I could use just a little guidance

[TonyL]

Jim,
what do you think about the apparent EasyMega family [for lack of a better label] issue on the other BALLS 2 stage flight?

 

[mikec]

How do you pick the minimum altitude for lockout?

I set it to 80% of the expected altitude at the staging time based on simulations and any programmed delay. Be aware that the reported time of burnout in simulations is usually the nominal time of the motor reaching low thrust, while altimeters sense burnout with deceleration, which usually happens earlier depending on drag and velocity.

The percentage is a balancing act between actual flight angle and all the slop in the sims, etc. If you needed tighter tilt control you would have to go to a tilt sensing solution. Without naming names, I remain unconvinced that some of the systems available really do this reliably under all conditions.

 

[BABAR]

I set it to 80% of the expected altitude at the staging time based on simulations and any programmed delay. Be aware that the reported time of burnout in simulations is usually the nominal time of the motor reaching low thrust, while altimeters sense burnout with deceleration, which usually happens earlier depending on drag and velocity.

The percentage is a balancing act between actual flight angle and all the slop in the sims, etc. If you needed tighter tilt control you would have to go to a tilt sensing solution. Without naming names, I remain unconvinced that some of the systems available really do this reliably under all conditions.

I know I ask a lot of silly questions, but assuming tilt control and velocity locks, do you pick 80% for SAFETY reasons? At first glance, it seems like for safety considerations the MINIMUM altitude (again assuming tilt and velocity are nominal for safe sustainer flight ) would be that at which the booster has time to separate and safely deploy its recovery device. NON-safety reasons for selecting something higher may be on the order of, “Hey, if this doesn’t reach at least 80% predicted, I don’t want to ‘waste’ igniting my upper stage motor on a suboptimal flight. I’d rather just recovery the sustainer unfired and launch it again” Perfectly reasonable. I guess the question is, if it stages at 1,000 feet instead of 20,000 feet, but all recovery devices deploy correctly and all parts are recovered intact, is that a “failure?” That would be in the eye of the builder.

I guess if you DON’T have tilt and velocity locks maybe you can assume that if you hit at least 80% of your predicted staging altitude the rocket is probably operating nominally and it is safe to light the sustainer. Or, since you are out in the middle of nowhere, and therefore the greatest concentration of catastrophic targets (flyers and spectators and vehicles) is AT the launch site, if for some reason the system screws up and sustainer fires pointy end down, it is less likely to unfortuitously wend it’s way back to the launch site to hit anybody if it ignites 20,000 ft AGL from the launch site than 1000 ft above it.

 

[mikec]

I know I ask a lot of silly questions, but assuming tilt control and velocity locks, do you pick 80% for SAFETY reasons?

I don't have tilt control on the altimeters I use and I don't use velocity gates.

Tilt control is ultimately a safety issue so the second stage doesn't go outside the waiver cylinder. For high flights that requires a pretty tight tilt limit. I haven't tried to go above 75K feet and at BALLS you don't really need tilt control at that altitude since it is physically impossible to land outside the 7-mile radius at any sane tilt, say up to 45 degrees, and you can get that with an altitude gate. At some point, certainly with Jim's flights, you need real tilt control just to stay within the 15-mile radius.

Of course, just because you set tilt control on an altimeter that provides it doesn't mean you are going to get it...

 

[JimJarvis50]

Jim,
what do you think about the apparent EasyMega family [for lack of a better label] issue on the other BALLS 2 stage flight?

I don't know. Maybe it will be recovered at some point. We found some bugs a few years ago, and they were addressed. My particular easymega flew to 175K last year, and I'm comfortable with it.

Jim

 

[JimJarvis50]

I set it to 80% of the expected altitude at the staging time based on simulations and any programmed delay. Be aware that the reported time of burnout in simulations is usually the nominal time of the motor reaching low thrust, while altimeters sense burnout with deceleration, which usually happens earlier depending on drag and velocity.

The percentage is a balancing act between actual flight angle and all the slop in the sims, etc. If you needed tighter tilt control you would have to go to a tilt sensing solution. Without naming names, I remain unconvinced that some of the systems available really do this reliably under all conditions.

