I could use just a little guidance

The Rocketry Forum

Help Support The Rocketry Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
IIRC, you right-click in the parameter window (the one that allows you to select what gets graphed.) It's not the most intuitive interface.

Yep, that's how it works.

The graph with velocity added is attached (with velocity / 10 on the right axis). Looks like the flight was straight down to 200 ft/s.

Jim

Tilt w_velocity.png
 
Jim,

Have y'all considered using Full State Feedback(FSF) instead of PID? FSF allows you to place the poles of your system anywhere you want, so you can specify parameters such as settling time and overshoot, use those to calculate your desired natural frequency and damping ratio, and from those calculate your desired system poles. There are practical limits into how much force your canards can actually produce but it may be useful to try that method instead of guesstimating PID gains and looking at the response.

Elijah
 
Yep, that's how it works.

The graph with velocity added is attached (with velocity / 10 on the right axis). Looks like the flight was straight down to 200 ft/s.

Jim
Thanks for adding the velocity, it makes explaining the observed vehicle dynamics possible.

My interpretation is the minimum effective airspeed to the canard of this size and lever arm was 250 fps. That's the velocity when the weather cocking starts to be corrected, and the correction becomes more effective as the velocity increases to maximum. Then as the velocity drops, the corrective effects are reduced and the tilt angle increases and but stabilized until the velocity dropped to below 250 fps at which point the weather cocking started accelerated the tilting till apogee.

My conclusion it that the canards need to be larger (perhaps by increasing the aspect ratio to increase lift), and you need a longer lever arm to lower the minimum effective canard velocity. Once that is done, directional control will be attained a lower velocity and the initial tilt angle will be lower and stay lower for a longer fraction of the flight.

Bob

Bob
 
Jim,

Have y'all considered using Full State Feedback(FSF) instead of PID?
Elijah

I haven't considered that method because I wasn't aware of it. Thanks for pointing it out. A quick look at it suggests that it will take a little time for me to figure out what it is. In the interest of keeping things simple, I suspect that I won't be using it anytime soon. But I do appreciate you mentioning it.

Jim
 
Thanks for adding the velocity, it makes explaining the observed vehicle dynamics possible.

My interpretation is the minimum effective airspeed to the canard of this size and lever arm was 250 fps. That's the velocity when the weather cocking starts to be corrected, and the correction becomes more effective as the velocity increases to maximum. Then as the velocity drops, the corrective effects are reduced and the tilt angle increases and but stabilized until the velocity dropped to below 250 fps at which point the weather cocking started accelerated the tilting till apogee.

My conclusion it that the canards need to be larger (perhaps by increasing the aspect ratio to increase lift), and you need a longer lever arm to lower the minimum effective canard velocity. Once that is done, directional control will be attained a lower velocity and the initial tilt angle will be lower and stay lower for a longer fraction of the flight.

Bob

Bob

Good catch on the effectiveness of the canards for increasing velocity.

The planned next iteration of the canards is on the left in the pic below. They will be about 2.5x larger than the canards from the last flight (on the right). That size increase is required for stability of the three stage rocket (i.e., the canards could be larger but not smaller).

Jim

All Canards.JPG
 
Sorry, have not made any comments in awhile. Impressive results with tour last flight, and great that you added onboard video to be able to truly see how well the system worked.

Could see that it made a minor roll correction smoothly and did not overcontrol.

I second the idea of doing a mid-flight maneuver to test out the system effectiveness.

What I'd suggest is that 1 second before maximum velocity, for the controller to begin a say 45 degree roll, but 1 second later, at expected max velocity, to end the commanded 45 degree roll and force it back to the original roll position. That would allow you to see if it is over-controlling, as well as to get some good idea of how quick;y it does a roll maneuver at such a velocity, It could be that it make a 45 degree roll in 1/2 second and the rest of the 1/2 second is it settling out before it begins to resume the original roll orientation. Or it could be the roll control is not that rapid, so it may only say roll 30 degrees in 1 second, then would stop that roll (probably overshoot a bit) then you'd see how it returned back to the original position and how well it "settles down".

If it actually did do a 45 degree roll in a half second or less…. I'd be concerned it would have more roll control than would be needed. Since after all, you do not really need for the rocket to make a programmed 45 degree roll, you only need for the roll control to be able to keep the bird steady in roll. But a test maneuver like that would help give you some idea.

