Lessons from Final AIT

Final AIT (Assembly, Integration & Testing) refers to the stage of the satellite in which all the flight hardware is assembled, integrated together, and tested.

This is the final test of all the systems in terms of design & concept, and is the last stage for troubleshooting before everything is epoxied, vibration tested, and ultimately integrated into the rocket to be launched into orbit. This presents unique opportunities for last-minute troubleshooting. In our specific case, we noticed a few things pop up in the process:

• We had an antenna deployment cable get pinched between the frame wall & the rest of the frame/ADCS. This led to that channel immediately powering off as a preventative measure & an overcurrent fault. What this meant was that a wire was shorting either the ground or the frame.

  • We solved this by rerouting the cable such that it didn't come in contact with the frame.

• After the LEOPS (Launch and Early Operations) procedure, the antenna wouldn't deploy. This was traced back to a capacitor shorting with the frame's endplate. Had this been tested without being connected to the frame, it wouldn't have been detected - there was nothing within the documentation suggesting that this would've occurred. This highlights the importance of:

  • Realizing that documentation and reality are different. In our case, it was the orientation of the connector and the capacitor on the antenna that were different, meaning our models were incorrect.
  • You need to inspect and document these differences between the models & reality throughout an AIT campaign. This is done preferably by:
    • Measuring with detailed photos taken that have a ruler in frame - do this immediately when they are first received and have that saved in a folder on the shared GDrive that people know about. Then compare it to the CAD models by at least 2 people, and correct from there
    • Deprioritize simulations & models, especially thermal simulations - they're incredibly hard to do well, especially when the components don't perfectly match their CAD models & such. It is better to empirically test your hardware because of this, and speaks to what more proper project management could look like, with that time having been better spent doing practice CAD projects, starting cable harness models earlier, machine shop training, or real world thermal testing in this case.
  • That you need to be methodical and comprehensive in testing - that it was wise to test everything in the frame as if it'd be in orbit and on an individual-component-level, that you need to check if the components respond as expected before & after a test, and that this needs to be done slowly, with rest breaks in between tests.

• The Deployable Composite Lattice Boom needed its deployment mechanism troubleshooted due to concerns about its length and boom design mount being changed. The DCLB's length was changed by its researcher, and the deployment mechanism was changed after our failure at the vibration testing campaign we were part of at the ESA. These changes were designed to make it harder for the boom to release, which would've let us more easily reach our vibration goals then, but lead to the longer-length boom not being deployable.

  • We resolved this by adding a nitinol leaf spring and taking careful measurements on how far we stowed the boom into the mechanism. If this were to fail in-orbit, we've also done some calculations on what spinning the satellite to deploy it might look like, proving that it made for a feasible alternative.

• While assembling the solar panels onto the frame, we found that it might interfere with both the MPI's measurements & be unable to be flipped to fix it without damaging either the solar panels or the frame. Due to these concerns; one rendering the mission scientifically irrelevant and the other leading to concerns about the vibrational qualities of the satellite, a bit of troubleshooting was done. A few options were considered:

  • Addition of washers as spacers under the panels to raise them 1-2mm, which would allow for the solar panels' passives & connectors to not interfere with anything. This had the downsides of vibe testing complications, and wouldn't allow for the panels to be flipped to ground on the MPI side.
  • Water jet or laser cut spacer brackets similar to the ones used between our in-house solar panels & the MPI, 1.5mm (1/16") thick. This didn't allow for the flipping of the panels but did have less vibe risk & would be easier than assembling option 1.
  • Option 2 for the rear two COTS 2U panels, and a thicker 4.8mm (3/16") material for the front 2 COTS panels near the MPI. This would allow for the solar panel ground to be on the MPI side and lots of room for wires, but would take the longest & would throw off our CoG calculations & calibrations.

  These options were considered and then reconsidered as we discovered that our solar panels had the strange property of having the middle of the panels be where the highest & lowest potentials are. This meant that flipping the panels would have no difference on MPI effects, and so we were left worrying about spacing, specifically to fit small surface mount components that weren't in our provided CAD models. This wound up being resolved by going with option 2; 1.5mm (1/16") thick spacer brackets were manufactured to allow for the solar panels' surface components to fit, which would be placed on all sides to minimize changes to the CoG.

