2017-09-26

The test stand blast deflector for vertical testing is complete. It uses three 24 x 24 x 2 inch concrete pavers at 30 degree increments to deflect the exhaust horizontally down the flame trench in my test pit. The frame is made of 1.5 x 3/16 inch steel angle. A negative step of about 0.1 inches between pavers will hopefully minimize erosion on the leading edge of the downstream ones.

I had to re-spin the signal conditioning PCB used for the bridge sensors. The original design used an INA125 instrumentation amplifier in a single supply configuration which limits the bridge common mode voltage to a range of about 1 to 4 V with a +5 V supply. I didn't think that was a problem because the pressure sensors I used for checkout were all within that range. However, some new Kulite sensors I purchased for the vehicle had common mode voltages around 0.75 V which was out of the amplifier's range with a single-ended supply. The fix was add a TL7660 negative voltage converter to generate -5 V so the op amp can handle inputs closer to 0. Even with a fair amount of filtering, there is still about +/- 10 mV of ripple on the -5 V rail but I ran it through the full input and common mode range and didn't observe any adverse effects. I also took the opportunity to calibrate the three new transducers I recently purchased off eBay.

To minimize the chance of overheating the Digi XT-09 modem, I added a heatsink bracket that conducts the heat away from the back side of the XT-09 into the aluminum airframe longerons. Tests showed the internal temperature dropped about 23 deg C with the addition of the heatsink when running at 1 W. Since there won't be a lot of convective airflow in the avionics bay while sitting on the launch pad in the sun, I may end up adding a small 5 or 12 V fan as cheap insurance against the flight computer overheating. The CDI box and power panel have been mounted along with a battery holder. The electrical tasks remaining are to mount the GPS antenna, mount the TM antenna, and to fab the cable between the CDI sense lines and signal conditioning board.

On the engine, I replaced the aluminum Cv plugs that were used to close out the cross-drilled converging holes with threaded holes for use with 1/4-28 bolts and Parker Stat-O-Seals. I had wanted a way to clean out those passageways but couldn't find a suitable seal at the time. I plan to use the fluorosilicone Stat-O-Seals which should be good to about 400 deg F for a short time test. Opening up those holes allowed me to use a wire tube brush to clean out the gunk and mild corrosion left over from some water tests I ran last fall. All the plumbing for the vehicle has been re-cleaned and is ready for the upcoming hot-fire.

To hold down the rocket and test stand during the vertical test, I had originally planned on pouring a concrete pad in the test pit and bolting the test stand to the pad with concrete anchors. However, I discovered earth anchors and I believe they will be less work and should be just as effective for this application. There are many types but the ones I found are steel rods with a corkscrew plate on the end that you screw in the ground. They have >1000 lb of holding power each so I plan to use four of them with appropriate steel cables to tie down the test stand and keep the rocket from flying away during the test.

The next task is to build the support structure to attach the rocket vertically to the test stand.


2017-06-10

For breakaway connectors on my rocket, I had originally planned to use DE-9 connectors and RJ-45 jacks w/o the tab. But I was having trouble figuring out to keep the protuberances to a minimum and I didn't like the the off-axis loads on the pins and shells. I remembered the MagSafe connectors for Apple laptops but I couldn't find anything off the shelf that was: 1) orderable, 2) had enough pins, and 3) was easy to integrate into the skin of a rocket.

It doesn't look too complicated so I decided to try making a custom version. I found these Mill-Max spring loaded pins which are a ready-made solution. Connectors with multiple pin arrangements are orderable from Digikey and Mouser. Some round magnets with a countersink are the last piece needed to make the connector. Some photos of the prototype:

The white acetal blocks are 0.5 by 2.0 inches and the pin spacing is 0.1 inches. The parts cost for the prototype is around $15. I went with the concave target pins instead of the flat ones for the vehicle side because I thought they would be better for the battery charging connections. The contacts are rated at 2 A continuous, 3 A peak. The spring loaded pins have a travel of 0.055 inches with a spring force of .15 lbf at mid-deflection so a connector with 8 of them needs at least 1.1 lbf of compression to hold it together. The magnets have a spec of 4.2 lbf but it's not clear if that's with another magnet or a steel plate.

To register the two assemblies together, I recessed the magnets by 0.02 inches on the vehicle connector and then let the magnets protrude an equivalent amount on the removable connector. The magnets are strong enough that if you get them anywhere close to each other, they just want to snap in place. Opposite polarity magnets on each end of each connector allow it to be assembled in only one orientation. I drilled two holes in each end of the removable portion to attach a lanyard so when the vehicle lifts off, it won't pull on the wires. The two halves separate cleanly as expected. To test the signal integrity of the connector, I cut a network cable in two and wired it to each side. I'm only going to use it at 100Base-T speeds (and you only need 2 pair for that) but at GB speeds, iperf shows 894 Mb/s in one direction and 938 Mb/s in the other. The original unmodified CAT-5E cable ran at 938 Mb/s.