Remember that this particular flight was with active stabilization, and not a normal staged flight. It was set tight intentionally. Trust me when I say I have reviewed simulations versus flight data in great detail and for this rocket specifically. It doesn't bug me that much when the sustainer doesn't light if it's not a good flight. I just design things to handle the extra motor weight and then just try again (and the next time around is a "cheap date".

I have had the opportunity to investigate the tilt performance of the the EasyMega's (and the original Tiltometer), as well as with Kate. This is because I have gps as well as side-looking video on most flights. In most cases, the tilt predictions are pretty good. Just my experience.

Jim

 

[JimJarvis50]

I don't have tilt control on the altimeters I use and I don't use velocity gates.

Tilt control is ultimately a safety issue so the second stage doesn't go outside the waiver cylinder. For high flights that requires a pretty tight tilt limit. I haven't tried to go above 75K feet and at BALLS you don't really need tilt control at that altitude since it is physically impossible to land outside the 7-mile radius at any sane tilt, say up to 45 degrees, and you can get that with an altitude gate. At some point, certainly with Jim's flights, you need real tilt control just to stay within the 15-mile radius.

Of course, just because you set tilt control on an altimeter that provides it doesn't mean you are going to get it...

I have often said that an altitude check is for safety and tilt is for the waiver radius. Some people say that tilt only is fine, but I always use an altitude check even with tilt, although I relax the altitude check (per my posted example).

Jim

 

[Steve Shannon]

I have often said that an altitude check is for safety and tilt is for the waiver radius. Some people say that tilt only is fine, but I always use an altitude check even with tilt, although I relax the altitude check (per my posted example).

Jim

I’ve seen (and been responsible in one instance) three different flights or videos where the sustainer ignited and came to pressure at altitude but with the sustainer horizontal. I never worried about the waiver radius (although that could be an issue). My concern was always that a rocket under thrust heading horizontally or even down towards people or property could be very bad. In two of those an altitude constraint would have been met; the booster flew vertically. In my failed attempt I tried to set the timer to 1 second after MECO but my unfamiliarity with setting the beta version electronics resulted in returning to the default 10 second setting. Purely and simply that was my fault, but in my mind a tilt inhibit function would have enhanced safety and helped protect against waiver excursions.

 

[BABAR]

I have often said that an altitude check is for safety and tilt is for the waiver radius. Some people say that tilt only is fine, but I always use an altitude check even with tilt, although I relax the altitude check (per my posted example).

Jim

So what objective criteria are used to determine the minimum “safe” altitude to set the lockout?

 

[JimJarvis50]

I’ve seen (and been responsible in one instance) three different flights or videos where the sustainer ignited and came to pressure at altitude but with the sustainer horizontal. I never worried about the waiver radius (although that could be an issue). My concern was always that a rocket under thrust heading horizontally or even down towards people or property could be very bad. In two of those an altitude constraint would have been met; the booster flew vertically. In my failed attempt I tried to set the timer to 1 second after MECO but my unfamiliarity with setting the beta version electronics resulted in returning to the default 10 second setting. Purely and simply that was my fault, but in my mind a tilt inhibit function would have enhanced safety and helped protect against waiver excursions.

My point is that tilt (only) for safety is insufficient in my opinion. An altitude check should be used as well.

I'm curious about the two cases where you said that an altitude check would have been passed, but that the rocket went horizontal. With a reasonable altitude check setting, that should be rare.

Jim

 

[JimJarvis50]

So what objective criteria are used to determine the minimum “safe” altitude to set the lockout?

I think it "depends". The lowest-level approach would be an altitude check with something like 75% of the expected altitude. One thing that that does is provide one more permissive that would have to be met before the sustainer could fire on the ground. This approach at least gets the rocket away from people when the sustainer lights, and the flight is likely to be reasonably on plan if it gets to that altitude.

Next-level approach might be if you were using altitude only, but wanted to ensure a reasonably vertical flight. Then, you might use a tighter criteria. That's what I did on my LDRS flight.

