I could use just a little guidance

[JimJarvis50]

Hi Jim,
looks like you are getting your system pretty well dialed in.

br/

Tony

The recent changes represent about an order-of-magnitude improvement in the performance and capability of the system. The testing and flying of the system has been very interesting and will continue. Next, we'll try loops. OK, we won't try loops, but there are some interesting profiles we could try. One that interests me is to fly the rocket at some relatively high angle - say, 20°, and then bring the rocket back to vertical near the end of the flight to reduce the deployment velocity. We can do all of this now - just need to use a Raven or a timer to ground a pin at the appropriate time.

Jim

 

[heada]

I was curious if you could use a static fin can and negate the control reversal in software. If you knew the speed range when it happens and can accurately calculate speed, have the canards give reverse correction so that the control reversal is reversed. I don't know how hard that'd be and if it'd be easier than building the spin can.

 

[Hermesf]

Jim - excellent work...how exciting to have achieved yet another milestone in your pursuit! Gotta admire Bill Premerlani's ingenuity and creativity, and both of your persistence...Frank H.

 

[JimJarvis50]

Jim - excellent work...how exciting to have achieved yet another milestone in your pursuit! Gotta admire Bill Premerlani's ingenuity and creativity, and both of your persistence...Frank H.

He's a magician and I'm just along for the ride. But, we're both having immense fun, and we both agree that we have gotten much further than we ever envisioned when you got us together. My humble thanks for doing that by the way.

Jim

 

[Hermesf]

I am the one humbled, by your work...it has been so much fun to watch the ride...Frank

 

[UhClem]

The cause of tilt accumulation was error resulting from the use of a small angle approximation when the roll rate was high. Adding the third term of the Taylor's expansion fixed the problem.

Running the tilt calculation at a rate higher than the servo update rate would do the same thing. Using the small angle approximation (for sine) lets this run very fast. If speed isn't really a concern, then just whip a full sine on it. Hard to say which is better since the code is still a black box.

I still think that what you want is something that maintains the initial azimuth and elevation. This lets you adjust the trajectory to compensate for wind simply by adjusting the pad.

 

[JimJarvis50]

Guys, I wouldn't normally do this, but I reposted the video to add some new information on the flight. The new link is below. The main thing I wanted to add was information on the path of the rocket relative to the pre-set orientation. This info is summarized in the attached picture.

The information on the wind direction in the picture was obtained from the gps data during the descent of the rocket. Based on the wind direction, you can see how the rocket was blown sideways by the crosswind, particularly at the start and end of the flight where the velocity was slow. In the middle of the flight, though, the path was aligned with the pre-set orientation of the rocket. To me, this is the most interesting data from the flight.

Jim

 

[JimJarvis50]

Running the tilt calculation at a rate higher than the servo update rate would do the same thing. Using the small angle approximation (for sine) lets this run very fast. If speed isn't really a concern, then just whip a full sine on it. Hard to say which is better since the code is still a black box.

I still think that what you want is something that maintains the initial azimuth and elevation. This lets you adjust the trajectory to compensate for wind simply by adjusting the pad.

Dave, Bill Premerlani passed along some information on the nature of our roll error.

"The second term in the Taylor’s expansion prevents a cross axis coupling between roll and tilt estimation. Think of it like a gyroscopic effect. At high roll rates, without this term, the roll gyro signal cross couples into the tilt estimation in a nonlinear way that makes the estimate of the tilt grow even when it is really constant. The third term and the gyro calibration contribute to roll angle estimation accuracy. We need an accurate estimation of the roll angle, because we need it to transform earth frame aiming into body frame control."

What we have is an approach that maintains an azimuth relative to a point on the rocket. So yes, you would adjust the azimuth by adjusting the pad. We're flying in the direction of the rail buttons. The elevation is pre-set and not related to the launch position.

Jim

 

[JimJarvis50]

I had a chance to fly the stabilized rocket again this month. L-1000 to about 5,600 feet. The flight profile was to take off vertically, and then fly into the wind at an angle of 15°. Then, at 8 seconds, turn vertical again. The idea of this is to get recovery closer to the pad by flying upwind for a while, but without suffering the high velocity deployment that you would otherwise get if you flew into the wind.

