Some Work on Two-Stage Safety...

Ever since I designed, built and flew my first HPR two-stager this past year, Double Take, I have been contemplating two stage design safety especially with respect to some of my larger upcoming builds such as the RDS Orion. I know Adrian A. has been a long-time proponent of sustainer ignition disconnect and even some mechanism for measuring altitude before firing the second stage. Although good, I don't believe it gets to the highest risk situation, which in my estimation is horizontal, or worse yet, below horizontal ignition of the sustainer and the potential ensuing danger to launch attendees. There are some options I have seen to control such an event; one example is the Tiltometer, but these tend to be pricey. Thus I have been working on a circuit implementation to solve this problem. In this endeavor I have tried to make the design simple, cost-effective and, with this first prototype, using generally off-the-shelf components. It involves a combination of a the Wooshtronics µMAD (which I purchased to upgrade one of my night rockets), a latching relay and appropriate discrete power up components. The overall cost in single unit quantities is ~$40 and, if integrated in 100's parts quantities, I would expect a cost of less than $15.

Please be warned--THIS IS A WORK IN PROGRESS and is not currently recommended for integration into one of your rocket designs, i.e., I disavow all responsibility for premature/inappropriate use. But, saying that, I will be building/testing this over the coming months and let you know how things transpire.

To gather my thoughts and document a bunch of back-of-napkin "chicken scratch", last night I wrote up a draft design description. For those that have an interest and want to provide constructive feedback, it can be found here: Ignition Inhibit Circuit Description.

I too am interested in "two stage design safety." I elected not to launch my J-to-H 2-stage last summer as I realized I hadn't adequately considered the risks. I'm adding positive retention to the stages and will initiate separation early in the booster coast phase. However, if the rocket is horizontal etc. it's still a safety risk.

I'm planning a HPR 2-stage and would like to improve altitude by delaying separation and firing of the sustainer. Without knowing the angle from vertical at time of separation the safety and failure risks are still there.

I'm looking forward to tracking this thread to seeing your thoughts and other's input.

As I have no experience with launching staged rockets and little knowledge of electronics I can't contribute much. I do have a few questions that may be answered as the thread continues.

How accurate is the magnetic sensor? +/- number of degrees? Went to other websites couldn't find anything.
I wonder about response time between change in angle of attack and change in output signal.


Thanks,
StanO

StanO said:

How accurate is the magnetic sensor? +/- number of degrees? Went to other websites couldn't find anything.
I wonder about response time between change in angle of attack and change in output signal.

Click to expand...

Initial tests seem to be as you approach ~5 to 10 degrees of horizontal. Have not measured exact rads/sec but I've moved it at a fairly fast pace.

StanO said:

I too am interested in "two stage design safety." I elected not to launch my J-to-H 2-stage last summer as I realized I hadn't adequately considered the risks. I'm adding positive retention to the stages and will initiate separation early in the booster coast phase. However, if the rocket is horizontal etc. it's still a safety risk.

I'm planning a HPR 2-stage and would like to improve altitude by delaying separation and firing of the sustainer. Without knowing the angle from vertical at time of separation the safety and failure risks are still there.

I'm looking forward to tracking this thread to seeing your thoughts and other's input.

As I have no experience with launching staged rockets and little knowledge of electronics I can't contribute much. I do have a few questions that may be answered as the thread continues.

How accurate is the magnetic sensor? +/- number of degrees? Went to other websites couldn't find anything.
I wonder about response time between change in angle of attack and change in output signal.


Thanks,
StanO

Click to expand...

A compass needle is used to visually indicate the magnetic N-S vector location anywhere on earth. If you know the local variation you can easily determine the true N-S vector from your location. The MAD method of apogee detection uses the earths magnetic field as an external reference for orientation just a like a compass does, but uses the sudden large change in the magnetioc vector as an indication that the rocket orientation has change by a large amount assumeed to be apogee tip-over.