I am glad to see the larger canards. The ones you used before just seemed too small. Now, with the 2.5X larger canards, you also will be increasing your roll control authority by 2.5X. But if it turned out for example that you had just the right amount of roll control authority before, you portably will have too much by increasing it by 2.5X. So, if the roll control was
"juuuust right" before, you'd need something like 40% of deflection angle for roll control for the larger ones to have the same effective control authority in roll.

Back to doing a maneuver during a test flight, a pitch/yaw maneuver would also be good. I guess in the big picture, if you can only test for one thing next flight, it may be best to focus on the pitch/yaw to see the effects of the larger canards, and save any roll maneuver test for later. But definitely think you need to reduce the roll control to about 40% to avoid possible overcontrol issues. And perhaps a launch tip-off might impart a small roll to the model that will cause the roll control to get a bit of a test anyway, not as severe as a programmed one, but perhaps useful.

I do not recall, is the programming set to handle roll in the sense of simply nulling out roll to stop roll? Or is it trying to hold an actual compass heading? In the big picture you only need to make it to stop rolling, without caring where the roll orientation ends up, so that may be how it is set up. My above suggestion of a programmed 45 degree (or whatever) roll would depend on it being able to maintain a compass heading, which I now realize is not a necessary capability. If it is not trying to hold a compass heading, then a programmed roll could be to deflect the canards a few degrees CW or CCW for 1/2 second or so to cause a roll to begin, then go back to normal control mode and see how quickly it stops the roll and how much it may oscillate a bit before settling down.

- George Gassaway
 
I second the idea of doing a mid-flight maneuver to test out the system effectiveness.

What I'd suggest is that 1 second before maximum velocity, for the controller to begin a say 45 degree roll, but 1 second later, at expected max velocity, to end the commanded 45 degree roll and force it back to the original roll position.

I am glad to see the larger canards. The ones you used before just seemed too small. Now, with the 2.5X larger canards, you also will be increasing your roll control authority by 2.5X. But if it turned out for example that you had just the right amount of roll control authority before, you portably will have too much by increasing it by 2.5X.

I do not recall, is the programming set to handle roll in the sense of simply nulling out roll to stop roll? Or is it trying to hold an actual compass heading?

- George Gassaway

In the current version of the software, we can change the angle of "vertical" so that the rocket would tilt at some angle relative to vertical. Currently, we can't tell the rocket to go "north", but we can say go "left", and that orientation will stick even if there is roll. The go north version is being worked on, although this will probably be something that is controlled by the orientation of the rocket on the pad and not a magnetometer. The current plan is to try one more test flight with the larger forward canards tilting to 5 degrees near maximum speed and then back to vertical at a slower speed. The programming for that flight will reduce the roll response by a factor of 2 over what it was, but will leave the tilt response the same.

I have been a little more concerned about tilt for obvious reasons. However, it might be possible to command a roll by simply changing the roll setpoint from zero degrees per second to some positive value (in the same manner that the tilt target is changed from zero degrees to 5 degrees or whatever). I don't think we would be able to actually control the rate without a more complex control approach, but we could make it turn. Currently, we are not trying to hold a compass heading. Instead, the objective is to reduce whatever roll would otherwise be there.

Jim
 
Just a quick update ....

I have completed the larger canards. They are thicker (1/8"), so I beveled them using my router. Almost too easy! Made 6 of them in case a few break during testing.

I have mostly completed the lower stabilization section. This is the spool piece that will be installed between the 1st and 2nd stages. A key step here was to transfer the hole positions from the upper canard spool piece to the larger-diameter lower spool piece. There's probably a better way to do this, but I just traced the positions of the holes, "moved" them a little to account for the larger tube, crossed my fingers and drilled. Didn't turn out too bad. The stabilization equipment can be moved from one spool piece to the other without any adjustments.

The last pic just shows how the lower spool piece will be configured. The upper part of the 1st stage is on the left. For the test rocket, there is a larger diameter top piece glued to the airframe, as this is the size of the tube on the actual three-stager. The stabilization spool fits into the top of that and the two pieces are shear pinned (and maybe keyed) together. The chute for the stabilization spool will go in from the left side.

The rocket on the right side is the test rocket, which is taking the place of the second stage in the three-stager. There is a piece of airframe tube around the bottom of the motor. The stabilization spool will be shear pinned to this piece of tubing. In flight, a separation charge in the left side of the stabilization spool will drop the first stage. Then, there will be about 10 seconds of guidance with the spool piece attached to the bottom of the second stage, and then a separation charge will separate the stabilization spool as the 2nd stage lights. It could work.