The rocket flew in the correct direction (determined by the position of the pad). However, it took longer to get to the 15° tilt than I expected, and it was settling in at about 17°. At 8 seconds, it tried to go vertical, but only got back to about 10° before it slowed down too much to turn. I know the system, as operated, turns the rocket relatively slowly, but for some reason, I thought turning to a 15° tilt would occur faster. Nope.

For the purpose of getting upwind a ways to then recover back to the pad, it probably makes more sense to just launch it into the wind and let it hold the angle, but for the time being, it's more fun to see it turn in the intended direction and then get to the preset angle. I'll probably try this again, but with a bit more control authority. Tilt and gps pics are attached.

The heading hold function worked again as planned with no net roll through the flight. Maybe I'll post the on-board video, but it isn't all that exciting.

Jim

 

[Hermesf]

Geesus Jim, sit back and just look again at what you have described...hard to believe you have come so far with this all - kudos to you my friend, excellent work...

 

[ksaves2]

Well, How far away from the pad did it end up landing? That is a really cool graph. I'm one who usually ends up pointing the rocket a couple of degrees or so downwind (depending on sims) to try and have a curving weather cocking flight that gets to a minimum energy at apogee for deployment. Doing active control like that is really neat to achieve both a shorter recovery distance and a low energy apogee deployment. Kurt Savegnago

 

[JimJarvis50]

Well, How far away from the pad did it end up landing? That is a really cool graph. I'm one who usually ends up pointing the rocket a couple of degrees or so downwind (depending on sims) to try and have a curving weather cocking flight that gets to a minimum energy at apogee for deployment. Doing active control like that is really neat to achieve both a shorter recovery distance and a low energy apogee deployment. Kurt Savegnago

The gps groundpath is attached. The plan was to try to land it back on the pad. Since it didn't reach the 15° as quickly as planned, I "overshot" the landing a bit. I also launched into the wind direction based on the weather models. I could tell that the wind was actually a bit more from the south then the southeast, but I stuck with the plan becuase if the wind was as predicted above the ground, then an adjustment to the south could have caused the rocket to recover into the flight line.

I'd still like to have the rocket go vertical, but it turns out that if you're holding an orientation like 15° at apogee, the deployment is still relatively gentle. In the video from this flight, it almost looks like the rocket backslides. It's not quite like the vertical flights (where it really does backslide), but it is gentle.

Jim

 

[ksaves2]

I’ve only witnessed one tail slide with my ASP WAC Corporal on an H128. Had the scale yellow paint job. With the small club launches we used to have, we put all the pads “way out there” to accommodate up to an M motor pertaining to the safety rules. My WAC went up and because it’s heavy with paint and a MAD unit, it usually stays within visual range. I saw on one flight the neatest tail slide and after falling backwards a few feet the MAD deployed the apogee charge. Actually having some distance from the pads allows one’s eyes to have the best chance of seeing most of a rocket‘s flight. I had a chute release on the reefed chute so that made for a super low energy deployment. The main unfurled at 800 feet with the JLCR. I like to do the highest main deployment a venue can tolerate as it allows more time to get a visual on the now slowed descending rocket. If a high flyer goes out of visual range for the entirety of the flight, the Rf tracker modality needs to be depended upon entirely. Using electronics to control the flight mode is very interesting and I caution any nay-sayers who state this constitutes “illegal” flying I rapidly state it’s not true point A to point B flight. There are no rule issues here. Broken down simply it’s control of the flight path on the “up side” for stable flight. Descent is still random as one is trying to get it to land as close to one’s position for easy recovery and is at the whim of the wind gods still.