Next level after that would be to use tilt. I relax the lockout altitude in that case, but you still want to use an altitude that ensures that the rocket is probably following the desired flight path. Generally, it is possible to set a value of tilt that has to be maintained for some period of time prior to sustainer ignition (a second or so for example, and that capability should be used. I would not set a really low-altitude lockout - like a launch detect altitude - what value is that?

Jim

 

[JimJarvis50]

Just for clarity, no one likes to see a sustainer light while horizonal or pointing downward. However, if this happens at 10K feet, it is unlikely that anyone on the ground will be hurt. On the other hand, if a sustainer lights on the ground, whether horizontal or vertical at the time, the odds of injury are much higher because people are much closer. It doesn't matter all that much what altitude criteria you pick for this safety advantage to be realized. I would never do a staged flight that didn't include an altitude check of some sort - period.

Jim

 

[Steve Shannon]

My point is that tilt (only) for safety is insufficient in my opinion. An altitude check should be used as well.

I'm curious about the two cases where you said that an altitude check would have been passed, but that the rocket went horizontal. With a reasonable altitude check setting, that should be rare.

Jim

I absolutely agree that tilt inhibition by itself is insufficient. And I consider you as probably the person who has devoted as much effort to safe staging as anyone. In both cases the boost was vertical and the sustainers coasted vertically until they arced over. Then they lit. Tilt inhibition would have prevented that. So would other methods. I would think (not having used one but thinking I understand the concept) that an altitude check would have a very narrow window in order to prevent this case.

 

[JimJarvis50]

I absolutely agree that tilt inhibition by itself is insufficient. And I consider you as probably the person who has devoted as much effort to safe staging as anyone. In both cases the boost was vertical and the sustainers coasted vertically until they arced over. Then they lit. Tilt inhibition would have prevented that. So would other methods. I would think (not having used one but thinking I understand the concept) that an altitude check would have a very narrow window in order to prevent this case.

I think you make my point. For the flights you mentioned, it would probably have been possible to select a more appropriate lockout altitude, but because there was a lockout altitude, the sustainers were away from people when they lit. In those cases, the fact that the sustainers lit was less of a safety problem and more of a waiver problem (since there aren't many places, including Blackrock, where you can light a sustainer horizontally and stay within the waiver). Yes, use tilt too, but altitude is the first line of defense for safety.

A while back, I studied some flight data from FredT on his multi-staged flights. He sent data from quite a few flights, and I was impressed at how tightly he set his altitude criteria. As I recall, he had some cases where he had about a half second window. Based on his experience, I tightened up my criteria just a bit. It is entirely possible to do this - to set altitude criteria that ensures a reasonable flight (and that would prevent sustainers from lighting while arcing over).

Jim

 

[JimJarvis50]

I was able to do the flight today (the M3100 to M745 test rocket flight). It went quite well. I did the boost with roll control only (since yaw/pitch wouldn't do much with the longer stack anyway), separated the stages at 3 seconds after a 2 second burn, and then turned on yaw/pitch at 4 seconds. You can clearly see the yaw/pitch control come on in the video. At this point, I have not seen the gps or the IMU data, so there is still much to see. I do have a crappy ground video and both up and down on-board video, so there are some things to look at. When the system works well, the rocket backslides because it is vertical at apogee. This flight didn't quite backslide, but it was close (the pic is right at apogee). There were supposed to be pretty strong winds above 18K (I went to 22K) so I think the system actually did quite well and the rocket was close to vertical at apogee.

Jim

PS - The spin can got a thorough workout on this flight. Lots of wind shears on the way up.

 

[JimJarvis50]

The results from this flight continue to look really good. The attached graphs show the tilt through the flight as measured by the UDB5 (which is controlling the active stabilization) and the EasyMega (controlling the staging). The two graphs are in reasonbly good agreement. You might expect the UDB5 to look good, since it is controlling the flight to itself, but the EasyMega is just reporting what it thinks happened. The fact that they are in agreement is helpful.

