Auto Cal 101
Our flight test and code revisions have focused on the (mostly) automatic calibration capability we cleverly call the “auto calibration wizard.” The intent of this WiFi function is to walk the pilot through a calibration. An auto cal is performed by logging on to the OnSpeed WiFi network (password ANGLEOFATTACK) and selecting the calibration source on the SYSTEM CONFIGURATION menu (either EFIS or internal IMU) using a smart phone or tablet [Figure 1]. This option is found on the SETTINGS menu. We currently support Dynon legacy, Skyview and AFS; Garmin G5/G3X; and MGL; and the Vector-Nav Systems GNSS/INS as a calibration source [Figure 2]. Unfortunately, we don’t currently interface with GRT avionics.
After the calibration source is confirmed or changed as desired, the SYSTEM CONFIGURATION is saved using the red SAVE button at the bottom of the page. The pilot receives a CONFIGURATION SAVED message when the configuration is properly saved. The pilot then selects the CALIBRATION WIZARD from the SETTINGS menu.
This opens the first page of the wizard. The pilot must input maximum allowable gross weight, test weight (actual weight at the time of calibration), best glide speed at maximum allowable gross weight, and the aircraft G limit [Figure 3]. Best glide speed at maximum allowable gross weight is the AOA associated with L/Dmax. The velocity for L/Dmax will change as a function of gross weight, but the AOA will not. An onspeed condition is mathematically related to L/Dmax. Onspeed AOA equals 1.73 times L/Dmax AOA. These basic relationships are used with the pilot-input parameters to calculate the start of the fast tone (L/Dmax flaps up), the onspeed band (bound by onspeed “fast” and onspeed “slow). Stall warning AOA is calculated as function of actual stall AOA.
After pressing the CONTINUE button, the second page of the wizard opens with instructions on how to fly a calibration run. After reviewing the instructions, the pilot selects CONTINUE which opens the third page of the wizard [Figure 4]. This page shows the deceleration display. This display shows current flap position and calibration data source. For an AOA system to function correctly, it is critical that a calibration is performed for each flap position. The ONSPEED system accommodates up to 5 flap positions. The pilot only uses the number of flap positions required. For example, in my RV-4 with manual flaps, I have three curves: one for flaps 0, one for flaps 20 and one for flaps 40. Page 3 of the wizard also has a deceleration display. This allows the pilot to monitor deceleration rate during a calibration run. I’ve found during flight test that moving the smoothing slider to the position shown damps the arrow nicely. At the top of the page, there is a blue RECORD button. This is the button the pilot uses to begin an automatic calibration.
Before beginning an automatic calibration run, the pilot should check flap position is reading correctly and the data source (either EFIS or internal IMU) is set as desired. The pilot should also check to ensure the deceleration arrow is moving. We refer to the arrow moving as a “live wire” during flight test. If the arrow is frozen, it’s necessary to refresh the page. Once the moving arrow (“live wire”) is confirmed, the system is ready for an auto calibration run.
To fly a calibration, the pilot accelerates to maximum speed in level flight and trims the airplane. After pressing the blue RECORD button (which turns into a red STOP! button when the system is in the calibration mode), the throttle is pulled to IDLE. It’s important to smoothly increase the angle of attack as the airplane slows down. Exact deceleration rate isn’t critical, but a rate of about 1 knot per second is desired. This will put the deceleration arrow in the green band on the display. Do NOT chase the deceleration arrow. During the initial portion of the run, the deceleration rate will exceed 1 kt/sec. Continue the deceleration until the airplane stalls. When the stall is detected (See “Capturing the Stall” below), the system automatically generates a calibration curve [Figure 5].
It is important that the pilot smoothly increase angle of attack as the airplane slows down. We tend to associate pitch with angle of attack, and while we use the elevators to control AOA, a smooth increase in AOA generally requires varying pitch rates. You feel AOA, but you see pitch. Rather than making excessive pitch inputs, it’s fine to think “smooth” increase in back pressure and allow altitude to drift. Plus, or minus 500’ is fine. Trust your butt. It may take a few runs to generate a smooth curve with an acceptable “R2” value. And calibrate in smooth air! The system is designed to work well in rough air; but smooth flying and smooth air is the key to a good calibration.