The next test flight will be with the upper stabilization spool, the larger canards and the transition to/from a 5° tilt from vertical. That will be on a Gorilla L1150. In May, I'm hoping to test the lower spool location using an AMW M1850GG. The booster portion of the test rocket is in construction and should be done in a few weeks. The motor case will be present in the upper stage of the test rocket, but will not contain a reload. This flight will test the function of the canards in the lower position and the separation sequence.

Jim

DSCF1056.jpg

DSCF1040.jpg

DSCF1042.jpg

DSCF1045.jpg

DSCF1046.jpg

DSCF1047.jpg
 
Really nice work Jim.

Thanks Bob. This project is a lot of fun!

I just got the new guidance firmware installed and I've spent a few minutes playing around with it. I don't fully understand this yet, but it appears to me that the direction of the 5 degree bend will be established at power-up. When the change in the tilt target is requested (by the Raven in my case), the UDB will try to turn the rocket in that initially-established direction even if the rocket has rolled or is still rolling. Said differently, the target tilt direction is constant in the earth frame of reference (except for gyro drift). If it was constant in the rocket frame of reference, then that target point would move around the sky in a cone shape if the rocket was rolling - not so good.

The graph below shows how it works (the graph shows the signals to the servos with 3000 being neutral - 1.5 ms). There are three points where I changed the tilt target 5 degrees and then went back to vertical. For the first block, I powered up the unit and then turned the tilt on and off. Two opposing canards moved (in an attempt to move the rocket 5 degrees off vertical). Then, I rotated the unit 45 degrees. With that orientation, all four canards had to move to try and move the rocket in the direction set at power up. Then, I turned the unit 90 degrees and the other set of opposing canards were the ones that moved. So, the unit will try to keep the rocket pointed at a particular point in the sky, even if it's rolling.

None of this capability will be used in the Balls flight - it's just for the purpose of generating a known disturbance to look at the dynamic response. But it is a good example of the capability of this board. I am really looking forward to flying this.

Jim

Servo outputs with tilt firmware.jpg
 
Seeing those four small control fins behind the three much larger fixed fins makes me think of NACA 1307 which discusses fin to fin interference effects.

Normally you would rotate the fins so as to maximize the radial separation but with one set of four and another set of three you cannot get uniform spacing.

Better would be to use three control fins although that would require a little extra effort from the firmware.
 
Seeing those four small control fins behind the three much larger fixed fins makes me think of NACA 1307 which discusses fin to fin interference effects.

Normally you would rotate the fins so as to maximize the radial separation but with one set of four and another set of three you cannot get uniform spacing.

Better would be to use three control fins although that would require a little extra effort from the firmware.

Actually, the firmware changes would be very simple (I'm told). Bill will add this if I want it, and we could use 4+3=7 of the available 8 outputs to allow either 4 or 3 canard operation. The problems going to 3 canards are more on my end. In addition to requiring a completely new stabilization section, this would affect many aspects of how the rocket and the pad interact that aren't particularly easy to change (they weren't all that easy to get to how they are now). I've been giving this some thought though.

Jim
 
I'm reading this tread with great interest, I notice the worry about G force, with active stabilisation we can actually lower the G force with lower trust weight ratio. We have to make the rocket go slower. …I continue to read……
 
I did the fourth test flight over the weekend. This flight was with the forward canards (probably the last of these), the larger canards, and the programmed change in the flight path. The flight path change was a 5° tilt off vertical at 4 seconds and then a return to vertical at 6 seconds. I think I see those changes, but some additional data analysis will be needed.

One thing that happened on this flight is that we had more lateral acceleration. That caused the accelerometers to turn on and off during the flight, starting at 5.5 seconds, which we didn't want to happen. We need to modify the programming to avoid this. This is probably what caused the roll instability starting about 6 seconds into the flight. Prior to that, though, the rocket is traveling pretty close to vertical, even with a pretty stiff wind. Lots of data and information to extract here, and here's the video.

https://youtu.be/I4UA2I5psIw

I'll post more data over the next week or so.

I also did a two-stage flight at this launch. Nothing to do with guidance, just a nice flight. K630 to K160, 14K feet. Here's the video for that flight.

https://youtu.be/xngNu9mhnNE

Jim
 
For those interested, here's the flight data. For orientation, canard 1 is on the left and canard 2 on the right (in the video). The x axis is to the left through canard 1. The z axis is to the right through canard 2.

I haven't analyzed this data too much yet, so I don't have a lot of comments on it. The servo plot shows the canards were working hard from the start to deal with weathercocking from the wind.

