Preliminary Design Review Notes
- Yevgeniy Gorbachev (RIT Student)
- Mary Dertinger (RIT Student)
- Louis Fleisher (RIT Student)
- Ben Hyman (RIT Student)
All errors courtesy of Yevgeniy.
VIDEO HERE: https://drive.google.com/file/d/1afZ8f8lS1JmH5o4gT8L7zVZnEJK4N3cJ/view?usp=sharing
Overview
The backplane is the first team-wide flight computer project. We intend to establish a standard for interchangeable modules that can be worked on independently, to more easily spread the load between members.
Notes
- Amend autopilot module to add state estimation
| Question | Answer |
|---|---|
Jim Heaney (RIT Student) is this going to be used for future rockets? | Ben Hyman (RIT Student) That is the intention. |
Backplane
Notes
- Modularity is violated by not providing its own power
- Not actually using Molex Picoblades for the flight implementation
- Might have to be very clever with positioning the connectors because DE-9s are big
- 8 GPIO for chutes is dumb because your two redundant computers can't be the same computer
- Power module gives you voltages - the connector slide should say 2x <Voltage 1>, 2x <Voltage 2>
- The managed version has SWD and JTAG broken out from the STM32
| Question | Answer |
|---|---|
Jim Heaney (RIT Student) What is the practical upper limit to the number of ports? | Billy Siegener (RIT Student) The upper limit will be driven mainly by the system's physical size (the mathematical limit is whatever Ethernet's is, which is enormous) |
John Ye (RIT Student) Is there a specific standard for each GPIO | Billy Siegener (RIT Student) No, e.g. if some module wants to use an external battery, it can |
Jim Heaney (RIT Student) Is there any sort of monitoring system on the SwitchBlox that resets in error state, or is it not a concern, being a COTS product? | Billy Siegener (RIT Student) No concern, it's COTS Will Merges (RIT Student) Probably already exists on the board |
Jim Heaney (RIT Student) What is management and PoE? | Billy Siegener (RIT Student) PoE is power over ethernet Will Merges (RIT Student) Unmanaged switch is a repeater - packets get blasted everywhere, but this switch also prevents collisions that would cause. Managed switch can direct ports to other ports. |
Ben Hyman (RIT Student) Why only 2 pins Will Merges (RIT Student) Would it be worth allocating the pins to address now? | Billy Siegener (RIT Student) yes we should, more backward-compatible |
Will Merges (RIT Student) why two pins per rail | Billy Siegener (RIT Student) 3A felt too restricting |
Will Merges (RIT Student) limitation of two vs four pair? | Billy Siegener (RIT Student) limits to 10 Mbps, no need for gigabit |
Jim Heaney (RIT Student) can the switch IC communicate positon to modules | will - not without the STM32, in which case we could use DHCP |
Billy Siegener (RIT Student) why is mr backpain on here | Ben Hyman (RIT Student) he's everywhere |
Jim Heaney (RIT Student) Will all these DE-9s come out the same slot? | Billy Siegener (RIT Student) Early talks had these doubled up top/bottom, but we only have 4 modules, so it makes sense to put them all in one spot |
Yevgeniy Gorbachev (RIT Student) Why 9 pins and not something more? | Billy Siegener (RIT Student) BD misleading - if all the pins are on the same connector, the connector will need a new wiring harness, since you need to swap pins if you swap module positions. With this, switching mod positions just requires switching the connector |
Will Merges (RIT Student) DB-9 vs DE-9 | Letter is shell size - DE-9 is correct and DB-9 is incorrect but frequently |
Jim Heaney (RIT Student) Big connector - too big for cube? | Billy Siegener (RIT Student) Yes, that's a tradeoff we're willing to accept. We'd need to be clever. |
Jim Heaney (RIT Student) Can you fit four DE-9s next to each other inside the diameter of a rocket? | Billy Siegener (RIT Student) Two by two, yes |
Jim Heaney (RIT Student) Highest I/O count, and is that enough for future proofing? | Ben Hyman (RIT Student) Autopilot is highest, it already has 4 for just the chutes, not counting control signals Billy Siegener (RIT Student) We think yes, it is enough. Also 8 pins are enough for an RS-485 transceiver if we really need more I/O |
Will Merges (RIT Student) Could one module request that another module perform actions over GPIO? | Billy Siegener (RIT Student) Yes, because we have a shared data bus and positions on the board are known to each module |