Although I will state that if you entered a ”spot landing” competition, a contest director might balk with the electronic ”assist” installed. (n.b. A little attempt at jocularity here) Kurt Savegnago

 

[JimJarvis50]

I’ve only witnessed one tail slide with my ASP WAC Corporal on an H128. Had the scale yellow paint job. With the small club launches we used to have, we put all the pads “way out there” to accommodate up to an M motor pertaining to the safety rules. My WAC went up and because it’s heavy with paint and a MAD unit, it usually stays within visual range. I saw on one flight the neatest tail slide and after falling backwards a few feet the MAD deployed the apogee charge. Actually having some distance from the pads allows one’s eyes to have the best chance of seeing most of a rocket‘s flight. I had a chute release on the reefed chute so that made for a super low energy deployment. The main unfurled at 800 feet with the JLCR. I like to do the highest main deployment a venue can tolerate as it allows more time to get a visual on the now slowed descending rocket. If a high flyer goes out of visual range for the entirety of the flight, the Rf tracker modality needs to be depended upon entirely. Using electronics to control the flight mode is very interesting and I caution any nay-sayers who state this constitutes “illegal” flying I rapidly state it’s not true point A to point B flight. There are no rule issues here. Broken down simply it’s control of the flight path on the “up side” for stable flight. Descent is still random as one is trying to get it to land as close to one’s position for easy recovery and is at the whim of the wind gods still.

Although I will state that if you entered a ”spot landing” competition, a contest director might balk with the electronic ”assist” installed. (n.b. A little attempt at jocularity here) Kurt Savegnago

The system only attempts to control the orientation of the rocket. It is still subject to the aerodynamic ability of the rocket to hold that orientation, and other forces such as gravity and wind will move it away from where it is pointing. It is nice to have a reasonable idea of where the rocket is going to land. I don't like long recovery hikes.

I used to have a video showing three flights where the rocket actually back-slid on a flight. Can't find it for some reason.

Jim

 

[JimJarvis50]

I put together some of the flight video to document the flight. It didn't turn out quite as well as I had hoped, but I learned a few things for the next try.

Jim

 

[Dipstick]

I put together some of the flight video to document the flight. It didn't turn out quite as well as I had hoped, but I learned a few things for the next try.

Jim

I love this thread, I love each test flight you do, and so on...

I think Jim is one of the barnstormers of future amateur rocketry!

 

[JimJarvis50]

I love this thread, I love each test flight you do, and so on...

Bruce,

I'm glad you enjoy the flights! I do too. Good news, I did another one today. This one turned much faster off the pad and got more vertical at the end. Great day with Tripoli North Texas to fly rockets!

Jim

 

[Nytrunner]

I did another one today.

Did you adjust the settings or do a repeat?

 

[JimJarvis50]

Did you adjust the settings or do a repeat?

There were several changes. First, it was a larger motor (L1720 versus L1000), so about 7500 feet versus 5500 feet, and more early speed. I also reduced the static stability a little (about 1.1 with the larger motor but otherwise no changes), and I increased the control gain by about 20%. Those changes made things much more responsive. However, the test data, which I'll post at some point, suggests that the settings used for this flight are about the limit of the capability of the system as currently configured.

Edit - I also kept the turning point to vertical at 8 seconds. That should have resulted in a speed of about 500 ft/s going into the vertical turn versus about 390 ft/s with the smaller motor. That helped it get closer to vertical.

Jim

 

[JimJarvis50]

As it turns out, I can actually add a little clip of the takeoff here instead of just a picture. The turn was much faster, but you can see a little instability at maximum velocity (about 900 ft/s for this flight).

Jim

 

[cerving]

Wow, how did you get that video? The drone must have been at some serious altitude...

 

[JimJarvis50]

Wow, how did you get that video? The drone must have been at some serious altitude...

I'm pretty sure that's just an illusion from the fisheye lens.

The purpose of the flight profile I'm using is to land the rocket close to the pad, or at least in a location that's not bad. So, if you know the wind as a function of altitude, you can guestimate where the rocket will land. The attached pic shows what I planned to happen versus what actually happened. As shown, the rocket landed further to the east than I expected. "Windy" only has higher altitude wind velocities at altitudes of 5000, 6400 and 10,000 feet. I knew there was a strong northwest flow at 6400 and a strong westerly flow at 10K, but the transitions of those flows is not known. Also, in this case, there was a northwest flow at lower altitudes that Windy didn't pick up on. So, not quite a spot landing, but it is better than random.

Jim

Edit - I have adjusted the planned path to account for drift during ascent. It appears that the actual apogee was a bit further out than the planned apogee location. I think this is due to the rocket turning faster than I expected. In addition, the rocket had some oscillation and the result of that was that the net tilt was above the planned 15°. So, the rocket flew a little further from the pad before going vertical.