Bob Galejs, a former member of CMASS and employed at MIT Lincoln Laboratory, was the inventor of the MAD (Magnetic Apogee Detector) in the 90's. It was first published in the CMASS Senteniel and can be downloaded from Aerocon Systems which sells an Analog MAD kit. https://www.aeroconsystems.com/electronics/magnetosensor_images/Magnet3.PDF

The Analog MAD requires adjustment for local magnetic variations. The uMAD mentioned in Jim's post appears to eliminate any manual adjustments by adding a PIC uP to the analog sensor circuit. This make the device much easier to use, but IMO may not be good enough to use as a go/no go tilt sensor. The original MAD concept looks for the large change in magnetic field orientation when a rocket arcs over at apogee so high resolution is not required. My concern is that most rockets rotate, and unless great care is taken to insure that the sensore is precisely aligned with the ceter of rotation of the rocket, the magnetic field strength sensor output willl vary sinusoidionallly at the roll rate and degrade the tilt angle resolution. I'm not sure in practice that you can accurately measure a 10-15 degree tilt reliably with a single axis magnetic field sensor.

I speculate that if you want high angular resolution you might need to measure the magnetic field strength on all 3 axis. The first pages of the following reference do a good job describing how these magnetic field sensors work.

https://ve6sbs.sbszoo.com/projects/compass/kmz52an00022.pdf is a Phillips App Note on how to use a 1-axis KMZ51 and 2-axis KMZ52 coupled with a uP as 1 degree resolution 3-axis compass which is the type of tilt resolution I believe you need for this safety application. This would kick the retail cost of the $40 uMAD to $60 or so.

Once you go this far, with the addition of a 3-axis gyro chip and a 3-axis accelrometer chip you have a 9-DOF IMU (Inertial Measurement Unit. A 9-DOF German unit was mention here within in the last few week that cost just over $100 and it had communications ports for a GPS, so for about $200 you could have a complete INS (Inertial Navigation System). For a rocket where GPS lock can be lost under power, an external pressure sensor, barometer and larger range accelerometer would be required, but this is exactly what the Raven and other high end rocket altimeters already have.

Bob

bobkrech said:

Once you go this far, with the addition of a 3-axis gyro chip and a 3-axis accelrometer chip you have a 9-DOF IMU...

Click to expand...

I also have some reservations about the one-axis magnetometer but I haven't done the error analysis to understand the sensitivity. Note that the magnetic field of the earth is at various angles with the local vertical as a function of launch position, and a one-axis measurement only senses one angle, obviously.

That said, it seems to me that looking at this is well worth it. I could imagine that an integrated system with a one-axis magnetometer for 1-dof orientation and a barometric sensor for altitude might be very safe, and that has the promise to be at least 2-3x cheaper than a full IMU solution, and potentially more resilient to noise.

It has been a few years but I once flew a three axis magnetic sensor and recorded the data using an RDAS.

a sample

As luck would have it, I happened to fly it off of a rail which used iron pipe as reinforcement. The result was a significant change in the orientation of the local magnetic field and interesting effects as the rocket was coming off the rail. The point of this story is do not rely on a magnetic sensor if you use a launcher that is made from ferro-magnetic materials.

For measuring attitude all that is really required is a three axis gyro and a micro-controller to integrate the output. (The math behind that integration is a bit tricky but quaternions are your friend.) A quick check at Mouser finds three axis gyros for less than $10 each.

UhClem said:

A quick check at Mouser finds three axis gyros for less than $10 each.

Click to expand...

Fair point, but rate range, rate resolution, and drift could make this a challenging problem for many cheap MEMS gyros.

That said, it might be interesting to experiment with something like the ITG-3200 https://invensense.com/mems/gyro/itg3200.html

mikec said:

Fair point, but rate range, rate resolution, and drift could make this a challenging problem for many cheap MEMS gyros.

Click to expand...

L3G4200D:
Selectable range: 250, 500,2000 degrees/sec.
Resolution is 8.75 to 70 milli-degrees/sec depending on range.
Drift is low. (Drift over the time from launch to staging is going to be small anyway because the interval is short.) But missing is acceleration sensitivity so that would need to be tested.

If 2000 dps is too low, you should really reconsider what you are doing. In any case, exceeding some roll rate threshold should probably be a condition to lock out staging.

Another concern about gyros is the sampling rate the system can manage and the computational load of doing the integration (I'd have to look at how open-source code like ArduIMU is doing the DCM calculation -- I think they do it at 50 Hz and use single-precision software floating-point), and the resulting integration errors (though the device does have internal digital filtering). The nice thing about the magnetometer is that it provides direct sensing of orientation without relying on integration.

