On 5/18/12 North Texas Near Space (NTNS-2) achieved a maximum altitude of 87959 ft over a sparsely populated section of North Texas. The flight was partially successful and full duration telemetry and camera data were recoverd.
• Vehicle departed from controlled flight, did not navigate to the landing site and was severely damaged
• Modified SW version 2.50 to utilize new AHRS code
• Modified Stinger 64 airframe
• Surface winds calm
• Trajectory was close to prediction
• Ascent was nominal
• Ascent terminated early on balloon failure at 88k ft. APM released on free fall indication and routed autopilot outputs to servos. Balloon not recommended since bursts at 86k and 88k on NTNS-1&2. 2# free lift.
• Higher than expected descent rate experienced and the plane spun down to ~17000 ft and began semi-controlled flight. The A/F managed to fly despite issues with the attitude reference and control saturation.
• Impact occured due to additional loss of reference. The roll autopilot was tracking commands.
• Maximum Mach number was about 0.5
• Minimum internal temperature was 48F.
• Airframe held up very well. One aileron horn was damaged (inconclusive considering the impact).
• Just Plain Strange: Almost crashed on my ranch. The airspeed probe is re-usable.
• Likely root case (s):
- dynamic pressure measurement low by 0.5
- gain scaling issues causing pitch and roll oscillations.
- pitch reference and finally roll reference is lost
- roll reference loss responsible for final dive
- contributing: agressive airspeed targets
• Todo:
- duplicate pitch oscillation with the stinger UAV; may need to duplicate A/S measurement issue in SW
- simulation reconstruction
- new A/F build, upgrades, noted SW changes (see spreadsheet), merge to 2.60 or latest stable build, gains
Comments
Update: I found a pinhole in the pitot dynamic proble. This could be crash damage, a manufacturing defect, or an assembly issue. Potenially significant to others is that I have tested both my DIYdrones AS probes and they both leak from the dynamic to the static port. i.e. when I apply pressure to the dynamic port and plug the end I get "some" gas expelled through the static port holes. I would like to get a set of the old probes, but the seem to be out of print on all of the websites that I've checked. Decided to roll my own here since it's not exactly hard.
Starting NTNS-2 flight reconstructions with the Stinger UAV today or tomorrow. The issue here is that the pitch inertia is lower than what was flown on the HAB flight. LG
I agree with you Monroe but, as Larry points out, being rule-bound doesn't guarantee safety and harmlessness. The focus on achieving the safest practicable outcome is in my opinion much more important than meeting regulatory requirements. Most of the time these two goals intersect, but we all know that in sometimes lawyers will demand guarantees of 100% safety, whilst crossing the street to the courthouse to make the demand...
We are never 100% within the law on a daily basis and so I believe that whilst we shouldn't set out to break the law, like da Vinci and other great scientists before us, we shouldn't necessarily let the last letter of the law impede a worthwhile experiment...
This thread is really mostly about the data. Not necessary, but completed was a range safety analysis that showed < 1X10-7 probabilities of a major issue with aircraft. I haven't done the work but I believe it much more likely to hit a buzzard on the way down below 1500 ft (by several orders of magnitude). What is a major risk are the batteries and the likelyhood of grassfires. This is a safety consideration that did not get mitigated that a lot of us are VERY sensitive to. Besides considerations mentioned above I will be considering some mitigation to include elimination of the LiPO cells that were flown last time.
Once I get back from the Ranch and have some BW I'll post updated plot and data.
passerby, I do hope you take as much care to ensure that the Canada Geese and similar are all filing for the appropriate clearances and waivers before they V1 and rotate for migration. I'm certain they've all been TSA cleared and have filed IFR flight plans too...
The folowing link has video frame by frame from apogee + several other key frames where the autopilot was performing well. The orginal video was not completely recovered and video editing SW has some difficulty with frame count. This could be and issue later in the video when roll attitude and heading don't match up with telemetry for these frames. At the start of the video the attitudes seems at least plausable. I didn't include the raw magnetometer measurements on the video, but they are in the telemetry file. Note that I did update the original video with some altitude stamps. Still working on data, but thought I'd post the video. The attitudes appear to diverge later just before the final dive, but this video isn't available.
