Vertical Trajectory System (with some 3D printed parts)

Discussions on high/extreme altitude and mach busting rockets.

Moderator: Moderators

User avatar
OverTheTop
It's only money...
It's only money...
Posts: 2658
Joined: Sat Jul 21, 2007 2:20 pm
Location: Melbourne, Australia

Re: Vertical Trajectory System (with some 3D printed parts)

Postby OverTheTop » Fri Sep 07, 2018 5:20 pm

Flight Diagnosis
Over the past 1.5 weeks I have managed to download the altimeter data and video files from the flight. A quick glance about a week back showed some promise that something had at least worked somewhat, although there was some curious funkiness in some of the plots. Here is what I have found in my limited time so far.

The flight was on a fairly punchy K1103X motor, with predicted boost accel peaking at 16.4Gs. As it climbed off the pad it seemed to take a bit more of a turn to the left (towards horizontal) than I usually experience on my flights. Curious. The rocket deployed and recovered nominally about 1km from the pad, in the same paddock. A flight later in the day that did something remarkably similar I wasn’t particularly concerned.

Initial download of the logging showed that there was probably some smoothing in the data that seemed to iron out some of the high-frequency (2.5Hz) components in the PWM stream. A quick query in to Keith and Bdale regarding the frequency response of those analog inputs on the TeleMega confirmed the logging rate as about 3Hz, so 2.5Hz will not be visible. However, ignoring the smoothing that seems evident the plot of the four servo PWM signals showed some unexpected wanderings during boost. They should have showed flat-line during boost, control during coast and flat-line again after NC deploy. What it did show was that all four lines tracked each other during the boost phase, did their own things during coast and tracked together flat-line as expected after apogee. Seriously weird behaviour in boost.

Note region A where the PWM signals are tracking but varying. They should be all horizontal like in region C.
Region B is where the VTS should be attempting to control the flight and signals are expected to vary and differ.
Region C the flight controls should be back to a steady neutral position, and they are indicated as such.
Logging1.png


Looking at the Pitch Roll and Yaw, you can see two are in-phase (roll and yaw) and one is phase shifted (pitch). This is what showed up as Dutch roll in the flight and on video.
logging2.png

Zooming in a bit makes the phase relationships more lucid.
logging3.png


Finally, looking at tilt angle you can see that there was a bit of a tweak down low and then some progressive increase of tilt during boost. Just after commencement of coast the VTS kicked in and there was a sudden decrease in tilt (10degree improvement) and then resumption of original trajectory.
logging4.png


As mentioned earlier there was some apparent roll control, with the roll axis oscillating around a fixed value. It could be roll instability (too much gain) or some Dutch roll introduced by the control system and airframe interactions. Looking at the video from the side-facing camera it seems to be Dutch roll, and I will confirm this by looking at the logging soon.

Disassembly
So, with the unit back on the bench, batteries replaced and charged, I powered it up. Feeding in some deflections to the servos showed that only one of the four was moving. The servo supply 20A SBEC (switchmode battery eliminator circuit) output was measured. 12.1V in, 12.1V out. :( . That should be 5.5V out to the servos.

Everything was powered down and the servos individually tested. Three of the four were indeed completely non-functional due to being force-fed by the SBEC. A quick autopsy of one of the servos shows a crater in the bridge drive chip where the magic smoke escaped.
FriedServoSmall.jpg


Root Cause
The SBEC was cut out of the clear heatshrink that enclosed the unit. Inspection under the Mantis showed that there were quite a few large solder balls evident on the top of the PCB (circled in yellow).
SBECSolderBallsSmall.jpg

I am willing to bet somebody else’s left testicle (I never bet my own) that one of the solder balls (perhaps not even one of the pictured ones) moved around under boost acceleration and shorted out something on the PCA. This caused the main switching FET to go short-circuit and apply a very solid 12V downstream to whatever got in its way (servo x 4 and a servo reverser).

Pics of SBEC

Luckily the pulse generator and autopilot run off a separate BEC so they should be in much better condition than the servos. The two PCAs I designed are capable of running up to 30V input supply so they will also be not a problem.

Next Steps
* Continue to analyse data and see if anything worthwhile can be extracted
* Purchase new servos and SBEC
* Flight-proof the SBEC by cleaning/inspection/conformal coating the PCA
* Rebuild the VTS and test
* Fly it again!


All that will likely happen next year now, as the crops are getting up and fire season is almost upon us.
TRA #13430
L3
"Everybody's simulation model is guilty until proven innocent" (Thomas H. Lawrence 1994)

User avatar
snrkl
Rocket Flyer
Rocket Flyer
Posts: 94
Joined: Wed May 10, 2017 2:07 pm
Location: Brisbane

Re: Vertical Trajectory System (with some 3D printed parts)

Postby snrkl » Mon Sep 10, 2018 11:01 am

Any ideas where the solder balls came from?