Jim Heaney (RIT Student) Why constant current drivers? | Billy Siegener (RIT Student) We do not know each rail's voltages in full generality because that's determined by the power module |
Ben Hyman (RIT Student) You noted through-hole LEDs | Billy Siegener (RIT Student) They're ubiquitous |
Jim Heaney (RIT Student) does having a standard connector for LEDs so that they are visible outside the rocket make sense? | Will Merges (RIT Student) No, status will be obvious without them - notable failures like no telemetry will be easily detectable, and cause a scrub and disassembly anyway Billy Siegener (RIT Student) Externally visible LEDs will not be externally visible for any reasonable distance; these LEDs are just nice for your lab bench |
John Ye (RIT Student) V1 by FDR - any changes foreseen by V2 after FDR? | Billy Siegener (RIT Student) V1 means the 2023 version, basically. V2 would be for a different vehicle or different year |
Alex Speyer (RIT Student) What improvements would be on V2 | Billy Siegener (RIT Student) More modules; Ethernet is arbitrarily expandable. You can actually just daisy-chain backplanes if you really wanted to. |
Autopilot
Notes
- NOT the brain board
- Deployment controlled by high-side switch; we know this circuit works because it was on DARWIN
- Jim Heaney (RIT Student) Might be worth looking into I2C hotplug ICs
- Billy Siegener (RIT Student) Offboard I2C is not generally a great idea
- Billy Siegener (RIT Student) None shall ask why this isn't the backplane already
- We want to go to a lab to vibe and shock test.
- Billy Siegener (RIT Student) We should have a way to disable LEDs we can't see during flight
Action items
- Ben Hyman (RIT Student) Implement continuity checks on the deployment channels
- Ben Hyman (RIT Student) Figure out data budget
- Ben Hyman (RIT Student) Subtract WiFi from power budget (spreadsheet linked in speaker notes)
| Question | Answer |
|---|---|
Will Merges (RIT Student) Have you allocated the specific control signals on the GPIO | Ben Hyman (RIT Student) No. Definitely will include PWM. |
Will Merges (RIT Student) Is there a requirement to do charges? | Ben Hyman (RIT Student) Yes, detailed in the B.D. |
Billy Siegener (RIT Student) How many ethernet lines does the vocore have? | Ben Hyman (RIT Student) Up to 5, but steals from SPI, I2C, PWM pins |
Will Merges (RIT Student) Why a high side, not a low-side switch | Ben Hyman (RIT Student) If you accidentally ground a low-side switch, current will flow and the e-matches will blow Billy Siegener (RIT Student) You are switching the high-side - your charge is tied low, and you pull it high. |
John Ye (RIT Student) Is there SPI and I2C on this MCU? | |
Jim Heaney (RIT Student) Are you breaking out SPI and I2C on the GPIO? | Ben Hyman (RIT Student) Not at the moment but it's possible |
Yevgeniy Gorbachev (RIT Student) How many cores? | |
John Ye (RIT Student) How much does this cost? | Ben Hyman (RIT Student) $17, but takes a while to ship (still quicker than SwitchBlox lmao) |
Will Merges (RIT Student) How does this have higher processing power with a lower clock rate? | Ben Hyman (RIT Student) Not true, this is like 3x faster clock rate |
Will Merges (RIT Student) Bad vibes on the SD card? | Ben Hyman (RIT Student) No thought given Jim Heaney (RIT Student) Automotive-grade slots are like $100 Will Merges (RIT Student) Just hot glue it in |
Jim Heaney (RIT Student) What are you recording and how fast? | Ben Hyman (RIT Student) Unkown - depends on software team. They might want to record major events, RW state, etc. |
@everyone What's with the aversion against brain board? |
Power
Notes
- Umbilical babbyyyy
- More clean power is more future proof
- But arguably modules that need above-average power should do it themselves
- Ben Hyman (RIT Student) DDPAK is not as easy as you think, but I did it on a stove
- James May (RIT Student) We have a toaster
- Radio might change the power budget, don't update the budget just yet
Action items
- James May (RIT Student) Figure out EMI mitigations
- James May (RIT Student) Update power calcs
- James May (RIT Student) Figure out power cleanliness requirements
- James May (RIT Student) Confirm voltage ripple number
- James May (RIT Student) Sim the switcher ripple in Ti WebBench (spice would be good as well)
- Everyone: Sit down and don't leave until we have a power budget
| Question | Answer |
|---|---|