The offsets and tilt plot shows that the net tilt (in black) decreased as the speed increased. You can see the approximate 5 degree tilt when it turned on (turning the rocket closer to vertical).

The acceleration plot shows significant lateral acceleration associated with the turn.

The black line on the gyro plot shows where the accelerometers turned on (by mistake). We think this is what caused the roll instability (the accelerometers giving bad information as to where "down" was), but we are not sure yet.

The two gps plots show that the flight trajectory was almost perfectly vertical through about 8 seconds of flight.

There is a lot of data to extract from this, so feel free to share your opinions is you spot something.

Jim

Servo outputs.jpg

Vert offsets & Tilt.jpg

Acceleration.jpg

Gyros.jpg

GPS Elev.jpg

GPS Plan.jpg
 
Last edited:
Jim can you make 4 graphs that show what is driving each of the servos? IE put one servo output on the graph then add the tilt axis for that servo, then add the roll rate. Then add tilt and roll together (or subtract) to show the total desired correction. If might also be handy to put the accell data in there, this shows how much force is being applied to the rocket to correct the tilt (zero tilt should result in zero accell).
 
Jim can you make 4 graphs that show what is driving each of the servos? IE put one servo output on the graph then add the tilt axis for that servo, then add the roll rate. Then add tilt and roll together (or subtract) to show the total desired correction. If might also be handy to put the accell data in there, this shows how much force is being applied to the rocket to correct the tilt (zero tilt should result in zero accell).

I like that idea. Might make things easier to see. In the short term, z axis tilt drives servos 1 and 3, and x axis tilt drives 2 and 4. Roll affects all four, so you should be able to spot that in the graph you described.

Jim
 
Jim

Comments.

I think you pegged the vertical accelerometer during boost (it appears to be set at 2^14 = -16384) and it appear that -1 G is ~ 4096, so it a 4 G accelerometer?

Have you done any calibrations to the accelerometers and gyros to convert the digital counts into engineering units? I think it might be useful to simply run the unit on the bench and record the baseline output on all channels to measure short term drift on the gyros and accelerometers with the package oriented in the +X and - X. +Y and -Y and +Z and -Z axis. This would provide a basic calibration for the gyros and accelerometer so the plots could be calibrated in engineering units. Temperature should be noted for the calibrations.

This would make interpretation really easy.

Bob
 
Jim

Comments.

I think you pegged the vertical accelerometer during boost (it appears to be set at 2^14 = -16384) and it appear that -1 G is ~ 4096, so it a 4 G accelerometer?

Have you done any calibrations to the accelerometers and gyros to convert the digital counts into engineering units? I think it might be useful to simply run the unit on the bench and record the baseline output on all channels to measure short term drift on the gyros and accelerometers with the package oriented in the +X and - X. +Y and -Y and +Z and -Z axis. This would provide a basic calibration for the gyros and accelerometer so the plots could be calibrated in engineering units. Temperature should be noted for the calibrations.

This would make interpretation really easy.

Bob

Bob, time is really limited at the moment, so I didn't have time to label the charts. Yes, it's a 4G accelerometer with gravity being about -4000. For the x and z offsets, these can be divided by 286 for the angle in degrees (or just use the calculated net tilt). The servo output is 2000 to 4000 (for the 1 to 2 ms pulse range). However, the servo gain factors into this. I don't have a number in hand that would give you a conversion between servo signal and servo angle. My best off-hand guess is that the full range on the output would be a 45 degree canard change. The gyro values are in degrees per second.

Much testing of drift has been done (with good results). I've just not presented that data here. Temperature is a variable that we are not dealing with yet.

Jim
 
Jim can you make 4 graphs that show what is driving each of the servos? IE put one servo output on the graph then add the tilt axis for that servo, then add the roll rate. Then add tilt and roll together (or subtract) to show the total desired correction. If might also be handy to put the accell data in there, this shows how much force is being applied to the rocket to correct the tilt (zero tilt should result in zero accell).

I came up with the attached plots. They show the requested correction (in black) and the corresponding servo movements (blue and red). The way I have it plotted, they should all agree unless there is roll (in gold). Then, the servo movements should diverge, which they do. Although it's hard to link cause and effect with the video, everything is moving as it should.

Bill has been able to prove that the accelerometers coming on (which we didn't want) is what caused the flight to go bad (to arc over prematurely). I wish I could fully appreciate his explanation, but in a nutshell, he can compare the pitch and yaw rates that were calculated during the flight (from the gyros and accelerometers when they were on) versus those calculated from the gyros only (which is what we wanted). The error is substantial, partly due to the lateral accelerations during the flight (including the ones we added).