The main flight events included burnout at 2 seconds, stage separation at 3 seconds, activation of yaw/pitch control at 4 seconds, 2nd stage ignition at 15 seconds, burnout at 23 seconds, and apogee at 46 seconds.

In both cases, the data show that the rocket went off at an increasing angle, reaching about 9 degrees at the point when yaw/pitch control was engaged. At that point, both instruments show that the rocket moved back t o nearly vertical and then stayed there through the sustainer burn until just before apogee. The oscillations in tilt just after the control was turned on (at 4 seconds) are evident in the data and in the flight video. Both instruments show more tilt around the time of sustainer ignition, and the video shows that too.

Before too long, I'll post the gps results. The flight was nearly vertical, although the rocket does drift downwind a little (as it should if it's vertical and there is wind), and the gps flight path shows the correction in the flight trajectory at 4 seconds. This is just about everything I could have hoped for from this flight.

Jim

 

[rharshberger]

Sounds like you are definitely getting there Jim, great work and its fun follow your progress

 

[OverTheTop]

Nice results Jim. Great progress .

 

[Serac]

Very nice results!

Are you scheduling/changing the gains over time? Over dynamic pressure? There seems to be a potential issue between 4-11 seconds. Is this related to the booster? You haven't plotted altitude and speed. But I would wager that it is related to the dynamic pressure at this time - you are still fairly low at this point.

I've read this thread from the beginning - but it is a lot to digest.

 

[JimJarvis50]

Very nice results!

Are you scheduling/changing the gains over time? Over dynamic pressure? There seems to be a potential issue between 4-11 seconds. Is this related to the booster? You haven't plotted altitude and speed. But I would wager that it is related to the dynamic pressure at this time - you are still fairly low at this point.

I've read this thread from the beginning - but it is a lot to digest.

Post #631 has the simulation for the flight. It should be pretty close because I got 22.7K out of a predicted 24K. The only difference will be that the sustainer motor came up to pressure at 15 seconds (a couple seconds later than the simulation).

No, I'm not using gain scheduling. Might consider that though if I try this system at supersonic speed. I would be interested in what folks think of the "control" response when the yaw/pitch went active at 4 seconds. Yes, there is some oscillation, but remember, this is simple proportional control. It would have been moving at a pretty good clip at the point of activation, and the response time for moving back to vertical (a few seconds) seems pretty good to me. I did have some roll on the flight, so I probably cut the roll gain back too much from previous flights (I cut back roll gain by a factor of three relative to yaw/pitch, and then reduced the overall gain by about a factor of two because of the near-mach flight speed). I only need to reduce roll to the point where yaw/pitch has a chance to work, but it looks like maybe twice the current gain might be about right.

Jim

 

[Serac]

Post #631 has the simulation for the flight. It should be pretty close because I got 22.7K out of a predicted 24K. The only difference will be that the sustainer motor came up to pressure at 15 seconds (a couple seconds later than the simulation).

Thanks for that. It doesn't look like dynamic pressure because that range of time has the slowest speed and altitude isn't that high. Your control loop is really trying to control two different rockets with one tuning: with the booster, and without the booster. I suspect that your controller is not as well tuned for the early stage with the booster than without. It may be helpful to look at the moment arms between the canards and the centers of mass for the two configurations. With the booster attached, the moment arm may be longer. This would essentially increase the proportional gain for that configuration.

No, I'm not using gain scheduling. Might consider that though if I try this system at supersonic speed. I would be interested in what folks think of the "control" response when the yaw/pitch went active at 4 seconds. Yes, there is some oscillation, but remember, this is simple proportional control. It would have been moving at a pretty good clip at the point of activation, and the response time for moving back to vertical (a few seconds) seems pretty good to me. I did have some roll on the flight, so I probably cut the roll gain back too much from previous flights (I cut back roll gain by a factor of three relative to yaw/pitch, and then reduced the overall gain by about a factor of two because of the near-mach flight speed). I only need to reduce roll to the point where yaw/pitch has a chance to work, but it looks like maybe twice the current gain might be about right.