Here's the production version with 18 pins:

The GSE (Ground Service Equipment) box will be located about 25 feet away from the launch pad (inside, rear). The enclosure is a Ridgid toolbox from Home Depot and contains the following:

  • Battery charger and ground power supply
  • CDI power supply for the igniter
  • Solenoid power supply
  • Beaglebone Black for remote reset and serial console debug of the onboard flight computer
  • WiFi router (cabled to the flight computer for ground operation) with extra ports on the rear
  • AC switch and receptacle for powering test equipment on the pad

To minimize onboard complexity, the flight battery connects directly to the ground power without a dedicated onboard charging circuit. This is not optimal from a battery life perspective but is commonly used in low-cost applications and is certainly acceptable here. An external li-ion charger won't work since it can't distinguish between the load and battery so I made an external current-limited power supply using an LT1185 which limits the charge to 2 A (safe charge rate for the battery). Trickle charging at 4.1 V/cell does not fully charge the battery but significantly increases the lifespan when using this scheme. The voltage drop in the cables, input diode, etc. are significant but adjusting the offboard regulator to approach 16.4 V (4.1 V/cell) as the charge current tapers off to 0 works well.

The vehicle power panel is finished and will be mounted in a bay near the flight computer. It has a 5 A circuit breaker for the onboard battery, an external power switch, and an ARM/SAFE switch.

To minimize interference between the TM transmitter and GPS receiver, I'm using a PCTEL 3961D-HR high-rejection active GPS antenna. With the TM transmitter at 1 W on 900 MHz, most of the other GPS antennas I tried lost all satellites, even with the antennas 1-2 feet apart. I really wanted a active quadrifilar helix antenna with an appropriate filter but was unable to find one readily orderable. I also discovered that I need to send the UBX-AID-HUI message to the GPS receiver at startup or the UTC time is off by 2 seconds for up to 12.5 minutes until it receives the full navigation message.

The CDI electronics box for the igniter is complete. It houses the RC-EXL module plus a small PCB containing the opto isolated current signal conditioning, a solid state relay, and a 555 timer to generate the pulses for the CDI module.


2016-12-29

Some progress has been made the past few months to prepare for the static test, just not as much as I had hoped. Most of the design for the static test holddown is complete. The existing test stand structure will be reused but extra components have been added to secure the rocket vertically. Two 8-foot 3030-S beam extrusions from 80/20 Inc. will be attached to the test stand and supported by aluminum angles. The 80/20 beams are more than I need for this application but I plan to use them later during final assembly to help with alignment of the skin to the frame. The rocket will be mounted 6 inches away from the beam using 3 x 6 x 1/4 inch square tubes as spacers. The thrust load will be reacted through two 1 x 1 x 1/8 inch angles that tie into the injector manifold at the top of the engine and then to the vertical beam. The rocket will also be held in place at one or two other points using a cradle and hose clamp arrangement around the fuselage frames. I also plan to pour a concrete pad with mounting studs in my test pit to secure the test stand. I haven't designed the blast deflector yet.

One feature I wanted onboard was feedback from the CDI ignition module to verify it is working before opening the igniter valves. However, the CDI power supply comes from an isolated offboard source so I had to design an optoisolated current monitor circuit that ties into the onboard data system. The CDI module and a small circuit board with the Vishay IL300 linear optocoupler and op-amp will be housed in a small plastic enclosure that will be located near the main propellant valve. All the parts are here, I just need time to wire up the circuit.

I did make a lot of progress on the flight and ground software in preparation for the static test. The flight software now has full discrete support and proper support for a "current value" or CVT message. At a few places in the code (in the state machine for example), I need the latest value of a parameter but I don't want to have to buffer up a bunch of messages and then throw them all away just to get the latest value. I realized a double-buffer scheme wouldn't work since there is a case where the writer can be blocked so I implemented a triple-buffer scheme which always has a place for the writer to store new data. Afterwards, I realized that this is already a solved problem in the computer graphics area. For the ground software, more functionality was added to support the static test. Each numeric indicator has an optional set of limits with green, yellow, or red that corresponds to normal, warning, and error values for a parameter. Additionally, if a value goes stale because of loss of data from the vehicle, a red X will be displayed over it to indicate the value is unknown. The states from the state machine are shown in a table that tracks the progress of the countdown along with any trigger events and associated reasons for abnormal transitions. More features will be added later to support the flight (vehicle position/velocity, map location, etc.)