So, even though this flight wasn't as aesthetic as it might have been, I think it was very successful. The new canards seemed to work pretty well, although the data suggest that both the tilt and roll gain could still be increased for this particular flight and rocket configuration.

The next flight will be with stabilization at the rear of the test rocket.

Jim

Controls 1.jpg

Controls 2.jpg
 
It looks like large corrections are inducing a roll. I think this is because of uneven loading on the 2 fins involved. Not sure if there is anything you can do to negate this other than increase the gain.
 
It looks like large corrections are inducing a roll. I think this is because of uneven loading on the 2 fins involved. Not sure if there is anything you can do to negate this other than increase the gain.

I have a feeling that this stabilization system is best used with four-fin rockets. If there is wind, then you are intentionally flying with a non-zero angle of attack. With three fins, that will probably induce some roll. The first portion of the last flight looked good, though, before the gyro problem.

I've made progress on the test rocket in the two-stage configuration. The new booster will be electronic apogee deploy, and it's complete except for rail guides, shear pins and vents. The stabilization section, which is longer than the one used for the upper canards, is mostly complete too. There are a couple of sets of shear pins on that, but the main challenge is to figure out how to run the wires for the various separation and recovery charges.

I'm hoping to launch this in a couple of weeks. Ideally, it will launch in the two-stage configuration (pic 1) with the canards providing mainly roll control (I need to arrange for the CG to be just ahead of the canards at launch. Then, the booster will drop off and the sustainer/stabilization section (pic 2) will coast and go vertical. After about 8 seconds of coast, the stabilization section will separate and recover on its own (electronic apogee also). There will be no motor in the sustainer, so it will just coast and then recover normally.

Jim

IMG_0094.jpg

IMG_0097.jpg
 
Jim, why not keep the guidance with the rocket?

I wonder if you keep it with the top stage of that might not get you better altitude by keeping it vertical? Not sure if that versus addition weight/drag wins though.
 
Jim, why not keep the guidance with the rocket?

I wonder if you keep it with the top stage of that might not get you better altitude by keeping it vertical? Not sure if that versus addition weight/drag wins though.

It would probably be a good thing, but isn't practical with my existing rocket. Obviously, the stabilization section as used behind the second stage has to separate before the motor ignites. It would be interesting to have a second, or perhaps only, system near the top of the 3rd stage, but with the current design, it would be difficult to transmit roll forces through the rocket. I also think that further development of the control system would be required, such as changing the gain as a function of velocity (or time). Just a project for another day. My current goal is to be somewhat straighter than I would otherwise be without stabilization.

It won't be long before this tool is available to a wider group. Then, there can be a wide range of goals.

Jim
 
Jim,
Get the MARSA54 Wireless system to control your staging....no wires to run!

BTW: this is a candidate for "best thread of the year" IMHO

Keep up the good work.
FredA
 
Folks, I have stated before that one of my goals for this project is to be able to make the technology available to a wider group. To start this process, I am seeking a handful of rocketeers (maybe 3-5) that are interested in attempting a stabilization project. My goal is to take the experience from this group to develope a guidance document (sorry) that addresses mechanical and electrical design, flight issues, and safety. This would be the basis for a potential wider distribution of the technology later on.

If you are interested in doing this, please contact or PM me. If there are more than 3-5 people interested, I'll have to pick and choose, as I won't be able to support more projects than that. Those participating will receive the equipment at cost (I'm not a vendor and don't plan to become one), some modest documentation, and kibitzing from me. One requirement for participating will be to share your equipment design, experience and flight data. Projects need not be high altitude rockets at Balls.

Jim
 
I wasn't able to do my test flight with the rear canards (see post #189) due to a waiver issue. Unfortunately, it looks like it will be at least 3 weeks before I get another chance.

Between floods, though, I did have a chance to document my experience using the UDB5. It's a reference for me and for the guys that will be doing test flights. If you're interested in the details of how the device is used, take a peek.

Jim

View attachment UDB5 Intro.pdf
 
So, it turned out that 5 flyers (or groups) want to do a stabilization project with the UDB5. I hope they'll report progress as they get further along. One flyer has completed his design and is ready for a test flight. His approach for mounting the servos and the canards is quite different than mine (see the pics), so we'll see how that works out.

I'm still on hold for my 5th test flight. Maybe this coming weekend or the next.

Jim

IMG_0324.jpg

IMG_0362.jpg
 
I cant believe I missed this thread until now.

Definitely got me thinking of starting a new rocket project.
 
Back
Top