Sent from my iPhone using Tapatalk Pro

drew
Southern fried goodness
Southern fried goodness
Posts: 947
Joined: Sat Aug 01, 2009 10:20 pm
Location: Perth

Re: Vertical Trajectory System (with some 3D printed parts)

Postby drew » Mon Sep 10, 2018 12:51 pm

I gotta say OTT the above root cause analysis of your electronics fault is absolutely amazing! Well done on finding the likely candidate for the fault, I'd still be scratching my head if it were me! Good luck on the next test.
Andrew Hamilton
AMRS 28 L3
AMRS Records Committee Chairman
Max Alt AGL - 23,908ft - K300 - Balls 22
Max V - 2,488 ft/s, ~Mach 2.2 - M2250 - THUNDA 2015

User avatar
OverTheTop
It's only money...
It's only money...
Posts: 2658
Joined: Sat Jul 21, 2007 2:20 pm
Location: Melbourne, Australia

Re: Vertical Trajectory System (with some 3D printed parts)

Postby OverTheTop » Mon Sep 10, 2018 2:01 pm

Solder balls are a manufacturing defect. Typically seen from reflow processing of a surface-mount PCA. Given the size of the balls I almost think they were from the sloppy hand-soldering of some of the components. It was basically suffering from what I call "Chinese Toy Syndrome" where a toy works for a while and then fails due to bad soldering because they are paid on a piece-part basis. Normally such workers are members of the Solder Conservation Society, but in this case they, unusually, were providing excess on most joints.

Normal surface mount solder balls are an artifact from the solder paste (mix of tiny solder balls and a flux) used for the reflow work. With correct process parameters all the tiny balls form the correct joints with no leftovers. If process is not sufficiently good small clumps of them form into a larger ball. They will quite often stay put, stuck in flux residue, if the board is not mechanically shocked.
TRA #13430
L3
"Everybody's simulation model is guilty until proven innocent" (Thomas H. Lawrence 1994)

User avatar
snrkl
Rocket Flyer
Rocket Flyer
Posts: 94
Joined: Wed May 10, 2017 2:07 pm
Location: Brisbane

Re: Vertical Trajectory System (with some 3D printed parts)

Postby snrkl » Mon Sep 10, 2018 6:29 pm

Well it’s a good thing we’re not using them for anything with any kind of force is applied to the units then...

#sarcarsmHasAPlaceInRocketScience #JustNotABigPlace


Sent from my iPhone using Tapatalk Pro

User avatar
OverTheTop
It's only money...
It's only money...
Posts: 2658
Joined: Sat Jul 21, 2007 2:20 pm
Location: Melbourne, Australia

Re: Vertical Trajectory System (with some 3D printed parts)

Postby OverTheTop » Mon Sep 10, 2018 6:34 pm

:D :D

I am pondering whether to split open the other electronic items and flight-proof them. In my spare time I think...
TRA #13430
L3
"Everybody's simulation model is guilty until proven innocent" (Thomas H. Lawrence 1994)

User avatar
OverTheTop
It's only money...
It's only money...
Posts: 2658
Joined: Sat Jul 21, 2007 2:20 pm
Location: Melbourne, Australia

It's Back!

Postby OverTheTop » Sun Sep 16, 2018 3:07 pm

Good News
After the maiden flight that saw the servos fried by a SBEC that passed 12V directly to them, things are looking better. Following a short period of convalescence and some R&R (that’s Remove and Refit :wink: ) the VTS is again functional. All four servos (I didn’t trust the remaining “good” one) and the offending SBEC were replaced.

The new SBEC was cut out of the heatshrink wrapping and inspected for solderballs. Sure enough, there were a couple on the surface of the PCB. The whole thing was given a bath in flux remover and inspection under 10x magnification on the Mantis. Once it looked clean and dry it was given a good coat of polyurethane conformal coating to insulate any exposed metal and pin down any solder ball I may have (unlikely) missed, and shrinkwrapped again. The heatshrink actually pushes the heatsink onto the back of the PCA for thermal transfer.
SBEC2resize.jpg

I also remove all those jumpers on the right of the picture to get rid of another possible source of unreliability. The voltage jumpering is soldered in place.

I have to say I am still nervous about killing another gaggle of servos, so I may actually put some crowbar protection on the output of the SBEC. This deliberately shorts the output if the voltage exceeds a trip point, but protects everything downstream from overvoltage.

I could also feed that servo SBEC with only a 2s battery (8.4V), not the 3s (12.6V) that is currently fitted. That might take a little electrical stress off it and perhaps not fry everything downstream if it does fail. I need to do some testing that I can still get decent current delivery with the reduced input voltage before I fly in that configuration.

I did have one of the axis gains set deliberately higher than the other so I could effectively get two different data points from the flight. That is where the system oscillation crept in I suspect. You can see it in the magnitude of the oscillations in the earlier graph, where one is much higher than the other two. I will adjust that gain down to be similar to the other one for the next flight.

I might have to try to get to a MARS launch in October for a second flight. With crops and bees at Serpentine, and a looming fire season I don’t think I will get to light one off there until next year. I just found a nice K815 in the cupboard that will make for a decent flight, right in the part of the flight envelope where I want to be testing currently, at Drouin.