Reducing a large number of points on a scatter plot to a single line is a statistical technique called regression. “R square” value is a measure of statistical goodness. A high value means there is good correlation between the data points and the line derived from those points that can be expressed as a mathematical equation, referred to as an algorithm. The computer uses the algorithm to compute actual angle of attack by computing the coefficient of pressure and inserting it into the calibration equation. A good calibration will have a high R square value, usually .98 or greater. The data will form a relatively smooth line, and the regression line will be nearly linear or slightly concave.
The “Save Data to File” and “Save Screenshot” buttons can be used to save data after a run. This is helpful for troubleshooting, so it’s recommended that both functions be used each run. The red “Save Calibration” button is used once the pilot is satisfied with the quality of the calibration.
A video demonstration of automatic calibration using an EFIS calibration source can be viewed here https://youtu.be/4ZN5XmLhtl0 and a video using the internal IMU as calibration data source can be viewed here https://youtu.be/DdqQm3fRhdY .
WiFi Network Stability
We’ve had issues with network stability in the RV-4 when navigating to and using the SYSTEM CONFIGURATION and AOA CALIBRATION WIZARD menus. When a glitch (technical term for “malfunction”) in the WiFi occurs, the system re-boots. A new log file is created each time there is a re-boot. Glitches also manifest as missing data or a random data point (or several) when the auto calibration generates a calibration curve. We don’t know what causes the problem yet. Lenny’s personal WiFi performance has been rock-solid, so he hasn’t been able to duplicate the problem. The RV-4 is max’d out with two different WiFi networks, a wireless air data boom and a high rate GNSS/INS system; but some of our beta testers have also suffered the occasional random re-boot; and errant auto calibration data, so it’s not aircraft specific. Overall, more annoying than serious issue, but we’ve got more work to do.
Capturing the Stall
As we’ve installed ONSPEED boxes in more airplanes, we encountered an issue we hadn’t seen in the RV-4: failure to detect the stall during auto calibration. The reason a stall wasn’t detected was due to our definition of stall. Stall was defined as a smoothed pitch rate change of 5 deg per second in either direction AND the nose below the horizon (negative pitch angle). Well, we proved once again an airplane can stall at any attitude (and any airspeed, for that matter). Figure 7 is a plot of a series of four “mushy” stalls. AOA, pitch (both IMU-derived and VN-300 reference gyro) and smoothed IMU-derived pitch rate (degrees per second) are plotted. Here you can clearly see the airplane generates sufficient pitch rate at the stall (about 6-7 deg/sec with the solid red line at -5 deg/sec) but the nose remains above the horizon (positive pitch angle vs dotted red line at 0 deg). To terminate the run, a healthy unload (stick forward) was flown. The unload resulted in significant pitch rate change as well as a negative pitch angle.
Figure 8 is the actual auto calibration curve generated during this run. I was using the secondary system in the wing with the “dirt cheap” probe attached. The large number of data points on the right side of the plot are a result of spending lots of time in and out of the stall prior to the unload which triggered the system to calculate a curve. Here's a video of this test: https://youtu.be/3SxYjuOkUV4 .
As a result of these tests, we changed the definition of stall to a pitch rate change of greater than 5 deg/sec. We also found out we had a sign error in the code and had recorded pitch rate reversed. Some airplanes, depending on flight control rigging and center of gravity location will simply "buck" at the stall as opposed to generating a clean break, but the pitch rate should be sufficient to stop auto calibration and generate a curve. 'Course, we haven't tried this yet in an Ercoupe...
