A couple of blogs ago, I introduced the background on what we call the “hard tune” calibration technique for the Gen 2 AOA system. By hard tune I mean manually inputting an “aircraft curve” and “set points” into the code—an allusion to manually tuning a radio, like we used to back in the late Paleozoic era. This is a two-step process: first you program an aircraft curve, then you program AOA set points. We are currently working on methods to simplify this and, ultimately, to automate it; but for folks that are interested in the sausage making, I thought I’d share our progress.
Lenny has developed a probe configuration tab in the Arduino software where we can input aircraft curve equations derived from flight test. This page in the software is shown in Figure 1:
As I pointed out in my last calibration blog, the flight test technique to find the aircraft curve is to fly three or four-leg GPS speed runs at different flap settings at 5-7000 feet pressure altitude. Post-flight, we look for stable parameters on each leg, and average low-amplitude pressure data. The pressure sensors are very sensitive, and we are recording data at 64Hz, so we have lots of data to cull. In the case of the RV-4, we fly runs at Flaps 0, flaps 20 and flaps 40 (each flap setting we want to develop a curve for). The autopilot helps with high speed runs, but slow speed runs need to be hand-flown.
I use an in-flight run card where I record manifold pressure, RPM, OAT, IAS, fuel on board, GPS ground track and GPS ground speed. I also have a forward-looking camera on board, which is helpful for verbal notes and post-sortie reconstruction. I’m used to doing this alone; but there is a lot going on and you may find it more efficient to bring along a flight test engineer/spare set of eyes/data recorder to help out. This allows the pilot to concentrate on precise aircraft control. As you fly high-power, low-speed points, keep an eye on CHT and oil temperature.
Post-flight, I conduct a review of the video and pull the 1 Hz Dynon log file to refine the run card with some averaged data that will subsequently be input into our data reduction spreadsheets:
Our test RV-4 is equipped with dual Dynon DY-10A EFIS, which record these data 1 HZ but do not transmit all of them via serial, which means I have to download the 1Hz log file from the EFIS in addition to data from the ONSPEED box. Newer EFIS systems usually transmit all of the data required for reduction, so it’s easier to find what you are looking for in a single step. With the first-generation Dynon system, I have to look through the 1Hz log file I download from the EFIS to find the second or two of high-rate data I want to look at in detail to find pressures. Once I know where to look, I pull the appropriate high-rate data lines for analysis. Needless to say, this takes a bit of time; but the good news is you only have to go through this process once to establish the aircraft curves—they won’t change.
We are still developing the spreadsheets that we are using for data reduction; and will eventually post them, but at present they aren’t ready for prime time. Figure 4 shows an example of what the 1Hz Dynon data look like, and Figure 5 shows an example of high-speed data captured by the ONSPEED box. When the flight test boom is installed on the airplane, those data are also recorded in this file.
With good pressure data for each test point, we can generate a curve for each flap configuration. For the Dynon sensor, the curve is parabolic, so a second-order polynomial function is used by the software. The good news is that computers are excellent at math, and Excel is excellent at deriving 2nd order polynomials for the curve; so even history majors can generally work unsupervised. The R squared value under the equation in the chart is indicative of how well the data correlate to the curve equation. It's a statistical value, and the closer the number is to 1.0, the better the correlation.
To QC each curve during post-flight analysis, we normalize the pressure difference between the forward and angled port by dynamic pressure. When you plot these normalized data, the points should correlate linearly—if they do, then you know the curve itself is “tight” and can be programmed in the software.
Figure 8 and 9 are the Flaps 20 and Flaps 40 curves respectively for the RV-4. Note the similarity between these curves for the RV-4—again, more testing required to determine the feasibility of fusing them together.
The shape of the aircraft curve is important and differs based on the type of sensor that we are using. Each type of sensor has unique pressure signatures. So far, we’ve only tested the Dynon sensor as well as the Alpha Systems sensor used in the original FAA project; and we are still learning about those. In a couple of weeks, we’ll start testing a G3X system with the Garmin AOA probe. In a future post, I’ll talk about how we derive the tone set points (i.e., specific angle of attack references) to program into the software.