Jim Heaney (RIT Student) Are you concerned at all about buck EMI, how are you mitigating? | James May (RIT Student) Yes, TBD - originally wanted to use integrated inductor, selected one does not have that though |
Ben Hyman (RIT Student) How are you picking your inductor? | James May (RIT Student) Billy Siegener (RIT Student) there are equations for that Jim Heaney (RIT Student) some inductor has an integrated case |
John Ye (RIT Student) Are the batteries included in the 3U requirement? | Billy Siegener (RIT Student) Yes Billy Siegener (RIT Student) 3U is a decent upper bound |
Jim Heaney (RIT Student) Any specific power dissipation considerations? | James May (RIT Student) Have not ~ Rationale for switcher through LDO is to minimize heat dissipation? |
Billy Siegener (RIT Student) Still have reservation about LDO to switcher because that gives you a fixed power dissipation, this probably results in throwing away power. Have you looked into other ways to reduce rail noise; is this scheme absolutely necessary? Yevgeniy Gorbachev (RIT Student) Why not leave it to the modules to filter the power? | James May (RIT Student) Not strictly necessary. We're looking for future proofing, so I want to send very clean power to future boards that might want it. James May (RIT Student) comes down to what each board needs Billy Siegener (RIT Student) We are using a lot of power already; we should not make it worse by cleaning up power in this way unnecessarily |
Jim Heaney (RIT Student) Does the datasheet give expected ripple? | James May (RIT Student) Yes, probably 100mV Billy Siegener (RIT Student) Depends on lots of stuff |
John Ye (RIT Student) Do you have plans for upgrading to support for higher voltage batteries? | James May (RIT Student) The switcher already can do 36V, the limiting factor is trace spacing |
Jim Heaney (RIT Student) Any plans for over voltage or reverse polarity protection? | |
Yevgeniy Gorbachev (RIT Student) Any mitigations for human error | James May (RIT Student) Most of that is battery protection Billy Siegener (RIT Student) There are ways to limit current but Don't be an idiot Jim Heaney (RIT Student) Sprinkle some polyfuses in there |
Will Merges (RIT Student) Where are we measuring current and voltage? | James May (RIT Student) Current and voltage sense off the primary switcher input |
Jim Heaney (RIT Student) How are you connecting the two batteries? | James May (RIT Student) Switch-over IC, TBD |
| John Ye (RIT Student) Any plans to measure output voltage? | Yevgeniy Gorbachev (RIT Student) Might want to, in case we need to investigate an in-flight failure James May (RIT Student) Jim Heaney (RIT Student) Will Merges (RIT Student) We won't see transients at our sample rate, and that same faulty power is going to power the logging circuit anyway |
Alex Speyer (RIT Student) What's a transient | James May (RIT Student) "Woop!" Ben Hyman (RIT Student) Current jump or dip for a very small period |
John Ye (RIT Student) Would it be a good idea to have the MCU be powered by a separate battery? | James May (RIT Student) Jim Heaney (RIT Student) Will Merges (RIT Student) Not really necessary, all this tells us is "we lost power" which we'll know anyway Jim Heaney (RIT Student) Each module should have a fuse or something Billy Siegener (RIT Student) Automotive fuses? Jim Heaney (RIT Student) No, polyfuses |
| Are we using local or synced clocks? |
Radio
Notes
- Will Merges (RIT Student) If we have to choose UART/SPI, use the slower one for the GPS and the faster one for the transmitter
- We hate U.FL connectors
- Will Merges (RIT Student) The radio module is a layer 4 switch, if you squint
Action items
- Dante Sivo (RIT Student) Explain why radio wants a more accurate clock
- Dante Sivo (RIT Student) Evaluate current driving potential of a GPIO out on this STM32 (Ben Hyman (RIT Student) answers 25 mA)
- Dante Sivo (RIT Student) Try to lock down numbers on things like precision and altitude lockout
- Dante Sivo (RIT Student) Get a more realistic power numbe
| Question | Answer |
|---|---|
Jim Heaney (RIT Student) Have you considered hand-assembly? | Dante Sivo (RIT Student) Yes, probably 30-40 components, having passives placed. Probably need to hand solder the big bois (XBee) |
Billy Siegener (RIT Student) Why do you want a "more accurate" clock? | Dante Sivo (RIT Student) We need to review; we should leave a footprint for a clock anyway |
Will Merges (RIT Student) What is the importance of not using UART? | Dante Sivo (RIT Student) I don't remember Will Merges (RIT Student) Because UART is slower than the radio |