IMU Vibration
From the “geewiz” department, it’s interesting to look at vibration effects on the IMU. With the fixed-pitch Catto prop on the RV-4, we typically see quite a bit of noise at high power settings. We don’t employ any shock mounts, so airframe noise is translated directly to the IMU. The VN-300 reference gyro is “hard mounted” to the top of the ONSPEED box, so it too suffers from engine/propeller vibration translated through the airframe. Figure 9 is a plot of IMU and VN-300 pitch rate, which shows the engine/propeller vibration effects. Note the vibration present at high power that smooths out as the throttle is retarded to IDLE. Also note that pitch rate is recorded in deg/sec for the IMU and radians/sec for the VN-300. The zero axis are offset slightly: IMU is the solid black line and the VN-300 is the dashed black line. Less than stellar plotting technique, but it gets the point across by showing relative magnitude.
Terry’s RV-8 is equipped with a Lycoming O-360 and a three-blade Whirlwind controllable propeller. Figure 10 shows the vibration characteristics of this engine/propeller combination. In this plot, Terry was using two throttle reductions to control the deceleration rate for an auto calibration test. Note the vibration as power is reduced through the “keep out RPM zone” where the high compression engine is prone to increased vibration.
“Pressure” Noise
Since integrating V3 hardware in the RV-4, we’ve swapped out three different boards. The quality of the auto calibration was not as good as we had previously seen with the old V2 hardware. The stick monkey noted more variability in the plot produced by the calibration wizard, which at first, the engineers naturally denied since the hardware was “identical.” Pilots have been causing engineers significant stress since December 1903, when Wilbur lost the flip 😊. After much nashing of teeth, we started looking at Pfwd and P45 pressure signals. After we agreed we had a noise issue, the next step was to figure out where it was coming from.
Our first rabbit hole was the Dynon AOA probe and associated plumbing. Since the noise was present in a pressure signal, we thought we may have had a plumbing leak or an issue with the P45 port on the tube itself. So, our first test was a leak-down check and modification of the cockpit P45 manifold to eliminate the D10A EFIS’s as a possible leak source. After we figured out the plumbing was tight, and ensured that there wasn’t any blockage, we looked at the face of the probe. There was some corrosion where the factory anodizing had failed, so we cleaned that up and alodined the surface. The next test flight showed no improvement; and since we previously ruled out any ringing issues, we started looking at an electrical issue.
It's important to note that Lenny, Terry and Tron's airplanes equipped with V3 hardware haven’t suffered any noise issues with pressure signals, so whatever we were seeing was RV-4 specific. The only difference is that the RV-4 incorporates a digital OAT sensor that interfaces directly with the V3 hardware since the antique Dynon EFIS doesn’t include OAT in the data stream. Lack of density ratio in the data stream (and high quality TAS information) has been an issue; and I’m glad we were able to rectify that. So, likely whatever was causing the issue had something to do with the addition of the OAT probe…Figure 11 shows the problem, a noisy P45 signal with the airplane sitting on the ground:
Note the difference in magnitude with the old V2 signal for comparison. Since we ruled out a pneumatic/plumbing or probe issue, we initially focused on the new OAT circuit. The digital sensor is mounted next to the Dynon probe under the right horizontal stab on the RV-4. A shielded three wire harness extends the length of the fuselage. We initially installed this without connecting the shield to a ground; so, the first mitigation measure was to ground the shield. Testing showed this to have no effect, P45 noise remained. After additional head-scratching, we started to disconnect the components directly connected to the V3 hardware: the VN-300 GNSS/INS reference gyro, the air data boom, and the modified M5 Stack visual display. Figure 12 shows the result. In this case, we started off with the air data boom disconnected:
And, voila, we had our suspect: the circuit powering the M5 visual display…Only took four and half months to get to the head slap moment—engineering in progress! Turns out there was a difference between the rather primitive V2 design and improved V3 hardware: the addition of an RS-232 chip. In over-simplified pilot terms, something in the M5 power circuit (likely the custom-made power supply in the M5 stack) doesn't play well with others. The problem only manifests itself if the V3 box powers a digital OAT AND modified M5 visual display simultaneously. The RV-4 is the only airplane flying in that configuration. I’m currently flying with the M5 display disconnected while the engineers work a fix.