 

[TimothyG]

His rocket drew a better head in profile than most of the art students i know.

 

[JimJarvis50]

His rocket drew a better head in profile than most of the art students i know.

Thank you! The wind wasn't right to draw the outline of Texas, so I settled for the Prefect's head.

Here's some additional data from the last two flights, which shows tilt versus time. Both flights were set to turn to 15° after launch and then go vertical at about 8.5 seconds. The first flight turned relatively slowly to 15°. In contrast, the second got to 15° more quickly, got closer to vertical, and also showed some oscillations in tilt through the flight. The physical configuration of the rocket was the same in both flights, but the flights differed as follows:

- Stability. The first flight launched at 1.8 calibers versus 1.0 in the second, acheived via the use of a heavier motor.
- Control action - Canard angle maximum of 6.25° at 7.5° tilt versus 7.5° at 7.5° tilt.
- Higher maximum speed - 670 ft/s versus 850 ft/s.

It's interesting that the oscillation took place even though the canards were not reaching the maximum allowable 7.5° angle.

In view of the flight results from the second flight, we are adding a derivative term to the control equation for yaw/pitch. Then, we'll try a flight both without and then with the rate term included. I'm going to set this up with the smaller of the two motors (L-1000) with a vertical launch and then a 10° turn after 3 seconds of flight (just after motor burnout). The initial stability will be 1.35, acheived by taking 1/4" off of the fin span. We'll determine the gain for the rate term from the flight data.

Jim

 

[Alan15578]

That is all great stuff, but I would also like to see the angle of attack. lateral g loads or acceleration. and flight path angle. Tying to change the flight path angle 15 degrees at high speed is impressive.

 

[JimJarvis50]

That is all great stuff, but I would also like to see the angle of attack. lateral g loads or acceleration. and flight path angle. Tying to change the flight path angle 15 degrees at high speed is impressive.

Always good to look at more flight data. I don't have a way of calculating angle of attack (I don't think so anyway), but the changes in the tilt angle might give an idea of that.

The gps trajectory for both flights is shown in the graphs (the black dots). In the first flight, the trajectory is generall lower than the tilt because the rocket was flying into the wind. In the second flight, there was less wind early, but later in the flight, there was a strong cross wind. So, the trajectory becomes higher than the tilt.

The acceleration for the first flight are shown in the third graph. A value of 4000 is approximately equal to 1 G. The rocket flies about mid-way between the X and Y plains, so if the rocket is in its original roll configuration, changes in yaw/pitch should have equal X and Y values but with opposite signs. The roll angle is shown in black (right axis) and there was a roll angle (around 45°) in the initial turn but the roll angle was close to zero for the second turn. Looks like the initial turn was a litle under 2 G's and the turn to vertical was a little under 1 G. The second flight is too noisy to see much, but the accelerations were somewhat higher due to the higher speeds where the turns happened.

Jim

 

[JimJarvis50]

So, the thing we're working on now is to dampen the oscillations that occur when we make the rocket less overstable. The plan is to add a rate term to the currently proportional-only control of yaw/pitch. Yesterday, I did a baseline flight with no damping, and the data from that flight plus a previous flight will be used to determine the gain on the damping term. Then, we'll repeat yesterday's flight with damping included.

The program for the flight yesterday was to control to vertical for 3 seconds and then turn to 10° for the time between 3 and 8.5 seconds, and then turn back to vertical at 8.5 seconds. In the flght (see the pic), the rocket weathercocked at little at launch and was working back towards vertical at 3 seconds. Then, the change at 3 seconds caused the oscillations that we wanted to see. The rocket didn't get back to vertical after 8.5 seconds, but I didn't expect it to. I think we got the baseline test we wanted and the next test will be with damping.