Regards, LG
Frames Video
Thanks Carl. The A/F was at angle of attack and pulling about 4G in the spin... which gets erratic and looks like a departure stall. We'll see where this goes.
Mostly what I would have wanted was near zero tip-off. Of course we got a bunch as Andrew observed. I don't know about the attitude, but it seems that distance could help with disturbance loading. I also wonder what axial loads are being applied. The response would depend on the initial condition.
The best answer is probaly dropping before burst. I will use a longer line next time out.
It would also be great to be wings level on ascent for pictures and such. I didn't figure out a way to do this and cleanly release. I'm sure there's a way.
Still working on video...
Hey thanks for the raw data Larry! It's fascinating!
You can actually see the massive pitch-up from nose-down to nose-up when the balloon ruptures!
Actually, thinking about this, I wonder if the explosive release of gas in such a low density environment is enough of a wind to disturb the airframe orientation? I guess with the balloon at full stretch, it could concievably have several "atmospheres" of gas pressure in it, given that an atmosphere at 88000' is 30mBar or thereabouts...
It makes me wonder whether the aircraft should be oriented nose-up on ascent.
BTW Larry, if you look at |A|, it seems like you could detect rupture release nearly 10s earlier than you did. Might be useful to help avoiding balloon debris
I read Carl's post and he mentioned departure stall (can't find this on this thread but don't know why). The Stinger will depart, but it takes an aft CG or really slowing it down. That doesn't mean that this isn't occuring (not sure the exact frame; yet). I had been attributing the departure to attitude matrix issues and A/P wrap point. But, it's best to open the aperature.
For those interested in the data, I have posted file that I'm currently working in (36M, .xlxs) Corrected Link
Stunning pictures Larry, less is more http://www.gpsboomerang.com/
The plane was about 20 ft from a 3 ft chute, attached to the balloon. These were components left over from my unguided APM ride along. You can see a picture or two by watching that video.
I started out trying to work with the APM climb_rate code that's commented out. This has issues with integer math at altitude and couldn't be used as is. I then used their idea and did a floating point version of a least squares fit, but was having issues with it. This turned out to be a C++ 101 duh, but by that time I had stripped out the code and replaced with a 10 point moving average for dgps_alt/dt. I looked at both algorithms during van testing and found the attenuation for the dumb filter to be about the same as the more elegant calculation of the derivative. The phase is a bit different, but since -7500 ft/min was arbitrary anyway the average was better anyway since it uses a few number of little numbers instead of a big number of big numbers. Both algorithms didn't seem to be issues for APM throughput.
I did consider |A| but the algorithm has a similar amount of pitfalls. Also, note that I'm only checking climb_rate after 60k ft (upper stratosphere) and only calculating it when NSATS >5. The latter condition was derived from NTNS-1 TM for this receiver. Summary: this worked great, but there could be a cleaner way.
I was also triggering on 100k altitude and 2.5h time.
I used a sailplane towhook release, used a servo on CH5 and modified the flap code. I had only used this method because I started out thinking that I wanted a two stage release with a drogue. The drogue code is still there, but the delay time is set to zero. This value was initially set based on time to reasonable AS (vaccum).
Results: it turned out that the vehicle doesn't seem to have issues with low Re and was about as stable as I see when put in a hammer head stall. I set the CG at the location where I have been flying the Stinger. By now I've had >150 flights on various versions. Unless... we weren't clear of the balloon debris and the vehicle is extremely stable because it has a big drogue.
BTW, I don't buy in here because of V v. t history. NTNS-1 and NTNS-2 had very different velocity profiles. I probalby should plot them.
I also haven't determined why we're rolling and why the A/P seems to catch the roll; almost, at various times in the flight. This could be asymmetric aero and not enough deflection and wrap points, or it may be just algorithmic.
LG
-
1
-
2
-
3
of 3 Next