  To prevent this sort of change occurring so last minute, we've come to realize that we should've asked for better solar panel documentation earlier, especially about surface-mounted components & what the charges are at various points on the panels. This was also partially caused by the centralization of our CAD'ing capacity in a single member, whom was the only one with both experience and hardware powerful enough to manage the whole satellite, but that didn't have access to all the components - the team must have more distributed CAD'ing capacity to prevent such a bottleneck from leading to these last-minute changes, whether that be through training more people and/or running off a cloud-based collaborative CAD software is TBD.

• Construct the cable harnesses correctly i.e. put the crimps in the correct way. Backup cable harnesses are highly recommended to prevent needing to scramble for more harnesses in case of harness failure, assembly failure, etc.

• Most of the firmware should be done before assembly starts.

• The use of epoxy as conformal coating made our OBC look unprofessional - other options might've been buying clear conformal coating farther in advance (it tends to have long lead times) or to use lead solder (which also helps prevent tin whiskers).

• A combination of ultrasonic cleaner & IPA was used to clean flight hardware. The ultrasonic cleaner was good for electrical components but it needs water soluble flux and its water bath took a while to warm up which was inconvenient. A better method to clean mechanical parts should be considered, such as greasy parts that came from the machine shop, as those took a long time to clean.

• Cable harnesses need to be completed before assembly starts.

• Extra spacers came in clutch for the stack assembly as the components' reality diverged from their CAD.

• Do extensive safe-to-mates & cable harness checks. We had 3 different people check each one, which made sure they were all electrically correct.

• Have more people involved in assembly. We had 5 main people throughout the whole campaign, with a few more in the beginning. This is a lot of work to be done by 5 people!

• For any given component ordered, it's best to have the capacity to turn most things on/off by themselves, if possible.

• We should've used a different ADCS or find a way to utilize more of the stack pins as the ADCS we used needed a lot of the stack pins, leading to a lot of harnesses & more complicated cable routing because of that.

• To test the flight GNSS without it leaving the lab, we routed some coaxial cables out of the cleanroom to a window. This led to it getting 6 satellites in view with the window after a bit of troubleshooting.

• The Flight GNSS is a controlled good, which led to a few complications. It's a tradeoff, but we considered it worth it.

• At the ESA, we learned to take pictures often. However, we didn't start organizing them until halfway through. Make sure to start organizing (making a GDrive folder, labelling, etc) as soon as you take pictures, if not earlier.

• Keep a lab binder in the cleanroom. This led to some good documentation

• Space-grade heat shrink (PTFE) is really big & doesn't shrink very much. This makes it less secure, which was revolved by pinching the heated heat shrink with pliers to make it fit better.

• Though flight hardware is worn down throughout assembly steps, you still need to balance keeping flight hardware safe with adequate testing time & adequate assembly checks

• Stay calm & determined in troubleshooting. A level head & resilience will get you far when things don't go exactly as planned.

• Kapton tape should be avoided in areas of tight tolerance.

• Jank setups are not ideal for testing, but with enough though put into them, they can turn out very helpful.

• Set up a mock satellite assembly to practice cable routing with non-flight harnesses & cables. We wanted to do this, but hadn't budgeted enough time for it to be useful before assembly started.

• Consider what steps are irreversible or will make further assembly more difficult. E.g.) epoxying cable harnesses - though this wasn't as bad as we initially thought, this made it more difficult to move or access certain parts of the satellite.

• Ketchup cups & plastic shot glasses make great places to organize & hold fasteners, along with having a bit of IPA in to clean stuff.

• You ideally want to epoxy things only when you're very sure about where it will stay. It's difficult to remove, but not impossible.

• Expect the unexpected. When everything is put together, it won't be the same as you predict.

• Know when to step back & talk about things more before moving on. Know when to take a break. Ignore the desire to just see it progress - you need to be careful, you need to go slow.

• A laminar flow table with access from all sides or a rotating table to do assembly on without rotating the whole LFT would've been helpful.