I have also been pondering setting the roll fins deliberately off neutral for launch. That way if there is a non-neutral input from either or both of the other two axes then the rocket will continue to fly in the direction it was launched, albeit with a bit of a corkscrew motion.. Zero roll and input from one of the other axes will cause it to arc over. Looks like the obvious "right" answer was perhaps not the best, after a bit of thought.
TRA #13430
L3
"Everybody's simulation model is guilty until proven innocent" (Thomas H. Lawrence 1994)

User avatar
OverTheTop
It's only money...
It's only money...
Posts: 2658
Joined: Sat Jul 21, 2007 2:20 pm
Location: Melbourne, Australia

Preping for Second Flight

Postby OverTheTop » Sat Sep 29, 2018 4:56 pm

Preping for Second Flight

Not My Biggest Fan
So, servos were replaced after failing on the earlier flight due to the SBEC going short circuit.

When thinking about failure modes for that SBEC on the first flight I am fairly certain it was solder balls moving around. The other possibility is that the heatsink overheated when the servos were loaded up pegged to the mechanical stops. I noticed how hot the heatsink got during load testing on the bench of the replacement unit. In case the heat dissipation issue may have been a problem I have purchased a small (25x25x7mm) on eBay. This will spin up when the VTS is turned on and give a good forced-convection flow across the heatsink of the SBEC in its confines. That should keep the temperature of the heatsink down and mitigate any failure mode in that area.

A piece of G10 was mounted on the dividing head and a mount was machined out.
Millingresize.JPG

Milling2resize.JPG

BeforeFile.JPG


A bit of a cleanup after separation and some M3 tapped threads added
afterfile.JPG

FanAssembly.JPG


The plates in the VTS were marked out and slots for locating the fan assembly were cut with the dental drill.
DentalDrillresize.JPG

The completed assembly
Assembledresize.JPG


In the interest of lower temperatures when it is sitting on the pad I will also include three breather holes in airframe for that compartment. That should keep the temperature to not too much above ambient.

Gaining…
The response of the system on the first flight was surprisingly fast where it was correcting the tilt and was also oscillatory after that. Both of those indicate that the overall system gains are too high. The gains on the autopilot were dialed back, and since I am not looking for neck-snapping performance I also had some fins fitted with less mechanical travel. Both of those treatments should tame the overall system gains and give a smoother journey all around.

I will see if I can get this thing in the air during October. I have a K815 that should do nicely :)

Flight Controller
I am also starting to think about inserting one of the open-source flight controllers like the Naze32 for the next slightly more productionised version, and less of a rough test breadboard. They are available for less than $30 :shock: .
naze32-rev6-2.jpg
TRA #13430
L3
"Everybody's simulation model is guilty until proven innocent" (Thomas H. Lawrence 1994)

User avatar
SpaceManMat
Astronaut
Astronaut
Posts: 2095
Joined: Mon Dec 15, 2008 9:56 pm
Location: Brisbane

Re: Vertical Trajectory System (with some 3D printed parts)

Postby SpaceManMat » Sat Sep 29, 2018 10:04 pm

That’s a neat little flight computer and great that it’s open source.

With heat concerns, I’m reasonably ceartain that this has caused a number of failures over the years. You’ve heard the story “it was working on the pad, but then no events” always missing the part about baking in the sun for half an hour. With only small holes in the av bay means any heat that gets in there stays in there.
QRS: 124
AMRS: 32 L2 RSO
Highest Altitude: 13,647 feet
Fastest Flight: Mach 1.55
Largest Motor: CTI 1115J530 IM
Current Project: X Wing

User avatar
OverTheTop
It's only money...
It's only money...
Posts: 2658
Joined: Sat Jul 21, 2007 2:20 pm
Location: Melbourne, Australia

Another flight for the VTS

Postby OverTheTop » Sat Dec 08, 2018 5:36 pm

Flew this again last weekend on an L935, getting the speed up a little from last time. Gains were turned down a little from last flight, and the failed power supply was of course fixed. Flight was essentially nominal, and I have had a quick look at the downloaded data and it seems to have stabilised on roll, pitch and yaw axes during coast! Interestingly the vector it tracked was not vertical :? . So there seems to be some sort of algorithm in the autopilot for trimming the level very slowly. This adjusted the vertical vector in the time between turning it on and hitting the launch button. No power supply failures (I assume) this flight. I will investigate further in the coming weeks.

Interestingly I flew four cameras in the new shroud and one suffered a failure (stopped recording and scrambled SD card) that I think was thermally induced. I need to keep the sun off them before flight for next time. Camera shroud details here: viewtopic.php?f=32&t=6230

Stay tuned!
TRA #13430
L3
"Everybody's simulation model is guilty until proven innocent" (Thomas H. Lawrence 1994)


Return to “Higher, Faster, Further”

Who is online

Users browsing this forum: No registered users and 1 guest