But talk is cheap and flight testing is clearly required for both approaches.

I can't make any serious contribution to this discussion, as I am completely unfamiliar with all these 'exotic' electronics devices.

However, I commend all of you guys for being aware of flight safety as well as concerned about it, and for being willing to DO something about it.

We need more hobbyists like you guys.

Please keep us all posted on how the new hardware/techniques play out--

mikec said:

Another concern about gyros is the sampling rate the system can manage and the computational load of doing the integration (I'd have to look at how open-source code like ArduIMU is doing the DCM calculation -- I think they do it at 50 Hz and use single-precision software floating-point), and the resulting integration errors (though the device does have internal digital filtering). The nice thing about the magnetometer is that it provides direct sensing of orientation without relying on integration.

But talk is cheap and flight testing is clearly required for both approaches.

Click to expand...

You need to perfrom a fair bit of computations if you are using a magnetometer for tilt orientation to filter out noise spikes as shown in Dave's flight data, and it's alot simpler if the data is digital to start with.

The Do It Yourself Drone community is way ahead of the rocket community in the development of GPS synchronized 6-DOF and 9-DOF IMUs and autopilot systems.

https://www.diydrones.com/notes/ArduPilot

Their 6- and 9-DOF IMUs use digital, not analog, gyros, so the outputs are already converted at the sensor and linearized into digital degrees/second value. A 50 Hz sampling rate is more than sufficient for normal attitude monitoring and flight contol corrections as the normal flight dynamics of any vehicle can be predicted and rates in excess of this value are only possible if something bad happened to the vehicle.

The same is true for any digital sensor: pressure sensors report a linearized digital pressure value; accelerometers report a linearized digital acceleration value, and magnetometers report a linearized digital field strenght value. You really can't set up an IMU to monitor the flight of your rocket unless you understand the normal dynamic behaviour. For exampe if your rocket spins at 10 Hz (3,600 degrees/second) you can use a 2,000 degree/second gyro. If your rocket accelerates at 100 g, you can't use a 50 g accelerometer. Similariy if your rocket accelerates at 10 g, you don't want to use a 200 g accelerometer as you are throwing away a lot of resolution.

Digital sensors usually contain a processing unit that can apply calibration factors, correct offset and can correct for temperature dependent sensor response. They are far more acccurate than most conventional analog sensors and that's one of the reasons why they are taking over most control applications. Unless you need extremely high response times, digital sensors are the preferred way to go as they off-load the CPU of the digitization of the sensor data and take on the sole role of number crunching. This does not mean you can get away with a slow processor, but it does mean you can spend virtually all your cpu time for determing your vehicle position instead of digitizing sensor data.

Bob

A few thoughts on this:

The Earth's magnetic field lines in North America range from about 15-30 degrees from vertical.

https://en.wikipedia.org/w/index.php?title=File:World_Magnetic_Inclination_2010.pdf&page=1

It's actually easier to measure up and down with a magnetic field than measuring North. That's a good thing for rocket tilt detection, because you can keep the sensor axis aligned with the rocket and be pretty insensitive to the rocket roll. The bad news is that the Earth's magnetic field is still pretty weak when it comes to being immune to disturbances. Electronic circuits, ferromagnetic materials near the pad, etc. can disturb the field enough to prevent a sensor from working. The magnetic sensors used for the Featherweight magnetic switches and av-bays trigger on fields that are much, much stronger than the Earth's field that you would be trying to sense for this. Even so, I know of at least one case in which an av-bay threaded rod was accidentally magnetized, causing the switch to be inoperable while in the vicinity of that rod.

But for an airstart inhibit, the most likely outcome if something goes wrong with sensing the Earth's magnetic field is that you'll fail to ignite the upper stage. That's an annoying and somewhat expensive failure, but not a safety-critical one. You would need a combination of a magnetic-based inhibit failing to notice the rocket tipped over, plus whatever failure caused the rocket to tilt over at the ignition time, for there to be a real safety issue. So for this application, perhaps a magnetic sensor is viable.