You're right. The few second rise time during the 4-11 second time window seems reasonable. It doesn't seem too fast. However, whenever I see oscillations like this on a closed loop system, alarm bells go off in my head. The oscillation could mean that it is in a cusp of stability, and any disturbance could cause the system to go unstable. However, it sounds like you've already reduced the gain, so the oscillation may be due to "laziness" and not near-instability. It may help to add a sign to the tilt angle calculation. If the tuning is indeed lazy during this period, I would expect it to bounce around a mean of zero. Since your tilt angle calculation is always a positive angle, this can't be seen. Without adding a derivative feedback term (D), you may not be able to damp out this oscillation.

 

[JimJarvis50]

Thanks for that. It doesn't look like dynamic pressure because that range of time has the slowest speed and altitude isn't that high. Your control loop is really trying to control two different rockets with one tuning: with the booster, and without the booster. I suspect that your controller is not as well tuned for the early stage with the booster than without. It may be helpful to look at the moment arms between the canards and the centers of mass for the two configurations. With the booster attached, the moment arm may be longer. This would essentially increase the proportional gain for that configuration.



You're right. The few second rise time during the 4-11 second time window seems reasonable. It doesn't seem too fast. However, whenever I see oscillations like this on a closed loop system, alarm bells go off in my head. The oscillation could mean that it is in a cusp of stability, and any disturbance could cause the system to go unstable. However, it sounds like you've already reduced the gain, so the oscillation may be due to "laziness" and not near-instability. It may help to add a sign to the tilt angle calculation. If the tuning is indeed lazy during this period, I would expect it to bounce around a mean of zero. Since your tilt angle calculation is always a positive angle, this can't be seen. Without adding a derivative feedback term (D), you may not be able to damp out this oscillation.

For this flight, only roll control was on at launch. There was a two-second burn, then the booster separated at three seconds, and then yaw/pitch stabilization started at 4 seconds. So, only the sustainer was vertically stabilized, starting at 4 seconds. I think the stack could be stabilized even though the moment arm is longer, but being a fiberglass rocket, there is enough movement through the air frame to where I would be worried about adding a "delay" to the system response. As it is, you can see the air frame bend a little when the yaw/pitch is engaged.

What I plotted in the graphs was tilt, but with the UDB5 data, I can tell the sign of the tilt. I don't think there was much roll going on during the initial adjustment to vertical, so I should be able to tell if the rocket is oscillation towards zero or through zero. Good suggestion.

Jim

 

[VernK]

What I plotted in the graphs was tilt, but with the UDB5 data, I can tell the sign of the tilt.
Jim

What does the sign of tilt indicate? I think of tilt as being away from the vertical and there is no sign. It can be tilted to the north, south, east, west or anything in between but it is still a positive tilt number. What am I missing? Or are you referring to the sign on the pitch and yaw?

 

[Serac]

What does the sign of tilt indicate? I think of tilt as being away from the vertical and there is no sign. It can be tilted to the north, south, east, west or anything in between but it is still a positive tilt number. What am I missing? Or are you referring to the sign on the pitch and yaw?

The sign could be made up, as long as it was consistent. The control system is working on roll/pitch/yaw, not tilt. So, being able to see sign changes in pitch and yaw could be instructive (using a definition of 0-degrees pitch and yaw to be zero tilt). You are essentially losing information by looking only at tilt off vertical. For example, imagine the rocket ascending in a coning pattern. This would look like a constant tilt angle. But, the pitch and yaw are changing signs. A signed tilt angle could show this movement.

 

[JimJarvis50]

What does the sign of tilt indicate? I think of tilt as being away from the vertical and there is no sign. It can be tilted to the north, south, east, west or anything in between but it is still a positive tilt number. What am I missing? Or are you referring to the sign on the pitch and yaw?

Yes, I can look at positive or negative pitch and positive or negative yaw with the UDB5 data. With no spin at that particular time (I think), it will be easy to tell how the rocket was moving.

