using the Dynon DY-10A results in flyable cuing. Overall, the two-probe configuration is a viable sensor for use with the ONSPEED system. Flight test videos for 11 and 13 May are up on our YouTube channel, and we'll make data available for anyone that wants it--drop me a line.