In the longer run, gyros measure quite directly the orientation of the rocket you need for a safe airstart, without the sensitivity to getting spoofed by the magnetic field The Invensense gyros are in the $10-$15 range, and have sufficient drift performance to get the job done for this application. The math to convert the local rates into propagated attitude is pretty computationally expensive, however. Given the relatively small subset of this niche market, I can't really justify adding $30+ to the retail cost of a Raven, plus development time, plus throttling back recording rate to make room for the calculations, for this capability. Maybe sometime I'll make a special version or a stand-alone project. Invensense is apparently also moving in the direction of propagating the attitude on the chip and outputting quaternions, which would reduce the computational burden and make it much more attractive to integrate onto a standard altimeter.

But take a step back for a moment and consider all the failure modes that could cause the upper stage to be pointed in a bad direction several seconds after launch, including CATOs, airframe breakup, motor underperformance, dynamic or aerodynamic instability, etc. Probably 80%-90% of those scenarios would be covered by just verifying that the rocket is at a predicted altitude before a predicted time before igniting the upper stage. I'm all for advancing the state of the art to improve airstart safety, and Frank Hermes' RocketTiltMeter is a really excellent advance in that area. But several altimeters right now can prevent the vast majority of unsafe airstarts, just with careful use of the ignition settings that are already available.

Making new devices won't much to help safety of the hobby as a whole as long as most people doing high-powered airstarts continue to use blind timers to do it. There's a growing recognition of the risks of on-board ignition systems when the rocket is on the ground, but not many are concerned about the practice of blindly igniting the upper stage after launch after the rocket's orientation is out of anyone's control.

Here's another thread that I started in the wake of the LDRS29 accident that goes over pre-launch and post-launch airstart safety:

https://www.rocketryplanet.com/forums/showthread.php?t=6757

bobkrech said:

My concern is that most rockets rotate, and unless great care is taken to insure that the sensore is precisely aligned with the ceter of rotation of the rocket, the magnetic field strength sensor output willl vary sinusoidionallly at the roll rate and degrade the tilt angle resolution. I'm not sure in practice that you can accurately measure a 10-15 degree tilt reliably with a single axis magnetic field sensor.

Click to expand...

Bob,

Thanks for pointing out rotation as a potential trigger. I had already planned this as part of the test suite, but may look at adding to this area of testing.

-Tim

The math to convert the local rates into propagated attitude is pretty computationally expensive, however.

Click to expand...

I don't think that it is as bad as you think. For one floating point (software or hardware) isn't required because of the limited dynamic range. In addition it is possible to do it without a single trig function.

When using the quaternion method you do need a sine and cosine when building the rotation quaternion from the latest gyro reading. But if you keep the sample rate high enough you can replace those with the small angle approximations. Which saves a very expensive trig library call.

Once you have rotated your attitude finding how much you have departed from the initial attitude is a simple dot product. While converting the result of that to an angle requires a trig function, you don't need to compute that in real time. The cosine of the angle works just as well as the angle for deciding when to enable staging.

I played around with this a while back after SparkFun launched a model rocket with one of their loggers and a gyro in it. Looking over that code it appears to be mostly multiplication, addition, along with one square root and a divide that can't be avoided. I am pretty sure any reasonable micro-controller can handle the task. And if that doesn't work then there is this unreasonable one (STM32F407) I have been playing with lately...

I really do have to stop following this thread or the next thing you know I will be designing hardware.

Adrian A said:

But take a step back for a moment and consider all the failure modes that could cause the upper stage to be pointed in a bad direction several seconds after launch, including CATOs, airframe breakup, motor underperformance, dynamic or aerodynamic instability, etc.

Click to expand...

I agree. The only thing that occurs to me is a CATO leading to a big acceleration spike that causes the accelerometer altitude to look much higher than it really is. I've had similar things happen that inhibited any deployment because the velocity lockout got faked out by the spike. Of course, if you're using barometric this won't happen. [Edit:never mind, this would only affect airstart safety if you were using only a velocity threshold, which no one would do.]

I think people are still using timers because they're cheap and simple. If one could integrate some safety check into a cheap and simple device, that might fill a niche. I always enjoy trying to do something with the fewest number of components.