I really only posted the two tilt graphs so that the EasyMega and UDB5 data could be compared. The EasyMega only gives tilt and not the individual axes. It is important, to me at least, that the two are in agreement.

Jim

 

[JimJarvis50]

So, here's some more data.

With respect to the positive and negative yaw/pitch values, the first picture below shows the vertical components of yaw and pitch, along with the resulting tilt. There was a little roll during this period, which makes it a little harder to interpret the data, but it looks to me (as you might expect), that the system is trying to control across zero tilt (versus working back towards zero tilt).

There are three gps graphs. The first is the "side wind" graph, which shows the angle of the ascent and descent. A waypoints are shown for the apogee position at 46 seconds. The descent part clearly shows the effect of the wind that was present. Roughly, while on drogue, the rocket drifted to the landing location at an average speed of about 25 mph (on the order of the forecast speed). On ascent, the rocket moved from the launch location to the apogee location at an average speed of about 24 mph. So, if the rocket was vertical, but was blown sideways at 24 mph on the average, then the apogee location was very close to where it should have been. Vertical stabilization isn't all that great when there is wind because you drift downwind on ascent instead of weathercocking into the wind. That can increase the recovery distance. On the other hand, you know pretty closely where the rocket is going to land. I estimated a landing location a few hundred yards from where it actually landed, just based on the wind-versus-altitude forecast.

The second gps plot is the "down wind" plot. The ascent part of the plot is pretty close to exactly vertical.

The third gps plot is the data around the 4-second point where yaw / pitch was turned on. The 4 second waypoint is shown on the plot. The change in trajectory at 4 seconds is pretty obvious.

Jim

 

[JimJarvis50]

I just posted the flight video from Distant Thunder. Enjoy!

Jim

 

[tim cubbedge]

Wow. That was a great flight.

 

[Serac]

Been meaning to reply - this is a fantastic flight! The figures and the video are great.

The video says "Raven altimeter - to activated roll & yaw/pitch during flight". Can you explain this more? So, the Raven is controlling the servos? Clearly, the UDB5 board running a modified MatrixPilot code base is doing the control. I just don't understand how a deployment altimeter runs the servos (although, I understand that the Raven is overengineered a bit more than others).

The spec on the MPU-6000 IMU is not significantly different than other smartphone IMUs used on other UAV flight control boards. For basic attitude control/stability, MatrixPilot is essentially very similar to ArduPilot, CleanFlight, and their numerous derivatives.

Does Bill give you any kind of description as to what he did to MatrixPilot? Did you give him the basic dimensions of your rocket? RockSim drawing?

 

[JimJarvis50]

Been meaning to reply - this is a fantastic flight! The figures and the video are great.

The video says "Raven altimeter - to activated roll & yaw/pitch during flight". Can you explain this more? So, the Raven is controlling the servos? Clearly, the UDB5 board running a modified MatrixPilot code base is doing the control. I just don't understand how a deployment altimeter runs the servos (although, I understand that the Raven is overengineered a bit more than others).

The spec on the MPU-6000 IMU is not significantly different than other smartphone IMUs used on other UAV flight control boards. For basic attitude control/stability, MatrixPilot is essentially very similar to ArduPilot, CleanFlight, and their numerous derivatives.

Does Bill give you any kind of description as to what he did to MatrixPilot? Did you give him the basic dimensions of your rocket? RockSim drawing?

The UDB5 is set up with two pins that separately turn on yaw/pitch and roll control. The Raven is used to take one or both of those pins to ground when you want to turn on control. In this flight, I used the Raven to turn on yaw/pitch 4 seconds into the flight, just based on a timer. One advantage of using a Raven for this is that the output can be latched.

One of the slides in the video goes over the changes that we have made to be able to use the program for rockets. Most of them are discussed at some point in this thread, but all are important for a functioning system. Bill and I (and Frank Hermes of tiltometer fame) have been working on this on-and-off for several years. We do a flight, analyze the flight results, make modifications to the flight program and/or hardware, and then repeat the process. Although it has been pointed out that there are limitations in the equipment, the system does work, and the last half dozen or so flights have been pretty good.

Jim