Jim Heaney (RIT Student) How precise is this GPS module? | Dante Sivo (RIT Student) Unknown |
Jim Heaney (RIT Student) Alt or speed lockout? | Dante Sivo (RIT Student) Alt lockout |
Jim Heaney (RIT Student) Does this chip log # of satt? | |
Jim Heaney (RIT Student) Does 18 Hz seem too fast? | Will Merges (RIT Student) Good if it is, but not necessary |
Jim Heaney (RIT Student) Any other drawbacks to a soldered antenna, is the work to harden a soldered antenna better than a remote antenna? | Billy Siegener (RIT Student) Soldered antenna is not that beneficial |
Jim Heaney (RIT Student) Are you concerned about range drop-off on non-LoRa modules? | Dante Sivo (RIT Student) Yes, we want to use the 933 MHz |
Jim Heaney (RIT Student) Why is your MCU using 130 mA | Dante Sivo (RIT Student) Driving all GPIO |
Will Merges (RIT Student) What's a gain margin? | Billy Siegener (RIT Student) How much higher Rx sense is than Tx power |
Jim Heaney (RIT Student) What does gain margin give us practically? | Billy Siegener (RIT Student) Indicates signal strength - "can we hear the rocket?" |
Sensors
Notes
- Low-g orientation matters because it has a hard-coded offset
- BGA is Bad
- Hard to solder
- Hard to inspect
- Mechanically weak
- Don't need very fast RH or temperature
- If we change the temp sensor, choose a sensor that doesn't need a hole in the board
- Use the 500 dps setting
- Will has a repressed trauma involving RS-485
- RS-485 has a good tolerance on diffpair length, good because backplane doesn't make routing guarantees
- Filesystem not decided - Wear-leveling and specific FS are TODOs
Action items
- Louis Fleisher (RIT Student) Get low-g accel rating (Ben answers 5000 g)
- Confirm orientation dependency on low-g
- Figure out the full implications of orientation dependency on low-g
- Find a non-BGA chip (or package of the same)
- Figure out the pull-up resistance for the entire I2C bus
- Jim Heaney (RIT Student) Unify LEDs
- Figure out block size and write time, see if we can switch to the same chip everyone is using
| Question | Answer |
|---|---|
Yevgeniy Gorbachev (RIT Student) Does the alt measure pressure or altitude | Jim Heaney (RIT Student) Can be either, if you trust your MCU to do the math faster |
| Jim Heaney (RIT Student) What's the survivable G rating on the low-g? | Louis Fleisher (RIT Student) "Very likely" to be OK |
Jim Heaney (RIT Student) Are we likely to take advantage of dynamic range? | |
(Someone) Noise dependent on FS? | |
Jim Heaney (RIT Student) Does the accelerometer's orientation matter wrt Earth's gravity? | |
Yevgeniy Gorbachev (RIT Student) Any plan to use a DRDY interrupt (low-g)? Will Merges (RIT Student) Is there one? | Jim Heaney (RIT Student) Has one |
Jim Heaney (RIT Student) What's the data rate of the temperature sensor? | Louis Fleisher (RIT Student) 1/8 Hz RH, 1/5 Hz temperature |
Jim Heaney (RIT Student) 200 dps is near the upper limit - can you switch in-flight | Louis Fleisher (RIT Student) No, we should use the 500 dps setting |
Billy Siegener (RIT Student) Why is Jim's heart sideways | Mary Dertinger (RIT Student) Because otherwise his face would be sideways |
Jim Heaney (RIT Student) Why resistors on spi lines? | Louis Fleisher (RIT Student) Absorbs reflections |
Ben Hyman (RIT Student) Why not CAN? | Jim Heaney (RIT Student) RS-485 is cheaper and easier, and I don't know how to implement CAN |
Billy Siegener (RIT Student) Why not UART or other stuff? | Jim Heaney (RIT Student) RS-485 is resistant to noise, we don't need massive data throughput |
Will Merges (RIT Student) This is outside the original requirements, why? | Jim Heaney (RIT Student) This was brought up in a brainstorming session - e.g. external temperature monitoring would be really nice, and this future-proofs the board further. Very cheap for the future proofing. |
Jim Heaney (RIT Student) Why does the programming interface not draw power? | Mary Dertinger (RIT Student) Powered by USB |
Billy Siegener (RIT Student) How is this calculated? | Jim Heaney (RIT Student) Number of bits sent over I2C, for data rates in each flight phase |
Yevgeniy Gorbachev (RIT Student) Planning to change data rate throughout the flight? | Louis Fleisher (RIT Student) We can, but probably better to reduce that to "flying" and "not flying" |
Jim Heaney (RIT Student) What's the driver on data rates? | Will Merges (RIT Student) Very arbitrary - 100 "sounds good enough" |