I posted a short clip of the on-board video here:

Jim

 

[JimJarvis50]

So, just a quick update. The derivative control is programmed and ready to go. It's very simple. The proportional part of the term moves the canards to 7.5° if there is 7.5° of tilt. The canards stay at 7.5° max for tilts above 7.5°, and 7.5° of tilt corresponds to the saturation value of 250 us to the servos. The derivative term just opposes the change in the tilt. We think a gain of 50°/s would result in critical damping. So, as an example, if the tilt was 3 degrees and the rocket was moving at 20°/s towards 0° of tilt, the proportional correction would be 3 / 7.5 x 250 = 100 us in the direction of zero tilt. The derivative term would be 20 / 50 x 250 = 100 us in the direction away from zero tilt. So, for this example, the canards would be at 3° without the derivative term and 0° with that term. We will likely start with a lower derivative gain and work up.

Also, I was looking at more of the data from the flight. I previously posted tilt data from the EasyMega. However, the corresponding data from the UDB5 (which is actually doing the controlling) is slightly different. The data for the EasyMega and the UDB5 are posted below. What is interesting in the UDB5 data is that the rocket oscillates before it reaches 10° of tilt. At that point, the canards should still have been trying to get the rocket to the 10° set point. So, why did the rocket turn away from the 10° set point? I have concluded that what is happening is that the rocket flexes when requested to make a sharp turn to 10°. At 7.5°, the canards start to back off, and the rocket un-flexes, which shows up on the tilt graph as an oscillation that isn't real. This isn't the sort of response that is suitable for derivative control.

The "Test Rocket" was originally designed as a two-stage platform just to test the concept of controlling to vertical from below the sustainer. It was an inexpensive rocket, relatively speaking, made from thin-wall Wildman fiberglass tubing. It was supposed to fly a few times in preparation for Balls flights, and that was it. As it turns out, it has flow many more times than I ever expected under increasingly stressful conditions, and I added a spin can to the sustainer. So, for this on-going testing, I need to strengthen the upper air frame a little to elimimate this flexing. In keeping with the blue collar character of this rocket, I'm going to experiment with a Soller carbon sleeve with heat shrink tubing. I've never tried that, and never expected to try it, but it seems like the perfect application under the circumstances. I'll report back....

Jim

 

[JimJarvis50]

I did another flight of the stabilization system last weekend. The main objective of the flight was to test the derivative terms that we added to the yaw/pitch control scheme. Basically, the proportional term sets th canard angle in proportion to the deviation between the actual tilt and the set point tilt (although this is done separately for yaw and pitch). If the rocket tilt is moving towards the set point tilt, the derivative term (the rate of change of the tilt) moderates the canard angle to reduce the rate of change in tilt. Similarly, if the rocket is moving away from the set point, the canard angle is increased by the derivative term. The idea of this was to reduce oscillations around the set point.

Based on previous flight results, one thing we think we've learned is that air frame flex contributes to the oscillations we see during the flight. We don't want to try to control that, so I tried to minimize this by covering the upper air frame with some carbon. I used a carbon sock, and I can guarantee it is the first and last sock I will ever use. Nuff said on that.

The current flight was done on an L1050 Blue motor, The tilt program was to fly vertically for 4 seconds, tilt to 10°, and then go vertical again at 10 seconds. The "baseline" flight from last month was on an L1000, and the tilt graph from that flight was in Post 747. The current flight, with the L1050, had quite a bit more velocity when the tilt program engaged at 4 seconds. The tilt graph is attached, and there is less oscillation. Unfortunately, it's not possible to say whether to attribute this to less air frame flex (from the carbon) or the derivative term. The tilt graph shows that some air frame flex still occurred, which is indicated by an oscillation occuring before the tilt reached the 10° set point.

One other interesting thing that happened during the flight is that the rocket appeared to be unstable for about the first second of flight (and then went vertical once the velocity was high enough for the canards to kick in). The rocket is stable by 1.1 to 1.5 calibers, on paper, but is apparently not stable in actual flight. I haven't looked into this in any detail yet, but I suspect there is an issue associated with modeling the canards such that the normal rules for establishing the stability margin don't quite apply. Thoughts on that? If you look back in this thread to prior flights, you can see this has likely happened on other flights. Good thing I have a stabilization system on board to "rescue" the rocket! A short video snip showing this instability is linked and other pics and info from the flight are attached.

I should add that the "heading hold" feature added to the control logic resulted in no net roll through this flight, even with all of the events that occurred.

Jim