2016-09-16

The final cold tests with water and LN2 prior to the hot fire were completed this past weekend. I had several goals for the testing:

  • LOX fill technique - During the previous tests, I was unable to completely fill the LOX tank. For these tests, I used a modified fill technique which involved closing the LOX vent valve first, waiting a few seconds, then closing the LOX fill valve. This allows the 20 psi or so in the dewar to compress the ullage in the tank and get some more liquid in. I also added some temporary foam insulation to the tank and fill lines. As a result, the LOX tank pressurized in less than a second instead of the almost two seconds previously.
  • Sticky LOX vent - I wrapped a polyethylene bag around the LOX vent valve with cutouts for the plumbing and supplied low pressure N2 through a tube to displace the ambient air. This prevented moisture from condensing on the body of the valve and freezing it up on subsequent runs. The N2 flow rate is pretty low and in one case, the regulator shut off so I may want to get a lower pressure regulator for this application.
  • LOX fill valve leaks - The particular valve I'm using is the old main LOX valve and it has a provision for a pressurized seal cavity. I ran a 1/8 inch line from the LOX burst disc manifold and this eliminated any external liquid leaks. There is still a small gas leak when pressurized but it's good enough for now.
  • Regulator setpoints - I adjusted the static pressure setpoints to 550 psi for the fuel and 395 for the LOX to provide runtime pressures of 495 and 340 psi respectively.
  • Minimum helium pressure for run - During the previous set of runs, the starting helium tank pressure was too low to provide enough run time. I ordered a 3500 psi helium cylinder for this set of tests since the standard tanks are only about 2200 psi. Using the new LOX fill technique along with a helium fill pressure of around 2300-2500 psi provides about 17 seconds of runtime which is plenty for the first flight. One issue is that that the 3500 psi cylinder uses a CGA 680 fitting but I was able to get one in time for the test.
  • Self-pressurization of LOX tank - Even though I'm using helium for pressurization, I wanted to see how letting the LOX tank self-pressurize for a bit prior to the run would work. With the vent closed, the pressure rose from ambient to 370 psi in 5 minutes. I started the cold test at that point since I didn't want the pressure higher than the regulator static setpoint of 395 psi. It seemed to work without any issues but I can't really draw any conclusions since I didn't test the opposite condition of letting it vent for 5 minutes prior to a run.
  • New fuel fill technique - I machined a -6 boss port into the curved part of the fuel tank near the neck to allow filling it from the top. I was unable to find a suitable metering hand pump for a reasonable price so I gave up on the idea of filling the fuel tank through the drain QD fitting.
I also had a few problems during the test:
  • Clogged LOX fill line - After the last test, I carefully sealed up the ends of all my tubes to keep dirt and critters out but I missed one and as luck would have it, that's the one tube that some mud daubers decided to build a nest in. Since it was downstream of the external LOX fill filter, I now have to clean out all the LOX plumbing prior to the hot fire.
  • Frozen fuel line - During one of the six cold tests, the water in a portion of the fuel line near the bottom of the LOX tank froze up. This particular 3/8 inch tube is located about a half inch from the LOX tank so I was worried about this possibility but I believe the problem was caused by not cleaning off the external condensation from a previous run. Also, the temporary rubber pipe insulation was too thick to fit between the tube and LOX tank. Kerosene freezes at a much lower temperature (-40 deg) so that should help but I plan to put insulation between them for the flight vehicle.
  • Helium transducer range - I forgot to increase the range for the helium transducer and it pegged the ADC just above 2600 psi. Since 2500 psi proved to be enough for testing, I didn't bother to bump up the range during the tests but I plan to increase it to at least 4000 psi (the burst disc value) for subsequent tests.
Next up is a static hot fire of the rocket as soon as I can get ready for it. I need to figure out how to hold the rocket vertically in my test pit and I also need to fabricate a flame deflector since the exhaust will be pointing down. Ideas include angled concrete pavers, water soaked plywood, and curved steel plates.

2016-06-22

I ran some cold tests on my rocket this weekend in preparation for a hot-fire later this summer or fall. The main purpose of the tests was to check out the vehicle plumbing with LN2 and water along with using the real flight computer instead of the data and control system I used for my static tests. Overall things went pretty well but there are enough things to follow up on that I'd like to repeat the tests.

Things that worked:
  • No leaks at the bottom or top of the LOX tank - For the bottom boss port, I wanted to use a -6 Harrison K-Seal but none of the vendors were interested in small quantity sales. Instead, I used an AN901-6A aluminum crush washer which seemed to seal without leaks. For the top of the tank, these E-sized cylinders use a 0.750-16 UNF thread that matches a -8 AN fitting so I turned down some of the hex portion of an AN919-12D fitting and used a PTFE washer. I used the same fitting on the helium tank and didn't observe any leaks there either.
  • Helium fill valve - I'm using an AN6287-1 landing gear strut Schrader valve on the vehicle and a Schrader 556 fill valve. Both worked great once I figured out how to use it. With the 5/8 swivel nut on the valve, you open it a couple of turns but the tank only fills up to about 1000 psi. It freely spins a couple more turns, then you have to use a wrench to turn it another quarter turn or so to get it to open completely. I plan to use a 3500 psi helium cylinder next time since the standard one is only about 2250 psi when full.
  • New main LOX valve - This an update to my existing modified Swagelok SS-62T6 ball valve. It has two bearings (properly preloaded this time) and a single PTFE spring energized shaft seal. I initially used a vented ball and it leaked a little at the stem after the first actuation, just a hiss with no liquid. I changed it to a non-vented ball and it worked good for several runs and actuations. After several actuation on a given run, you could detect a very slight downstream gas leak. You couldn't hear it but if you put your finger over the downstream orifice I was using to simulate the engine, you could feel a bit of pressure. So despite my best efforts to eliminate shaft runout, those seals just don't work perfectly. I could have reconfigured it to use a pressurized seal cavity as I did previously but I'm happy with the performance as-is.
  • Flight computer - No problems even though it was over 95 deg F outside and powered on all day. The power supply heatsink was warm enough that I may want to tie it into the aluminum airframe to give it some more margin. The Digi RF modem was installed but not running so I expect that will increase the heat load on the power supply when I start using it. The 5V and 3.3V regulators on my power supply board are switchers so they should be pretty efficient.
  • New ground control S/W - I'm sending messages over UDP back between the flight and ground computers for commands and data. The ground S/W has a graphical representation of the onboard tanks and valves with custom Qt controls to indicate valve position. It's still a work in progress as I haven't added the RF telemetry capability to it yet.
  • Igniter GOX heat exchanger and valve - There is a tap off of the main LOX valve inlet and a 1/8 inch aluminum coiled tube heat exchanger to get GOX for the igniter. I didn't expect any surprises as I used a similar arrangement on my last static test.
  • Fuel fill/drain valve - I used a SnapTite quick disconnect fitting to fill and drain the fuel tank. As long as you're forceful and quick with the mating and de-mating, it only leaks a drop or so. If you fumble around with it (and the water hose is still turned on), it makes a big mess and leaks everywhere. I still need to find a hand pump that I can use to fill the fuel tank.
  • Aeroquip fittings - I was using several of these that had machined female AN ends and I was worried they might leak without a Del or Seco seal but I didn't notice any leaks.
Things that didn't work:
  • Following the procedures every time - A few times I tried to do some operations from memory during the helium and LN2 loading sequences and messed it up. This is really important when working with energetic systems and was a good reminder.
  • Filling the LOX tank - Apparently I didn't get the LOX tank completely full with LN2. Two indications were readily evident when looking at the pressure data recorded during the runs: 1) the helium tank pressure dropped way more then expected (1960 psig down to 1580 for example), and 2) the LOX tank took almost 2 seconds to pressurize. Ben Brockert provided some suggestions for improving the fill process:
    • Use insulation on all tanks and fill lines
    • Close the vent valve prior to closing the dewar valve to allow the dewar pressure to take up some of the tank ullage
    • Slow down the LOX flow during topping
    • Increase the size of the LOX vent if possible
  • LOX fill valve - The semi-custom ball valve isn't working out too well for this one. I think it's exposed to cold for too long during the fill operation to seal reliably. Using the new design that I tried for the main valve might work but I'm thinking a different stem seal is needed. Or, maybe a globe style valve might work better. Since it's a manual valve, there's no need for need for a quarter turn solution other than a desire for a large opening to help with the fill rate.
  • LOX vent valve - When closing it on a few runs, the shaft got to about 85 degrees closed and then stopped. That was enough but I think I can fix that with some servo tuning parameters. As long as I gave it enough time to dry out between runs it worked fine. However, on one back to back run, there was a lot of condensed moisture on the valve body and during the next fill, the water to ice and froze up the bearings. I am purposely using open bearings because I wanted to be able to clean out the grease for LOX use. Instead of taking the time to troubleshoot it properly, I had "get the test done itis" and tried to turn it with a set of pliers. The result was that I sheared off the setscrew. If I had just removed the setscrew, I could have easily determined whether it was the servo or valve that was stuck.
  • Transducer offset - One of my surplus pressure transducers had an interesting offset in the bridge that I didn't notice during calibration. The transducer works perfectly except the the balance is so far off that it exceeded the common mode range of my input amplifier (INA125 in single-ended mode) and I didn't design in a bridge balance or offset capability. Every other one I've used was very close to 0 mV output at ambient. I was able to swap it out with one from my test stand.
  • Aluminum NPT union leak - I was initially testing the tanks with an external regulator and as part of the setup, I was using an AN910-2D fitting I got from Aircraft Spruce. The tolerances on all of them seemed to be a little off because you could only thread in an AN816-6D fitting a couple of turns. I wonder if the anodizing on both parts messed with the dimensions too much. Anyway, I put two turns of tape on it and cranked it until I thought it was good. Well, after using it for a few minutes, I heard a loud hiss coming from the general area of that fitting and the first thing I thought of was a crack that was about to burst. I was able to depressurize without incident and I switched to a brass NPT union instead.
  • Valve orientation during assembly - Despite what I thought was a careful assembly procedure, I managed to get the orientation off by 90 deg on the fuel vent which I temporarily corrected in S/W.
  • Forgetting to turn off the lawn sprinklers - My wife woke me up early Saturday morning and said "Did you intend to water your rocket?" The plumbing was fine but I foolishly powered everything up before disassembling the servos to clean out any water and one failed to work. I replaced it but it eventually recovered a few hours later.
Interesting observations:
  • The helium tank got fairly hot when I was filling it. I could still touch it but I wonder if I was filling it too quickly. This tank is rated to 2015 psi, is normally hydro tested to 3358, and burst tested to 5037 but I was still slightly worried about being next to a pressure vessel with weaker material properties than expected due to temperature. I plan to fill it to around 3000 psi for the flight.
  • Pressure droop during run - Each of the Aqua 1247 regulators drooped about 50 psi when flowing, more than I expected. It shouldn't be a problem as I just need to add that much on to the initial setpoint. The pressure falloff as a function of supply pressure seemed reasonable (around 12 psi output change for 1000 psi of input change). When the helium pressure dropped below the setpoints, the output pressure actually rose a bit but not back to the original value (I'll post some plots on the website). I also noticed quite a bit of non-repeatability on the setpoint when initially setting the regulators, around 20 psi. It may have just been the gauge though because I didn't see it in the data during the runs.

There were some other minor issues with transducer gain and ADC ranges that I was able to resolve during the weekend. I plan to work a bit on the LOX fill valve and repeat the tests in a few weeks. I want to make sure that I have a reliable processes for filling, running, and de-tanking so things go smoothly on the hot fire. I also need to tweak the regulators and collect some data on the minimum helium pressure needed for a full run.


2016-04-10

I was invited to give a presentation on my progress at the Space Access '16 Conference this past week. A copy of the presentation is located in the Media section. It includes some block diagrams of the H/W and S/W aspects of the flight computer as well as the ground equipment configuration.

The main hardware components for the flight computer are complete - photos of the CPU and Signal Conditioning PCBs as well as the flight computer mounted in the rocket are up on the photos page. The remaining work includes:

  • Antenna brackets
  • Battery brackets
  • External power and debug connections
  • Transducer and servo wiring

After these items are complete, I can move on to proof tests, cold tests, and a hot fire test. I performed some preliminary EMI testing between the TM, GPS, and signal conditioning circuitry. As long as the TM antenna was at least a few inches away from the signal conditioning PCB, there was no measurable effect so it appears the filtering is working as intended.

As hardware is coming together, it seems like the pace of vehicle assembly is really picking up. Here is a photo of the overall vehicle progress so far (the nose cone is only temporarily mounted). There is a fair amount of work still to do for the S/W however. Most of the building blocks are in place so I'm going to incrementally add features as needed for the upcoming testing.

Instead of the five separate PCBs originally planned, I ended up with only three (Power Supply, CPU, and Signal Conditioning). A more distributed architecture is better for swapping out components and for testing but you also end up with more connections (that can fail) and a larger volume. After some thought, I decided to consolidate several functions onto the CPU PCB: the CPU, GPS, TM transceiver, IMU, digital pressure transducers, and the servo mux. It was a bit of a squeeze but I was able to get most of it on the front side of a 4x6 inch PCB with only the TM transceiver on the back. The signal conditioning PCB is the same as originally described. The Power Supply and Signal Conditioning PCBs are both two layers and the CPU PCB is four layers.


For archived articles, check out the News section.